*Review* **Fourier Transform Infrared (FTIR) Spectroscopy to Analyse Human Blood over the Last 20 Years: A Review towards Lab-on-a-Chip Devices**

**Ahmed Fadlelmoula 1,2, Diana Pinho <sup>3</sup> , Vitor Hugo Carvalho 4,5 , Susana O. Catarino 1,2 and Graça Minas 1,2,\***


**Abstract:** Since microorganisms are evolving rapidly, there is a growing need for a new, fast, and precise technique to analyse blood samples and distinguish healthy from pathological samples. Fourier Transform Infrared (FTIR) spectroscopy can provide information related to the biochemical composition and how it changes when a pathological state arises. FTIR spectroscopy has undergone rapid development over the last decades with a promise of easier, faster, and more impartial diagnoses within the biomedical field. However, thus far only a limited number of studies have addressed the use of FTIR spectroscopy in this field. This paper describes the main concepts related to FTIR and presents the latest research focusing on FTIR spectroscopy technology and its integration in lab-on-a-chip devices and their applications in the biological field. This review presents the potential use of FTIR to distinguish between healthy and pathological samples, with examples of early cancer detection, human immunodeficiency virus (HIV) detection, and routine blood analysis, among others. Finally, the study also reflects on the features of FTIR technology that can be applied in a lab-on-achip format and further developed for small healthcare devices that can be used for point-of-care monitoring purposes. To the best of the authors' knowledge, no other published study has reviewed these topics. Therefore, this analysis and its results will fill this research gap.

**Keywords:** blood cells; fourier transform infrared (FTIR) spectroscopy; functional group; lab-on-a-chip

#### **1. Introduction**

Millions of blood test analyses are performed every day worldwide in order to provide blood diagnostic services for the patients [1]. Usually, these tests are performed in clinical laboratories, simultaneously using different devices and relying on different specialties [2]. These devices are needed to run routine blood tests [2] and examine multiple parameters to assist the physicians in haematology-, chemistry-, and immunology-related diagnosis, among others. They require human resources, dedicated facilities, and time, which, in an ideal device, should be less than one hour from taking a sample to printing out the results [3,4]. Moreover, the reagents needed to run all these tests are expensive, and most of them are toxic, having a significant direct and indirect effect on the environment [5].

Diagnostic devices currently available on the market rely on the same measuring techniques developed in the last century (mainly spectrophotometry or electrochemical assays) [6]. Meanwhile, viruses, bacteria, and fungi are rapidly evolving [7], pushing further the need to develop new, quick, and reliable diagnostic tools. The primary, commercially available measuring techniques for such devices are spectrophotometry, enzyme-linked

**Citation:** Fadlelmoula, A.; Pinho, D.; Carvalho, V.H.; Catarino, S.O.; Minas, G. Fourier Transform Infrared (FTIR) Spectroscopy to Analyse Human Blood over the Last 20 Years: A Review towards Lab-on-a-Chip Devices. *Micromachines* **2022**, *13*, 187. https://doi.org/10.3390/ mi13020187

Academic Editor: Nam-Trung Nguyen

Received: 31 December 2021 Accepted: 24 January 2022 Published: 26 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

immunosorbent assay (ELISA), electrophoresis, and blood cell counting or complete blood count (CBC). However, all of these methods have limitations. In ultraviolet–visible (UV– VIS) spectrophotometry, the main limitation is the requirement for sample and setup preparation time to avoid light interferences [8]. ELISA limitations are related to the cost of the assays due to the use of antibodies, the risk of cross-reactivity, the high background noise, and extended analysis time [9]. Electrophoresis requires a large sample for the assays, as well as high analysis precision [10]. Finally, CBC limitations are related to the manual examination of blood smears, difficulty recognising abnormal red cell shapes (such as fragmented cells), and high running costs [11]. Hence, the pressing need for new, fast, and precise analysing techniques.

Fourier transform infrared (FTIR) spectroscopy is a field that has undergone significant development over the past decade, promising easier, more rapid, and more objective diagnoses [12,13]. FTIR spectroscopy studies the interactions between matter and electromagnetic radiation that appear in the form of a spectrum. Each molecule has a spectrum fingerprint that makes it unique and allows it to be distinguished from other molecules [14]. FTIR spectroscopy is also an effective and nondestructive method for monitoring cellular alterations [15,16]. FTIR spectral analysis has allowed the characterisation of several organs' diseases, as well as the quantification of different biomolecules such as proteins [16], nucleic acids [17,18], and lipids [19]. Several research documents highlighting the importance of spectroscopic techniques in cancer detection have been published in the literature [15,20,21]. FTIR focuses on the differentiation and characterisation of cells and tissues by looking at individual bands or groups to precisely identify the molecular conformations, bonding types, functional groups, and intermolecular interactions that compose the specimen [13,20]. Thus, this paper describes the main concepts and terminologies related to FTIR and presents the latest published research focusing on FTIR spectroscopy technology and its integration in lab-on-a-chip devices for application in the biological field. To the best of the authors' knowledge, no other studies have reviewed these topics, making this review the first to fill this research gap.

The paper is organised into five sections. Section 1 presents an introductory overview and the primary motivation guiding the study. Section 2 presents FTIR spectroscopy's theoretical, conceptual elements and clarifies the salient terminology, including the concepts of infrared (IR) regions, radiation, molecular vibration, FTIR, and Michelson Interferometer. Section 3 describes the methods used in the presented study, while Section 4 offers the results of the analysis of twenty-year-long research on the application of FTIR spectroscopy in the biological field, focusing on the possibility of applying this technology in lab-on-achip devices. Finally, Section 5 presents the conclusion and future trends.

#### **2. Theoretical Considerations**

Here the terminologies and concepts associated with FTIR spectroscopy, namely the IR region, IR radiation and molecular vibrations in biological matters, FTIR techniques and Michelson interferometer, are presented.

#### *2.1. Infrared Region*

IR radiation is a group of electromagnetic waves (EMR) with wavelengths longer than visible radiation, invisible to the human eye. The IR region of the electromagnetic spectrum ranges in wavelengths from 0.8–100 µm, illustrated in Table 1 [22,23]. Typically, the IR is broken into three ranges, near-IR, mid-IR, and far-IR. Most of the IR used in medical applications are in the mid-IR range, considering radiation from the electromagnetic spectrum, in the wavenumber interval from 4000 cm−<sup>1</sup> to 400 cm−<sup>1</sup> . The frequency of the absorbed radiation is responsible for each subatomic vibrational interaction, as schematised in Figure 1.


the absorbed radiation is responsible for each subatomic vibrational interaction, as sche-

matised in Figure 1.

*Micromachines* **2022**, *13*, x FOR PEER REVIEW 3 of 20

**Figure 1.** Scheme of the optical spectrum, focusing on the infrared region. Reprinted from [23], MDPI, under a Creative Commons Attribution (CC BY) license. **Figure 1.** Scheme of the optical spectrum, focusing on the infrared region. Reprinted from [23], MDPI, under a Creative Commons Attribution (CC BY) license.

#### *2.2. IR Radiation and Molecular Vibrations in Biological Matter 2.2. IR Radiation and Molecular Vibrations in Biological Matter*

As a type of electromagnetic wave, IR propagates energy and momentum, with properties similar to both a wave and a particle—the photon. IR is emitted or absorbed by molecules as they change their rotational, vibrational motions. It excites wave modes in a molecule by changing them instantaneously, making it a helpful frequency variation for studying the molecular energy states with correct symmetry. Therefore, IR chemical analysis studies the absorption and transmission of photons in the IR region [24]. The IR spectrum of biological samples consists of a combination of the characteristic absorption bands of proteins, lipids, nucleic acids, and carbohydrates within that sample [25,26]. As a type of electromagnetic wave, IR propagates energy and momentum, with properties similar to both a wave and a particle—the photon. IR is emitted or absorbed by molecules as they change their rotational, vibrational motions. It excites wave modes in a molecule by changing them instantaneously, making it a helpful frequency variation for studying the molecular energy states with correct symmetry. Therefore, IR chemical analysis studies the absorption and transmission of photons in the IR region [24]. The IR spectrum of biological samples consists of a combination of the characteristic absorption bands of proteins, lipids, nucleic acids, and carbohydrates within that sample [25,26].

The protein absorption bands are often assigned to amino acid side groups or peptide backbone in the 1700 cm−1–1500 cm−1 range. The vibrational modes of the peptide backbone produce the amide I and II bands. The amide I band (1700 cm−1–1600 cm−1) is mainly associated with the bending vibration of the N–H bond. The bands of amides I and II are usually used to analyse the secondary protein structure [27]. The presence of absorption bands at 1450 cm−1 and 1400 cm−1 is due to asymmetric and symmetric methyl bending modes, respectively [28]. The protein absorption bands are often assigned to amino acid side groups or peptide backbone in the 1700 cm−1–1500 cm−<sup>1</sup> range. The vibrational modes of the peptide backbone produce the amide I and II bands. The amide I band (1700 cm−1–1600 cm−<sup>1</sup> ) is mainly associated with the bending vibration of the N–H bond. The bands of amides I and II are usually used to analyse the secondary protein structure [27]. The presence of absorption bands at 1450 cm−<sup>1</sup> and 1400 cm−<sup>1</sup> is due to asymmetric and symmetric methyl bending modes, respectively [28].

In the spectra of lipids, absorption bands are found in numerous spectral regions: the range of 3050 cm−1–2800 cm−1 for asymmetric and symmetric stretching vibrations of -CH2 and -CH3, the range of 1500 cm−1–1350 cm−1 for deformation vibrations of -CH2 and -CH3 from the acyl chains of lipids, the range of 1745 cm−1–1725 cm−1 for symmetric stretching vibrations of ester–carbonyl bond (C=O), and the range of 1270 cm−1–1000 cm−1 for odd (1240 cm−1) and symmetric (1080 cm−1) vibrations of -PO2- in phospholipids [29]. The IR spectra of nucleic acids are characterised in four spectral regions: the region of 1780 cm−1– 1550 cm−1 for in-plane vibrations of double bonds of bases, the region of 1550 cm−1–1270 In the spectra of lipids, absorption bands are found in numerous spectral regions: the range of 3050 cm−1–2800 cm−<sup>1</sup> for asymmetric and symmetric stretching vibrations of -CH2 and -CH3, the range of 1500 cm−1–1350 cm−<sup>1</sup> for deformation vibrations of -CH2 and -CH3 from the acyl chains of lipids, the range of 1745 cm−1–1725 cm−<sup>1</sup> for symmetric stretching vibrations of ester–carbonyl bond (C=O), and the range of 1270 cm−1–1000 cm−<sup>1</sup> for odd (1240 cm−<sup>1</sup> ) and symmetric (1080 cm−<sup>1</sup> ) vibrations of -PO2- in phospholipids [29]. The IR spectra of nucleic acids are characterised in four spectral regions: the region of 1780 cm−1–1550 cm−<sup>1</sup> for in-plane vibrations of double bonds of bases, the region of 1550 cm−1–1270 cm−<sup>1</sup> for the deformation vibrations of bases that include the sugar vibrations, the region of 1270 cm−1–1000 cm−<sup>1</sup> for vibrations of -PO2- and, finally, the region of 1000 cm−1–780 cm−<sup>1</sup> for the vibrations of the sugar-phosphate backbone [30]. The carbohy-

drate spectra contain bands in the following ranges: the region of 3600 cm−1–3050 cm−<sup>1</sup> is assigned to the stretching vibration of O-H, the range of 3050 cm−1–2800 cm−<sup>1</sup> to the stretching vibrations of -CH<sup>3</sup> and -CH2, the region of 1200 cm−1–800 cm−<sup>1</sup> to the stretching vibrations of the C-O/C-C species, and, finally, the 1500 cm−1–1200 cm−<sup>1</sup> relates to the deformational modes of the CH3/CH<sup>2</sup> species [31]. In the blood analysis applications, the spectral bands of 3000 cm−1–2800 cm−<sup>1</sup> are the most relevant ones for analysing red blood cells and platelets, while for the white blood cells, the most relevant band ranges are 513 cm−1–1445 cm−<sup>1</sup> . Thus, those targeted cells and corresponding spectral bands are the most used for blood analysis, particularly lab-on-a-chip applications. stretching vibration of O-H, the range of 3050 cm−1–2800 cm−1 to the stretching vibrations of -CH3 and -CH2, the region of 1200 cm−1–800 cm−1 to the stretching vibrations of the C-O/C-C species, and, finally, the 1500 cm−1–1200 cm−1 relates to the deformational modes of the CH3/CH2 species [31]. In the blood analysis applications, the spectral bands of 3000 cm−1–2800 cm−1 are the most relevant ones for analysing red blood cells and platelets, while for the white blood cells, the most relevant band ranges are 513 cm−1–1445 cm−1. Thus, those targeted cells and corresponding spectral bands are the most used for blood analysis, particularly lab-on-a-chip applications.

cm−1 for the deformation vibrations of bases that include the sugar vibrations, the region of 1270 cm−1–1000 cm−1 for vibrations of -PO2- and, finally, the region of 1000 cm−1–780 cm−1 for the vibrations of the sugar-phosphate backbone [30]. The carbohydrate spectra contain bands in the following ranges: the region of 3600 cm−1–3050 cm−1 is assigned to the

#### *2.3. Fourier Transform Infrared Spectroscopy (FTIR) Techniques 2.3. Fourier Transform Infrared Spectroscopy (FTIR) Techniques*  FTIR spectroscopy is a technique used to obtain the absorption or emission infrared

*Micromachines* **2022**, *13*, x FOR PEER REVIEW 4 of 20

FTIR spectroscopy is a technique used to obtain the absorption or emission infrared spectrum of a solid, liquid, or gas [14,32]. The FTIR spectrometer simultaneously collects high-resolution information over a wide spectral range (between 4000 and 400 cm−<sup>1</sup> ), a distinct advantage over a dispersive spectrometer, which estimates power over a narrow range of frequencies at once. The aim of spectroscopy techniques (FTIR or bright perceptible (UV–Vis) spectroscopy) is to quantify how much light a sample absorbs at each frequency [14]. The most direct approach, the "dispersive spectroscopy" method, consists of focusing a monochromatic light beam at a sample, measuring the amount of absorbed light, and recalculating it for each frequency [14]. Fourier transform spectroscopy is a less instinctive approach for obtaining similar data. Rather than focusing a monochromatic (single frequency) light emission at the sample, this strategy might focus a bar, or array, which contains numerous frequencies of light at once and measures how much of that beam is absorbed by the sample. Then, the wave is changed to contain a different mixture of frequencies giving a second data point. This cycle is repeated many times within a short period of time, and the information is acquired by a computer. For instance, the wave plotted in Figure 2, called an interferogram, is created by applying a broadband light source-one that contains the entire range of frequencies to be estimated. The light sparks into a Michelson interferometer (detailed in the next section), consisting of a special array of mirrors, one of which is moved by a motor. As this mirror moves, each light frequency in the column is occasionally obstructed, mediated, impeded, and transmitted by the interferometer. The different frequencies are tweaked at different rates so that the column exiting the interferometer has a different range at each second or mirror position [14,32]. spectrum of a solid, liquid, or gas [14,32]. The FTIR spectrometer simultaneously collects high-resolution information over a wide spectral range (between 4000 and 400 cm−1), a distinct advantage over a dispersive spectrometer, which estimates power over a narrow range of frequencies at once. The aim of spectroscopy techniques (FTIR or bright perceptible (UV–Vis) spectroscopy) is to quantify how much light a sample absorbs at each frequency [14]. The most direct approach, the "dispersive spectroscopy" method, consists of focusing a monochromatic light beam at a sample, measuring the amount of absorbed light, and recalculating it for each frequency [14]. Fourier transform spectroscopy is a less instinctive approach for obtaining similar data. Rather than focusing a monochromatic (single frequency) light emission at the sample, this strategy might focus a bar, or array, which contains numerous frequencies of light at once and measures how much of that beam is absorbed by the sample. Then, the wave is changed to contain a different mixture of frequencies giving a second data point. This cycle is repeated many times within a short period of time, and the information is acquired by a computer. For instance, the wave plotted in Figure 2, called an interferogram, is created by applying a broadband light source-one that contains the entire range of frequencies to be estimated. The light sparks into a Michelson interferometer (detailed in the next section), consisting of a special array of mirrors, one of which is moved by a motor. As this mirror moves, each light frequency in the column is occasionally obstructed, mediated, impeded, and transmitted by the interferometer. The different frequencies are tweaked at different rates so that the column exiting the interferometer has a different range at each second or mirror position [14,32].

**Figure 2.** Example of a general FTIR interferogram. The central peak is positioned at the ZPD position (zero path difference or zero retardation), where the maximal amount of light passes through the interferometer to the detector. **Figure 2.** Example of a general FTIR interferogram. The central peak is positioned at the ZPD position (zero path difference or zero retardation), where the maximal amount of light passes through the interferometer to the detector.

Computational postprocessing based on Fourier frequencies is required to calculate the results (light pickup for each frequency) from the coarse raw information (light pickup for each mirror position), as presented in the example of Figure 2 [14,32]. Then, the Fourier transform converts a space (for this situation, the mirror's distance in cm) into its opposite space (wavenumbers in cm−<sup>1</sup> ).

The main limitations of FTIR spectroscopy relate to the tissue depth penetration of the infrared light, which only allows biochemical analysis of the tissues up to a few dozens of micrometres [20]. Additionally, in the conventional FTIR spectroscopy, which works in transmission mode and consequently with no incidence angle between emitter and sample, there is difficulty in assuring the reproducibility of the spacer thickness when using liquid samples [33].

The attenuated total reflectance Fourier transform (ATR -FTIR) technique as a complementary technique has helped FTIR spectroscopy overcome this limitation. ATR-FTIR is a particular FTIR spectroscopy method, which measures the reflected signal from a sample. In this reflectance setup, the IR radiation passes through a crystal with a high refractive index (typically with an angle of 45◦ ) and undergoes total internal reflection before exiting the crystal and being directed to an IR detector [33,34]. ATR-FTIR has a lower penetration depth than conventional FTIR (around 200 nm) but, since it measures the reflected light, it is an adequate method for measuring high absorbing and high thickness samples that typically do not allow the transmission of IR radiation [33]. Additionally, this technique can direct measurements of gas, fluidic and thin-film solid-state samples without complex sample preparation and with enhanced surface sensitivity [33,34].

Finally, microscopic FTIR (micro-FTIR) [35] relates to another particular FTIR technique that couples an IR spectrometer to a visible light microscope in order to achieve better sensitivity when detecting condensed-phase compounds [36] and is adequate for measuring solid or liquid thin films samples. In this technique, the microscope focuses the IR laser beam on the sample, and the measurement mainly comes from the target focal point, meaning that even a short displacement in the laser beam or the sample could provide a significant difference in the results. Therefore, micro-FTIR distinguishes by allowing local measuring of a particular point in the sample, while the conventional FTIR gives the average information from a complete homogenised sample [35,36].

Besides helping to identify organic compounds based on their specific IR spectral fingerprint, FTIR also has a relevant role in detecting alterations or pathological states of the molecules and samples, leading to different spectra between patients and healthy controls, as presented in several examples in Section 4. In the presence of pathology, the IR spectrum of a sample will change, either by changing its intensity or shifting its peak frequencies [37]. These shifts can be due to multiple chemical alterations in the molecules' composition, including weakening of the bonds, decreasing mass of the molecules, or even shifting the stretching vibrations due to temperature variations, which will change the vibrational frequencies of the bands. More details on this can be found elsewhere [37].

#### *2.4. Michelson Interferometer*

The Michelson Interferometer technique was adapted for FTIR so that the light from the polychromatic IR source, effectively a blackbody radiator, is collimated and directed onto a beam splitter, with 50% of the photons by the fixed mirror and 50% transmitted by the movable mirror [32]. In this configuration, light is reflected from the two mirrors back to the beam splitter, and some fraction of the original light passes into the sample compartment Figure 3.

There, the light is focused on the sample. When leaving the sample compartment, the light is refocused on the detector. The difference in the optical path length between the two mirrors to the interferometer is known as the retardation or optical path difference (OPD) [32]. An interferogram (as in Figure 2) is obtained by varying the retardation and recording the signal from the detector for different retardation values. When no sample is present, the interferogram profile depends on the variation of the source intensity and splitter efficiency with wavelength. This results in a maximum at zero retardation when there is constructive interference at all wavelengths, followed by a series of wiggles [32]. This problem is critical in the case of zero default when there is constructive interference within the smallest wavelengths followed by a series of wigglers. The location of the null default is determined by locating the purpose of the excessive intensity within the

interferogram. When a pattern is given away, the course interferogram is modulated with the aid of the absorption bands within the pattern (as exemplified in Figure 3) [32]. *Micromachines* **2022**, *13*, x FOR PEER REVIEW 6 of 20

**Figure 3.** Schematic diagram of a Michelson interferometer configured for FTIR. (**a**) An ideal Michelson interferometer; (**b**) a Michelson interferometer with the movable mirror tilting. The continuous and dashed lines represent the different directions of light. Reprinted from [32], MDPI, under a Creative Commons Attribution (CC BY) license. **Figure 3.** Schematic diagram of a Michelson interferometer configured for FTIR. (**a**) An ideal Michelson interferometer; (**b**) a Michelson interferometer with the movable mirror tilting. The continuous and dashed lines represent the different directions of light. Reprinted from [32], MDPI, under a Creative Commons Attribution (CC BY) license.

#### There, the light is focused on the sample. When leaving the sample compartment, the **3. Methods**

light is refocused on the detector. The difference in the optical path length between the two mirrors to the interferometer is known as the retardation or optical path difference (OPD) [32]. An interferogram (as in Figure 2) is obtained by varying the retardation and recording the signal from the detector for different retardation values. When no sample is present, the interferogram profile depends on the variation of the source intensity and splitter efficiency with wavelength. This results in a maximum at zero retardation when there is constructive interference at all wavelengths, followed by a series of wiggles [32]. This problem is critical in the case of zero default when there is constructive interference within the smallest wavelengths followed by a series of wigglers. The location of the null default is determined by locating the purpose of the excessive intensity within the interferogram. When a pattern is given away, the course interferogram is modulated with the aid of the absorption bands within the pattern (as exemplified in Figure 3) [32]. **3. Methods**  The research and data collecting strategy was based on evaluating a wide scale of papers (matching the topic keywords) published in the FTIR spectroscopy field, in the last decades, with all adequate reference and copyright permissions. For that, a comprehensive electronic search on ScienceDirect, Scopus and PubMed databases was performed (up to October 2021, Q3), as well as a direct search on different publishers' specific databases, such as MDPI, Wiley, or Nature, among others. Search keywords included: FTIR, spectroscopy, optics, infrared, blood, blood cells, Functional Group, Michelson Interferometer, lab-on-a-chip, microfluidics, microdevice and diagnostics. The search strategy was estab-The research and data collecting strategy was based on evaluating a wide scale of papers (matching the topic keywords) published in the FTIR spectroscopy field, in the last decades, with all adequate reference and copyright permissions. For that, a comprehensive electronic search on ScienceDirect, Scopus and PubMed databases was performed (up to October 2021, Q3), as well as a direct search on different publishers' specific databases, such as MDPI, Wiley, or Nature, among others. Search keywords included: FTIR, spectroscopy, optics, infrared, blood, blood cells, Functional Group, Michelson Interferometer, lab-on-achip, microfluidics, microdevice and diagnostics. The search strategy was established by combining several keywords and using AND/OR Boolean operators. The relevant studies resulting from the database search were manually analysed to identify other potential studies to be included. The exclusion criteria were: reviews, comments, overviews, case reports, viewpoints and perspectives, as well as documents reporting tests with data ambiguity. Studies not written in the English language were also excluded, as well as duplicate results. From there, titles and abstracts were screened. All abstracts were read, and those that did not fit the purpose of this review were excluded. The information regarding the application, quantitative outcomes, reported study limitations, and other relevant comments were selected and extracted from the remaining articles. Specifically, the authors selected papers that reported FTIR spectroscopy for analysing between normal blood samples and pathological blood samples for cancer detection, HIV early recognition in pregnant women, and blood grouping identification, among others. These applications were chosen to illustrate the wide range of FTIR applications. Finally, the most important conclusions and limitations on the analysed papers were summarised.

#### lished by combining several keywords and using AND/OR Boolean operators. The rele-**4. Results**

vant studies resulting from the database search were manually analysed to identify other potential studies to be included. The exclusion criteria were: reviews, comments, overviews, case reports, viewpoints and perspectives, as well as documents reporting tests with data ambiguity. Studies not written in the English language were also excluded, as This section presents the data collection results examining the most relevant studies over the last twenty years addressing FTIR in the biological field, organised by their application field.

#### well as duplicate results. From there, titles and abstracts were screened. All abstracts were *4.1. Collection of Data Seeking Applications of FTIR Spectroscopy*

read, and those that did not fit the purpose of this review were excluded. The information regarding the application, quantitative outcomes, reported study limitations, and other relevant comments were selected and extracted from the remaining articles. Specifically, the authors selected papers that reported FTIR spectroscopy for analysing between normal blood samples and pathological blood samples for cancer detection, HIV early recog-FTIR is widely used as a diagnostic method to analyse different materials and samples. According to the PubMed search results, FTIR spectroscopy was first studied as a new potential method in 1972, and until now (2021, Q3), more than 76,900 papers have been published on FTIR spectroscopy. A few years later, in 1982, researchers in the biological

important conclusions and limitations on the analysed papers were summarised.

field recognised the potential of the FTIR techniques and their suitability for diagnostics, including diagnoses of a long list of diseases that included cancer, microbes, bacteria, and viruses detection. The number of papers relating to FTIR published in the biological field in 2021 (up to Q3) is around 4810. Figure 4a,b illustrate the evolution of FTIR studies over the last two decades. logical field recognised the potential of the FTIR techniques and their suitability for diagnostics, including diagnoses of a long list of diseases that included cancer, microbes, bacteria, and viruses detection. The number of papers relating to FTIR published in the biological field in 2021 (up to Q3) is around 4810. Figure 4a,b illustrate the evolution of FTIR studies over the last two decades. logical field recognised the potential of the FTIR techniques and their suitability for diagnostics, including diagnoses of a long list of diseases that included cancer, microbes, bacteria, and viruses detection. The number of papers relating to FTIR published in the biological field in 2021 (up to Q3) is around 4810. Figure 4a,b illustrate the evolution of FTIR studies over the last two decades.

This section presents the data collection results examining the most relevant studies over the last twenty years addressing FTIR in the biological field, organised by their ap-

This section presents the data collection results examining the most relevant studies over the last twenty years addressing FTIR in the biological field, organised by their ap-

FTIR is widely used as a diagnostic method to analyse different materials and samples. According to the PubMed search results, FTIR spectroscopy was first studied as a new potential method in 1972, and until now (2021, Q3), more than 76,900 papers have been published on FTIR spectroscopy. A few years later, in 1982, researchers in the bio-

FTIR is widely used as a diagnostic method to analyse different materials and samples. According to the PubMed search results, FTIR spectroscopy was first studied as a new potential method in 1972, and until now (2021, Q3), more than 76,900 papers have been published on FTIR spectroscopy. A few years later, in 1982, researchers in the bio-

*Micromachines* **2022**, *13*, x FOR PEER REVIEW 7 of 20

*Micromachines* **2022**, *13*, x FOR PEER REVIEW 7 of 20

*4.1. Collection of Data Seeking Applications of FTIR Spectroscopy* 

*4.1. Collection of Data Seeking Applications of FTIR Spectroscopy* 

**4. Results** 

**4. Results** 

plication field.

plication field.

**Figure 4.** Published papers focusing FTIR: (**a**) Overall papers, since 1972; (**b**) papers in the biological field, since 1985 (until 2021, Q3). **Figure 4.** Published papers focusing FTIR: (**a**) Overall papers, since 1972; (**b**) papers in the biological field, since 1985 (until 2021, Q3). **Figure 4.** Published papers focusing FTIR: (**a**) Overall papers, since 1972; (**b**) papers in the biological field, since 1985 (until 2021, Q3).

In particular, by analysing the PubMed search results, a short number of papers addressed the use of FTIR spectroscopy to analyse and distinguish between normal blood sample cells and pathological blood samples. From the research undertaken until 2021, this number is less than 100 (Figure 5). In particular, by analysing the PubMed search results, a short number of papers addressed the use of FTIR spectroscopy to analyse and distinguish between normal blood sample cells and pathological blood samples. From the research undertaken until 2021, this number is less than 100 (Figure 5). In particular, by analysing the PubMed search results, a short number of papers addressed the use of FTIR spectroscopy to analyse and distinguish between normal blood sample cells and pathological blood samples. From the research undertaken until 2021, this number is less than 100 (Figure 5).

**Figure 5.** Published papers focusing FTIR for addressing differentiation between normal/pathological blood samples, from 1999 until 2021 (Q3). **Figure 5.** Published papers focusing FTIR for addressing differentiation between normal/pathological blood samples, from 1999 until 2021 (Q3). **Figure 5.** Published papers focusing FTIR for addressing differentiation between normal/pathological blood samples, from 1999 until 2021 (Q3).

Figure 6 summarises the number of published papers considering the FTIR subject (total of 76,900 papers) and its specificity in the application of the biology domain (4810) and subdomain for distinguishing between normal and pathological blood samples (around 100). (around 100).

(around 100).

*Micromachines* **2022**, *13*, x FOR PEER REVIEW 8 of 20

**Figure 6.** Summary of the total published papers focusing FTIR from 1999 until 2021 (Q3). *4.2. Applications of FTIR Spectroscopy in Cancer Diagnosis* 

**Figure 6.** Summary of the total published papers focusing FTIR from 1999 until 2021 (Q3).

#### *4.2. Applications of FTIR Spectroscopy in Cancer Diagnosis* Among the different spectroscopic techniques developed to distinguish between normal and cancerous blood tissues, Fourier transformed spectroscopy has shown tremen-

*4.2. Applications of FTIR Spectroscopy in Cancer Diagnosis*  Among the different spectroscopic techniques developed to distinguish between normal and cancerous blood tissues, Fourier transformed spectroscopy has shown tremendous potential. Additionally, biomedicine's IR-based techniques have become a reality with a large amount of information accumulated from clinical studies, trials, and devel-Among the different spectroscopic techniques developed to distinguish between normal and cancerous blood tissues, Fourier transformed spectroscopy has shown tremendous potential. Additionally, biomedicine's IR-based techniques have become a reality with a large amount of information accumulated from clinical studies, trials, and developments [38–40]. dous potential. Additionally, biomedicine's IR-based techniques have become a reality with a large amount of information accumulated from clinical studies, trials, and developments [38–40]. In 2013, FTIR spectroscopy was applied to study healthy and cancerous blood samples, using a diffuse reflectance technique from SHIMADZU 8000 series FTIR spectropho-

Figure 6 summarises the number of published papers considering the FTIR subject (total of 76,900 papers) and its specificity in the application of the biology domain (4810) and subdomain for distinguishing between normal and pathological blood samples

Figure 6 summarises the number of published papers considering the FTIR subject (total of 76,900 papers) and its specificity in the application of the biology domain (4810) and subdomain for distinguishing between normal and pathological blood samples

opments [38–40]. In 2013, FTIR spectroscopy was applied to study healthy and cancerous blood samples, using a diffuse reflectance technique from SHIMADZU 8000 series FTIR spectrophotometer. The spectra of cancerous and healthy blood were registered at a resolution of 4 cm−1 in the region of 900 cm−1 to 2000 cm−1, as observed in Figure 7a. The obtained results show that the bands of proteins, lipids, carbohydrates, and nucleic acids from cancerous samples are clearly different from the normal ones dominated by two absorption bands at 1643 cm−1–1550 cm−1, known as amide I and amide II. Amide I appears from the C=O stretching vibrations and amide II from the C–N stretching and CNH bending vibrations. In 2013, FTIR spectroscopy was applied to study healthy and cancerous blood samples, using a diffuse reflectance technique from SHIMADZU 8000 series FTIR spectrophotometer. The spectra of cancerous and healthy blood were registered at a resolution of 4 cm−<sup>1</sup> in the region of 900 cm−<sup>1</sup> to 2000 cm−<sup>1</sup> , as observed in Figure 7a. The obtained results show that the bands of proteins, lipids, carbohydrates, and nucleic acids from cancerous samples are clearly different from the normal ones dominated by two absorption bands at 1643 cm−1–1550 cm−<sup>1</sup> , known as amide I and amide II. Amide I appears from the C=O stretching vibrations and amide II from the C–N stretching and CNH bending vibrations. This wavelength band looks strong and sharp in a healthy blood sample [38], as seen in Figure 7b. tometer. The spectra of cancerous and healthy blood were registered at a resolution of 4 cm−1 in the region of 900 cm−1 to 2000 cm−1, as observed in Figure 7a. The obtained results show that the bands of proteins, lipids, carbohydrates, and nucleic acids from cancerous samples are clearly different from the normal ones dominated by two absorption bands at 1643 cm−1–1550 cm−1, known as amide I and amide II. Amide I appears from the C=O stretching vibrations and amide II from the C–N stretching and CNH bending vibrations. This wavelength band looks strong and sharp in a healthy blood sample [38], as seen in Figure 7b.

(**a**) (**b**) **Figure 7.** (**a**) FTIR absorption spectra of 'a' cancerous blood, 'b' normal blood and 'c' water samples using air as a reference; (**b**) detail of the FTIR absorption spectra of the normal and cancerous blood. Reprinted from [38], Copyright 2010 Convener, MMSETLSA-2009, with permission from the authors.

In 2016, label-free FTIR was used for early cancer detection in blood samples. This technique allowed detecting and verifying spectral biomarker candidate patterns to detect non-small cell lung carcinoma (NSCLC). The study was conducted on 161 patients where blood serum and plasma samples were analysed using an automatic FTIR spectroscopic system, together with pattern recognition algorithms, such as Monte Carlo cross-validation, linear discriminant analysis and random forest classification. Marker patterns for cancer discrimination (both from squamous-cell carcinoma and adenocarcinoma patients) from clinically relevant disease control patients were identified in FTIR spectra of blood samples. The analysis was constrained to the respective C–H-stretching in the 2800 cm−<sup>1</sup> –3200 cm−<sup>1</sup> region and the fingerprint regions 1750 cm−1–875 cm−<sup>1</sup> [39]. Accuracy of up to 79% was recorded [39]. According to the authors, the study demonstrates the applicability of FTIR spectroscopy using blood for lung cancer detection. Evidence for cancer subtype discrimination was given. With improved performance, the method could be developed as a routine diagnostic tool for blood testing of NSCLC [39].

Another study from 2018 demonstrated that ATR-FTIR spectroscopy is a potential technique that can be used for cutaneous melanoma (i.e., skin cancer) diagnosis and for differentiating the metastatic potential of cancer cells. By using IR spectroscopy, one can identify various types of cancer such as basal cell carcinoma, malignant melanoma, nevus, as well as metastatic potential by evaluating the alterations in hydration level and molecular changes [40]. The spectra obtained by the authors show different intensities and frequencies of normal and cancerous samples in the spectral range between 4000 cm−<sup>1</sup> and 400 cm−<sup>1</sup> . The region between 4000 cm−<sup>1</sup> and 3000 cm−<sup>1</sup> shows stretching vibrations of O-H and N-H corresponding to the spectral bands of collagens and proteins of the skin. As cancer changes, the permeability of the cells' membrane and the metastatic potential also change with the hydration grade of the cell membrane; the ATR-FTIR spectroscopy is an approach that allows successful differentiation of the metastatic potential of cancer cells [40]. In particular, the comparison between fewer and more metastatic cells shows that the hydration level of the plasma membrane leads to a significant difference between both states of cancer [40].

A study from 2020 presented an easy to use, a reagent-free method based on (ATR-FTIR) spectroscopy to quantify the protein content of extracellular vesicles (EV) samples with no sample preparation [41]. After calibration with bovine serum albumin, the protein concentration of red blood cell-derived EVs (REVs) was investigated by ATR-FTIR spectroscopy. The integrated region of the amide I band was calculated from the IR spectra of REVs, which was proportional to the protein quantity in the sample. Discriminatory protein bands of amide A, amide I and amide II were set at 3298, 1657, and 1546 cm−<sup>1</sup> , respectively. In the reported study, vibrations corresponding to the lipid components were also witnessed as antisymmetric and symmetric methylene stretching of acyl chains in the range 2924 cm−<sup>1</sup> to 2850 cm−<sup>1</sup> , and the C=O stretching at <sup>−</sup>1738 cm−<sup>1</sup> of the glycerol esters, respectively [41], as shown in Figure 8.

This new method presents a reagent-free alternative to traditional colourimetric protein determination assays and requires no special sample preparation to investigate EVs [41]. Therefore, this IR spectroscopy–based protein quantification method can be successfully adapted to the routine analysis of extracellular vesicles.

FTIR was also used to detect biomarkers for early screening of pediatric leukaemia [42]. In the reported study, the spectra were acquired from blood serum samples of ten child patients with B-cell precursor lymphoblastic leukaemia (BCP-ALL) and were contrasted with ten control samples. No clear peak shift was spotted between the averaged spectra of leukaemia patients and healthy individuals at the first trial. Thus, the authors applied the ratios of particular corrected peaks heights and the second derivatives analytical approaches to better distinguish between BCP-ALL and the control sample. A significant shift was observed for the peak corresponding to the amide I band (1700 cm−<sup>1</sup> to 1600 cm−<sup>1</sup> ) due to the C=O stretch vibrations of the peptide linkages. The frequencies of the amide I band are originally fixed to the secondary structure of the proteins. The position of the amide I band was at 1645 cm−<sup>1</sup> in the FTIR spectrum of the control group, whereas for the BCP-ALL patients, the peak was shifted to 1641 cm−<sup>1</sup> [42], as seen in Figure 9.

**Figure 8.** ATR-FTIR spectroscopy quantifies the protein content of extracellular vesicles (EV) samples. (**a**) Raw absorbance spectra after ATR correction. (**b**) Absorbance spectra after baseline correction and normalisation. (**c**) Absorbance spectra after buffer subtraction. (**d**) Zoomed absorbance spectra for calculating area under the curve (AUC) values of the amide I band by integration in 1700 cm<sup>−</sup>1–1600 cm−1 wavenumber region. Reprinted from [41], SpringerLink, under a Creative Commons Attribution 4.0 International License. **Figure 8.** ATR-FTIR spectroscopy quantifies the protein content of extracellular vesicles (EV) samples. (**a**) Raw absorbance spectra after ATR correction. (**b**) Absorbance spectra after baseline correction and normalisation. (**c**) Absorbance spectra after buffer subtraction. (**d**) Zoomed absorbance spectra for calculating area under the curve (AUC) values of the amide I band by integration in 1700 cm−1–1600 cm−<sup>1</sup> wavenumber region. Reprinted from [41], SpringerLink, under a Creative Commons Attribution 4.0 International License.

This new method presents a reagent-free alternative to traditional colourimetric protein determination assays and requires no special sample preparation to investigate EVs [41]. Therefore, this IR spectroscopy–based protein quantification method can be successfully adapted to the routine analysis of extracellular vesicles. Thus, the differences between the FTIR spectral profile of leukemic and normal serum may offer a potential route to the early identification of children with BCP-ALL, limiting the number of invasive procedures and accelerating the diagnosis of individuals. The possibility of the early detection of leukaemia in children based only on the FTIR analysis of their serum seems an attractive tool for routine medical practice [42].

for the BCP-ALL patients, the peak was shifted to 1641 cm−1 [42], as seen in Figure 9.

FTIR was also used to detect biomarkers for early screening of pediatric leukaemia [42]. In the reported study, the spectra were acquired from blood serum samples of ten child patients with B-cell precursor lymphoblastic leukaemia (BCP-ALL) and were contrasted with ten control samples. No clear peak shift was spotted between the averaged spectra of leukaemia patients and healthy individuals at the first trial. Thus, the authors applied the ratios of particular corrected peaks heights and the second derivatives analytical approaches to better distinguish between BCP-ALL and the control sample. A significant shift was observed for the peak corresponding to the amide I band (1700 cm−1 to 1600 cm−1) due to the C=O stretch vibrations of the peptide linkages. The frequencies of the

**Figure 9.** FTIR as a tool for detecting BCP-ALL biomarkers for early screening of pediatric leukaemia. Normalised average FTIR spectra of serum samples: control (black) and Acute Lymphoblastic Leukemia Precursor B (red). The presented spectra cover the range of 800 cm<sup>−</sup>1–3500 cm−1. Reprinted from [42], MDPI, under a Creative Commons Attribution (CC BY) license. **Figure 9.** FTIR as a tool for detecting BCP-ALL biomarkers for early screening of pediatric leukaemia. Normalised average FTIR spectra of serum samples: control (black) and Acute Lymphoblastic Leukemia Precursor B (red). The presented spectra cover the range of 800 cm−1–3500 cm−<sup>1</sup> . Reprinted from [42], MDPI, under a Creative Commons Attribution (CC BY) license.

#### Thus, the differences between the FTIR spectral profile of leukemic and normal se-*4.3. Applications of FTIR Spectroscopy in HIV Early Detection*

rum may offer a potential route to the early identification of children with BCP-ALL, limiting the number of invasive procedures and accelerating the diagnosis of individuals. The possibility of the early detection of leukaemia in children based only on the FTIR analysis of their serum seems an attractive tool for routine medical practice [42]. *4.3. Applications of FTIR Spectroscopy in HIV Early Detection*  In 2020, ATR-FTIR spectroscopy was considered for distinguishing HIV-infected patients from healthy uninfected controls [43]. This study comprised one hundred and twenty blood plasma samples of pregnant women and allowed to obtain good sensitivity (83%) and specificity (95%) using a genetic set of rules with linear discriminant assessment (GA-LDA). In the range of 1800 cm−1 to 900 cm−1, the spectra displayed some particular feature absorptions, including the amide I band at 1635 cm−1, an arm at 1560 cm−1 (due to C=O, Amide II) and three small depth absorptions at 1480 cm−1 (corresponding to the C-H asymmetric deformation of methyl agencies), at 1404 cm−1 (due to the COO−symmetric stretching of proteins and lipids) and 1060 cm−1 (due to the C-O nucleic acids). Due to the similarity between the spectral features in the groups (uninfected control and HIV in-In 2020, ATR-FTIR spectroscopy was considered for distinguishing HIV-infected patients from healthy uninfected controls [43]. This study comprised one hundred and twenty blood plasma samples of pregnant women and allowed to obtain good sensitivity (83%) and specificity (95%) using a genetic set of rules with linear discriminant assessment (GA-LDA). In the range of 1800 cm−<sup>1</sup> to 900 cm−<sup>1</sup> , the spectra displayed some particular feature absorptions, including the amide I band at 1635 cm−<sup>1</sup> , an arm at 1560 cm−<sup>1</sup> (due to C=O, Amide II) and three small depth absorptions at 1480 cm−<sup>1</sup> (corresponding to the C-H asymmetric deformation of methyl agencies), at 1404 cm−<sup>1</sup> (due to the COO−symmetric stretching of proteins and lipids) and 1060 cm−<sup>1</sup> (due to the C-O nucleic acids). Due to the similarity between the spectral features in the groups (uninfected control and HIV infected), chemometric patterns were used to identify spectral features responsible for class differentiation Figure 10. ATR-FTIR spectroscopy with multivariate analysis was able to accurately identify HIV-infected pregnant women based on blood plasma, showing the potential of this method for early detection of HIV in a fast and reagent-free approach. Successful development of this method in a clinical environment could aid early diagnosis of gestational HIV and help treatment [43].

fected), chemometric patterns were used to identify spectral features responsible for class differentiation Figure 10. ATR-FTIR spectroscopy with multivariate analysis was able to accurately identify HIV-infected pregnant women based on blood plasma, showing the potential of this method for early detection of HIV in a fast and reagent-free approach. Successful development of this method in a clinical environment could aid early diagnosis

of gestational HIV and help treatment [43].

**Figure 10.** ATR-FTIR spectra for distinguishing between HIV infected and healthy blood samples. (**A**) Mean raw IR spectra in the biofingerprint region (1800 cm<sup>−</sup>1–900 cm−1) for HIV-infected (HIV) and healthy uninfected controls (HC) samples. (**B**) Mean preprocessed IR spectra (AWLS baseline correction) in the biofingerprint region (1800 cm<sup>−</sup>1–900 cm−1) for HIV-infected (HIV) and healthy uninfected controls (HC) samples. (**C**) Discriminant function (DF) for the samples in the test set, where HIV stands for HIV-infected samples and HC for healthy uninfected controls, allowing their distinction. Reprinted from [43], Nature, under a Creative Commons Attribution 4.0 International License. **Figure 10.** ATR-FTIR spectra for distinguishing between HIV infected and healthy blood samples. (**A**) Mean raw IR spectra in the biofingerprint region (1800 cm−1–900 cm−<sup>1</sup> ) for HIV-infected (HIV) and healthy uninfected controls (HC) samples. (**B**) Mean preprocessed IR spectra (AWLS baseline correction) in the biofingerprint region (1800 cm−1–900 cm−<sup>1</sup> ) for HIV-infected (HIV) and healthy uninfected controls (HC) samples. (**C**) Discriminant function (DF) for the samples in the test set, where HIV stands for HIV-infected samples and HC for healthy uninfected controls, allowing their distinction. Reprinted from [43], Nature, under a Creative Commons Attribution 4.0 International License.

#### *4.4. Applications of FTIR Spectroscopy in Blood Grouping Analysis 4.4. Applications of FTIR Spectroscopy in Blood Grouping Analysis*

In 2017, a study explored the potential for the spectroscopic identification of blood antigens using an FTIR spectrophotometer (Shimadzu FTIR-8400S) within the range of 4000 cm−1 to 400 cm−1 [44]. The ABO blood type system is reflected in the FTIR spectra of human blood. Specific bands at 1166 cm−1 and 1020 cm−1 represent the fucose molecules linked glycosidically with galactose and -GlcNAc-, respectively, related to the O antigen. In 2017, a study explored the potential for the spectroscopic identification of blood antigens using an FTIR spectrophotometer (Shimadzu FTIR-8400S) within the range of 4000 cm−<sup>1</sup> to 400 cm−<sup>1</sup> [44]. The ABO blood type system is reflected in the FTIR spectra of human blood. Specific bands at 1166 cm−<sup>1</sup> and 1020 cm−<sup>1</sup> represent the fucose molecules linked glycosidically with galactose and -GlcNAc-, respectively, related to the O antigen.

When -GalNAc- is linked to -O antigen- through glycoside linkage, it exhibits a band at 1022 cm−1, due to the -A antigen-. A band at 1166 cm−1 reveals additional galactose glycosidically bonded to -O antigen-, as seen in Table 2. Summarily, the IR spectroscopic data on human blood of groups A, B, AB, and O explores the possibility of the nonlabelled and reagent free identification of blood antigens using FTIR [44]. When -GalNAc- is linked to -O antigen- through glycoside linkage, it exhibits a band at 1022 cm−<sup>1</sup> , due to the -A antigen-. A band at 1166 cm−<sup>1</sup> reveals additional galactose glycosidically bonded to -O antigen-, as seen in Table 2. Summarily, the IR spectroscopic data on human blood of groups A, B, AB, and O explores the possibility of the nonlabelled and reagent free identification of blood antigens using FTIR [44].


**Table 2.** Characteristic FTIR spectral data of human blood antigens (a–antigen) for blood grouping applications [44]. Reprinted with permission from the authors and the International Journal of Science, Environment and Technology. **Table 2.** Characteristic FTIR spectral data of human blood antigens (a–antigen) for blood grouping applications [44]. Reprinted with permission from the authors and the International Journal of Science, Environment and Technology.

#### *4.5. Applications of FTIR Spectroscopy in Blood Analysis 4.5. Applications of FTIR Spectroscopy in Blood Analysis*

*Micromachines* **2022**, *13*, x FOR PEER REVIEW 13 of 20

FTIR spectroscopy has also been considered in human blood analysis [13]. In 2004, a study presented a novel methodology for predicting the health status using FTIR-MC (micro-spectroscopy) data on blood components. In this study, FTIR-MC was complemented by cluster analysis algorithms (i.e., the task of grouping a set of objects in such a way that objects in the same group are more similar to each other than to those in other groups) [45]. The FTIR microscopic spectra of the major blood components, which include white blood cells (WBCs), red blood cells (RBCs), and plasma, were isolated from ten controls (average population). All the spectra were normalised to the amide I peak at 1643 cm−<sup>1</sup> . FTIR spectroscopy has also been considered in human blood analysis [13]. In 2004, a study presented a novel methodology for predicting the health status using FTIR-MC (micro-spectroscopy) data on blood components. In this study, FTIR-MC was complemented by cluster analysis algorithms (i.e., the task of grouping a set of objects in such a way that objects in the same group are more similar to each other than to those in other groups) [45]. The FTIR microscopic spectra of the major blood components, which include white blood cells (WBCs), red blood cells (RBCs), and plasma, were isolated from ten controls (average population). All the spectra were normalised to the amide I peak at 1643 cm−1.

The results reported by the authors showed that there are spectral variations between the three blood components to evaluate the validity of the method. Cluster analysis of the WBCs spectra in the 945 cm−<sup>1</sup> to 1282 cm−<sup>1</sup> range (comprises both symmetric and asymmetric regions of phosphate) and, more particularly, in the more specific range, from 1146 cm−<sup>1</sup> to 1282 cm−<sup>1</sup> , provided similar results, as shown in Figure 11a,b. The predictions (FTIR has also been used to analyse the body fluids for diagnostic and characterisation) matched the physician's diagnosis with 100% accuracy, proving the FTIR-MC as a potential tool to predict the health status of blood samples [45]. The results reported by the authors showed that there are spectral variations between the three blood components to evaluate the validity of the method. Cluster analysis of the WBCs spectra in the 945 cm−1 to 1282 cm−1 range (comprises both symmetric and asymmetric regions of phosphate) and, more particularly, in the more specific range, from 1146 cm−1 to 1282 cm−1, provided similar results, as shown in Figure 11a,b. The predictions (FTIR has also been used to analyse the body fluids for diagnostic and characterisation) matched the physician's diagnosis with 100% accuracy, proving the FTIR-MC as a potential tool to predict the health status of blood samples [45].

**Figure 11.** FTIR spectra of the major blood components: WBCs, RBCs and plasma, aiming for blood analysis. (**a**) Expanded region of FTIR-MC spectra (900–1500 cm<sup>−</sup>1) displaying the spectral differences in the symmetric and asymmetric stretching regions of the phosphate group, obtained by the average of ten representative controls; (**b**) FTIR-MSP spectra of the blood components of the averages of 10 representative controls in the 2700–3100 cm<sup>−</sup>1 region. (a) WBCs (blue); (b) RBCs (red); (c) Plasma (black) [45]. Adapted from [45] with permission from Wiley. **Figure 11.** FTIR spectra of the major blood components: WBCs, RBCs and plasma, aiming for blood analysis. (**a**) Expanded region of FTIR-MC spectra (900–1500 cm−<sup>1</sup> ) displaying the spectral differences in the symmetric and asymmetric stretching regions of the phosphate group, obtained by the average of ten representative controls; (**b**) FTIR-MSP spectra of the blood components of the averages of 10 representative controls in the 2700–3100 cm−<sup>1</sup> region. (a) WBCs (blue); (b) RBCs (red); (c) Plasma (black) [45]. Adapted from [45] with permission from Wiley.

Summarily, FTIR-MC can distinguish between the three main components of blood using spectral variations and cluster analysis. Specific spectral changes were observed between infected patients and age-matched healthy controls, providing good classification

[45].

Summarily, FTIR-MC can distinguish between the three main components of blood using spectral variations and cluster analysis. Specific spectral changes were observed between infected patients and age-matched healthy controls, providing good classification [45].

#### *4.6. Other Applications of FTIR Spectroscopy in the Biological Field*

Table 3 presents applications for the FTIR in the biological field tackled rather than blood cell distinction. As observed, FTIR is widely used in biology applications due to its potential to distinguish between different types of molecules.


#### **Table 3.** Examples of applications of FTIR in the biological field.

#### *4.7. Applications of FTIR Spectroscopy Integrated with Lab-on-a-Chip Devices*

Besides the macroscale FTIR applications, it has also been introduced into many modern technologies. In particular, lab-on-a-chip is a technology that has revolutionised and continues to revolutionise the medical field [60]. It intends to convert health care equipment into small devices that can be applied as point-of-care (PoC) methods for monitoring proposes. There are several examples in the literature of the combination of IR radiation and lab-on-achip technology [61]. Figure 12 presents an example of a pseudo-continuous flow FTIR system integrated on a microfluidic device for sugar identification [61]. Furthermore, the literature has already reported other miniaturised systems based on µFTIR for biological applications.

on µFTIR for biological applications.

*4.7. Applications of FTIR Spectroscopy Integrated with Lab-on-a-ChipDevices* 

Besides the macroscale FTIR applications, it has also been introduced into many modern technologies. In particular, lab-on-a-chip is a technology that has revolutionised and continues to revolutionise the medical field [60]. It intends to convert health care equipment into small devices that can be applied as point-of-care (PoC) methods for monitoring proposes. There are several examples in the literature of the combination of IR radiation and lab-on-a-chip technology [61]. Figure 12 presents an example of a pseudocontinuous flow FTIR system integrated on a microfluidic device for sugar identification [61]. Furthermore, the literature has already reported other miniaturised systems based

**Figure 12.** Schematic of the working principle of a pseudo-continuous flow FTIR system, integrated on a microfluidic device for sugar identification. The system includes a pumping station, a microfluidic device, a heating system (for temperature control), and a microscope-FTIR spectrometer. Reprinted from [61], MDPI, under a Creative Commons Attribution (CC BY) license. **Figure 12.** Schematic of the working principle of a pseudo-continuous flow FTIR system, integrated on a microfluidic device for sugar identification. The system includes a pumping station, a microfluidic device, a heating system (for temperature control), and a microscope-FTIR spectrometer. Reprinted from [61], MDPI, under a Creative Commons Attribution (CC BY) license.

A study reported by G. Birarda et al. [62] demonstrated a protocol to build a low-cost IR-Live microfluidic chip for real-time 2D infrared imaging of living cells or tissues with a resolution in the range of micrometres. In this study, FTIR compatible microfluidic chips were produced by direct photolithography of a resist layer coated onto one large IR window (40 mm diameter), with an inlet connected to a tubing system and an outlet attached to a circular reservoir [62]. In the centre of the device, there is an IR-transparent experimental chamber sandwiched between two CaF2 crystal discs. The results of IR imaging on migrating cells with the subcellular spatial resolution can distinguish different cellular organelles and identify their peculiar chemical composition at a functional group level. The authors, through the performed assays (n = 14), were able to show the characteristic shapes of the proteins (amide II bands) and lipids (CH2-CH3 stretching) in the cells. The spectrum has a sharp protein signal centred at 1654 cm−1, mainly attributed to an α-helix protein structure [62]. A study reported by G. Birarda et al. [62] demonstrated a protocol to build a low-cost IR-Live microfluidic chip for real-time 2D infrared imaging of living cells or tissues with a resolution in the range of micrometres. In this study, FTIR compatible microfluidic chips were produced by direct photolithography of a resist layer coated onto one large IR window (40 mm diameter), with an inlet connected to a tubing system and an outlet attached to a circular reservoir [62]. In the centre of the device, there is an IR-transparent experimental chamber sandwiched between two CaF<sup>2</sup> crystal discs. The results of IR imaging on migrating cells with the subcellular spatial resolution can distinguish different cellular organelles and identify their peculiar chemical composition at a functional group level. The authors, through the performed assays (n = 14), were able to show the characteristic shapes of the proteins (amide II bands) and lipids (CH2-CH<sup>3</sup> stretching) in the cells. The spectrum has a sharp protein signal centred at 1654 cm−<sup>1</sup> , mainly attributed to an α-helix protein structure [62].

Another method was developed for rapid ATR-FTIR monitoring solute concentrations in solutions flowing through microchannels [63]. The method involves the interface of commercially available ATR-FTIR instrumentation with a customised microfluidic device, which is sufficiently robust to withstand flow rates of the liquids of at least 20 mL h−1. The authors reported that the paper opened the way for on-chip identification of chemical compounds, measurements of their concentrations in solutions, and studies of Another method was developed for rapid ATR-FTIR monitoring solute concentrations in solutions flowing through microchannels [63]. The method involves the interface of commercially available ATR-FTIR instrumentation with a customised microfluidic device, which is sufficiently robust to withstand flow rates of the liquids of at least 20 mL h−<sup>1</sup> . The authors reported that the paper opened the way for on-chip identification of chemical compounds, measurements of their concentrations in solutions, and studies of reaction kinetics. Furthermore, the method can be used to characterise the adsorption of chemical and biological species adsorbed on the ATR surface under flow. From the spectrum in the region of 1400 cm−1–900 cm−<sup>1</sup> , the authors reported peaks at 1100 cm−<sup>1</sup> and 1250 cm−<sup>1</sup> , which correspond to the antisymmetric and symmetric vibrational modes of the COC groups, respectively, and a peek at 950 cm−<sup>1</sup> , which is from the C=C bonds in its phenyl ring. The authors also focused on the dominant band at 1100 cm−<sup>1</sup> and plotted the variation of its absorbance vs. concentration of TX-100 (*C*TX-100) in the solution. For *C*TX-100 ≥ 5 mM, the absorbance of the band linearly increased with the increasing solute concentration. According to the authors, the method allows the rapid acquisition of spectra and enables chemical characterisation and concentration measurements independent of the flow rate of liquids. The method enables the independent measurement of concentrations

of solutes with distinct spectral features in mixed solutions. For the polymer solutes, the authors report that the method has a sensitivity of at least 10 µM (0.01 wt%). The authors also proposed the method's applicability for the differentiation between dissolved and adsorbed amphiphilic species [63].

#### **5. Conclusions**

In this review paper, among the various spectroscopic techniques developed, FTIR is presented as a technique with the potential for distinguishing healthy from pathological samples. Numerous works have used FTIR with other techniques, such as ATR or micro-FTIR, to improve and simplify the spectral result of FTIR spectroscopy. The ATR-FTIR method promises the potential for the study of cells and tissues in general and, in particular, as a tool for estimating the metastatic potential of cancer cells. ATR-FTIR spectroscopy was also able to accurately identify HIV-infected pregnant women based on blood plasma, demonstrating the potential of this method for early detection of HIV in a rapid and reagent-free approach.

Label-free FTIR spectroscopy allows greater accuracy and reproducibility in cancer diagnosis while eliminating the need for complex and time-consuming clinical processing of tissue samples, currently required by existing computerised histopathological diagnosis. In addition, FTIR spectroscopy has also shown the potential to rapidly and objectively evaluate surgical resection margins to aid in surgical decision making, which may improve longterm survival and postoperative patient recovery compared with standard intraoperative pathological examination.

FTIR has also been used to monitor the response to cancer treatments and followup patients for treatment planning, early detection of recurrence, and assistance with psychological or psychosocial distress, with results that are faster, more sensitive, and more specific than conventional methods. Therefore, FTIR spectroscopy would be crucial to accelerate point-of-care decisions and potentially revolutionise cancer diagnostics in personalised medicine.

FTIR is the future measurement technique that shows tremendous potential and effective solutions to a large number of diagnostic complexities now faced by medical professionals. For instance, due to the limitations of the current gold standard techniques, FTIR may be advantageous to distinguish between normal samples and cancerous samples at an early stage, which offers the chance to diagnose and treat samples before any symptoms appear in patients. All these advantages force the researchers to dive deep into FTIR technology to move from a recognised to a viable technique used in the biological field, as in the other fields (environmental and chemical engineering, for instance) that are already recruiting FTIR for several applications.

#### **6. Future Trends**

Although FTIR is in continuous development in the biology field, the number of studies focusing on topics within the FTIR framework is steadily growing. In the future, FTIR may have a significant impact on various aspects of the medical field (i.e., hospital design, lab technician practices), including the financial. FTIR might bypass much equipment currently in use, as well as a large number of reagents used to perform the blood tests, thus proving to be a fast, convenient, economical, practical, and accurate method with high-quality results and minor environmental impact. However, to the best of the authors' knowledge, despite the great interest of the scientific community in FTIR, there are only a few microdevice platforms reported in the literature. Lab-on-a-chip devices, with integrated ATR-FTIR measurements for medical applications in real-time label-free living biological systems analysis, deal with the problem of water presence, either using cell culture medium, plasma, or serum samples, once the absorption values could overlap the bands of other components [64]. In this perspective, ATR-FTIR is the best option to be integrated to study both hydrated and dried biological samples, such as cells and fluid flow [64]. Specifically, in the field of biomechanics and living mechanobiology, ATR-FTIR spectroscopy visualisation

and quantification have also been demonstrated to be an excellent method for nondestructive biological analysis. Apart from all the cancer diagnostics biomarkers discussed in this work, cells' alterations related to diseases occur in most blood pathologies associated with mechanical and rheological changes. The detection and quantification of mechanical alterations have hundreds of applications in diverse fields, ranging from the analysis of cell biomechanics to the classification of tissue biopsies [64–66]. For example, mechanical differences in exosomes and microvesicles reflect changes in cell biomechanics and the cell type, state, treatment, and phenotype [66]. Thus, their quantification and analysis are important for diseases stratification and personalised medicine, showing ATR-FTIR as an advantageous strategy.

Another example is the RBC deformability analysis that is affected by several factors, such as ageing, high blood sugar levels, total cholesterol, or functional oxidative stress. Thus, their membrane and internal cytoplasm suffer changes which biochemical analysis (particularly ATR-FTIR) with morphological and rheological techniques can define and then provide a profile indicative of deformability alteration [67]. As ATR-FTIR technology for blood or body fluids analysis requires proper sample preparation, integration of microfluidics can play, once more, an important role in the development of strategies for sample preparation, such as cells or plasma separation devices, single-cell sorting, cell deformability devices, droplet generators, and cell traps, among others. It is known that spectral analysis demonstrated that deeply deformed cells have different cellular biochemistry compared to nondeformed ones. So, it is expected that significant improvements can be obtained by integrating sample preparation microfluidic systems, enabling the RBCs, white blood cells, or circulating tumour cells analysis, in terms of their membrane biochemical quantification and consequently their biomechanical behaviour [68]. Looking forward to applications in tumour-on-a-chip devices, transparent 3D microfluidic devices will allow ATR-FTIR microspectroscopy applications to monitor the biochemical response to both mechanical and chemical stimulations (i.e., drug resistance).

The development of such devices will be a step ahead in state of the art and will overcome the limitations of current technologies. Thus, to achieve such a goal, future works need to consider the design, fabrication, characterisation, and optimisation of lab-on-a-chip platforms, with IR radiation and Fourier Transform postprocessing, to examine blood cells, distinguishing between normal and pathological ones, and to better understand several mechanisms of treatment resistance and progression.

**Author Contributions:** Conceptualization, A.F., V.H.C. and G.M.; methodology, A.F., V.H.C., S.O.C. and G.M.; investigation, A.F., D.P., V.H.C., S.O.C. and G.M.; writing—original draft preparation, A.F.; writing—review and editing, D.P., V.H.C., S.O.C. and G.M.; supervision, V.H.C., S.O.C. and G.M.; funding acquisition, S.O.C. and G.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work results of the projects NORTE-01-0145-FEDER-029394, RTChip4Theranostics, and NORTE-01-0145-FEDER-028178, MalariaChip, supported by Programa Operacional Regional do Norte-Norte Portugal Regional Operational Programme (NORTE 2020), under the PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (FEDER) and by Fundação para a Ciência e Tecnologia (FCT), IP, projects reference PTDC/EMD-EMD/29394/2017 and PTDC/EEI-EEE/28178/2017. The authors also acknowledge the partial financial support by FCT national funds, under the national support to R&D units grant, through the reference project UIDB/04436/2020, UIDP/04436/2020 and UID/CEC/00319/2020. Susana Catarino thanks FCT for her contract funding provided through 2020.00215.CEECIND.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


### *Review* **Manual and Automatic Image Analysis Segmentation Methods for Blood Flow Studies in Microchannels**

**Violeta Carvalho <sup>1</sup> , Inês M. Gonçalves <sup>2</sup> , Andrews Souza <sup>3</sup> , Maria S. Souza <sup>4</sup> , David Bento 5,6 , João E. Ribeiro 6,7 , Rui Lima 1,5,\* and Diana Pinho 1,4,6**


**Abstract:** In blood flow studies, image analysis plays an extremely important role to examine raw data obtained by high-speed video microscopy systems. This work shows different ways to process the images which contain various blood phenomena happening in microfluidic devices and in microcirculation. For this purpose, the current methods used for tracking red blood cells (RBCs) flowing through a glass capillary and techniques to measure the cell-free layer thickness in different kinds of microchannels will be presented. Most of the past blood flow experimental data have been collected and analyzed by means of manual methods, that can be extremely reliable, but they are highly time-consuming, user-intensive, repetitive, and the results can be subjective to user-induced errors. For this reason, it is crucial to develop image analysis methods able to obtain the data automatically. Concerning automatic image analysis methods for individual RBCs tracking and to measure the well known microfluidic phenomena cell-free layer, two developed methods are presented and discussed in order to demonstrate their feasibility to obtain accurate data acquisition in such studies. Additionally, a comparison analysis between manual and automatic methods was performed.

**Keywords:** blood flow; particle tracking; red blood cells; manual methods; automatic methods; image analysis; biomicrofluidics

#### **1. Introduction**

Blood flow in microcirculation is crucial for the normal function of tissues and organs. Therefore, a detailed study of blood flow patterns and blood cells flowing in microvessels, microchannels and organs-on-chip is essential to provide a better understanding of the blood rheological properties and disorders in microcirculation [1–7]. One of the first techniques used for the study of flow patterns was the phase-contrast magnetic resonance imaging (PC-MRI). However, the technique requires long acquisition times and has low resolution [8,9]. Other techniques have been developed and combined to improve the acquisition and image processing. One of the most reliable ways to measure velocity fields in microcirculation is using Eulerian methods, such as the conventional micro-particle image velocimetry (PIV) [1,6,10–12] or the confocal micro-PIV [1,2,6,13]. The micro-PIV

**Citation:** Carvalho, V.; Gonçalves, I.M.; Souza, A.; Souza, M.S.; Bento, D.; Ribeiro, J.E.; Lima, R.; Pinho, D. Manual and Automatic Image Analysis Segmentation Methods for Blood Flow Studies in Microchannels. *Micromachines* **2021**, *12*, 317. https:// doi.org/10.3390/mi12030317

Academic Editor: Stefano Guido

Received: 20 February 2021 Accepted: 14 March 2021 Published: 18 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

technique is one of the best suitable methodologies to study blood flow phenomena in microcirculation. Some studies have also combined PIV with ultrasounds (Echo-PIV) [14,15]. However, most in vivo measurements contain physiological fluids with high concentrations of blood cells and as a result, the amount of tracer particles captured within the fluid is often very low [5]. Other approaches for blood flow studies are particle illumination photography, laser doppler velocimetry, fluorescent cytometry [16,17] and computer fluid dynamics [17,18].

In microcirculation, the study of red blood cells (RBCs) flowing in microvessels and microchannels and the study of the cell-free layer (CFL) thickness in different microchannels geometries are very important to get a better understanding of the blood rheological properties and disorders in microvessels in a fast and accurate way. The presence and physiological characteristics of other cell types are also of great clinical relevance [19]. In this kind of study, the image analysis has an important role to obtain crucial information about blood rheology. For blood flow in microvessels, where there is a large number of interacting cells, manual tracking methods have been used to accurately track individual deformable cells flowing through glass capillaries [1,11,20], straight polydimethylsiloxane microchannels [21], stenotic arteries [22,23], hyperbolic contractions [24], and bifurcations [25]. However, the manual data collection is extremely time-consuming to have a statistically representative number of samples and may introduce operators' errors that eventually limit the application of these methods many times at different conditions [26]. Hence, it is crucial to develop versatile and automatic methods able to automatically track and compute multiple cell trajectories and able to measure the cell-free layer thickness in a network of microchannels.

The purpose of this work is to review the state of the art of techniques used in in vitro blood flow studies and two developed methods (i) an automatic method to track RBCs flowing through microchannels and (ii) an automatic method to measure the CFL thickness in microchannels with bifurcations and confluences will be present and discuss.

This work is organized as follows, firstly an overview of methods used over the last years in the study of blood cells' morphology and tracking in in vitro blood flows is described. Secondly, a brief introduction to ImageJ, the image analysis software used to obtain manual data, will be made. Then, in Section 4, the results of manual and automatic methods applied were demonstrated and are discussed by the comparison with the manual data. Finally, a conclusion and future directions for the present work were discussed in Section 5.

#### **2. An Overview of Image Analysis Methods for Microfluidic Blood Phenomena Quantification**

#### *2.1. Image Segmentation and Thresholding*

Image analysis processing is a vast area that provides a large number of viable applications that can involve some steps such as image acquisition, image preprocessing, image segmentation, image post-processing and image analysis. Image segmentation is one of the most important and critical elements in automated image analysis, which consists in dividing a digital image into multiple regions, based on a set of pixels or objects, to simplify and/or change the representation of an image [27–29]. A variety of techniques can be applied: simple methods such as thresholding, or complex methods such as edge/boundary detection or region growing.

The literature contains hundreds of segmentation techniques [30,31], but there is no single method that can be considered good enough for all kinds of images. The main purpose of segmentation is to divide an image into regions of interest with similar gray-levels and textures in each region [32]. Segmentation methods change according to the imaging modality, application domain, method type—automatic or semi-automatic, depending on the image quality and the image artifacts, such as noise. Some segmentation methods may require image preprocessing prior to the segmentation algorithm [33,34]. Databases with algorithms to compensate for the uncertainties present in real-life datasets were developed [35]. On the other hand, some other methods apply post-processing to overcome the

problems arising from over-segmentation. Overall, segmentation methods can be grouped into thresholding, boundary detection, and region growing [27,29,31,36,37]. Those methods vary in the way that the image features are treated and the way the appearance and shape of the target are modeled [38].

Thresholding methods assign pixels with intensities below a certain threshold value into one class and the remaining pixels into another class and form regions by connecting adjacent pixels of the same class, that is, in the thresholding process, each pixel in a grayscale is recognized as either an object or background. The more advanced method creates histograms, oriented to the intensity of grayscale or color, showing the frequency of occurrence of certain intensities in an image so that the regions and objects are recognized from these data [28–30]. Thresholding methods work well on simple images where the objects and background have distinctively different intensity distributions. Boundary extraction methods use information about intensity differences between adjacent regions to separate the regions from each other. If the intensities within a region vary gradually but the difference of intensities between adjacent regions remains large, boundary detection methods can successfully delineate the regions [28–30,39]. Region growing methods form regions by combining pixels of similar properties [39,40].

#### *2.2. Blood Cell Image Segmentation and Tracking*

Over the last years, many studies have been conducted in the area of general segmentation methods that can analyze different types of medical images. Most used images are acquired during a diagnostic procedure and useful information is extracted for the medical professional. The development of image analysis in biomedical instrumentation engineering has the purpose of facilitating the acquisition of information useful for diagnosing, monitoring, treating or even investigating certain pathological conditions. It is important to always have in mind that the main purpose of biomedical imaging and image analysis is to provide a certain benefit to the subject or patient [41,42].

In normal human blood microscopic images, a high accumulation of RBCs could be observed, which results in the existence of touch and overlap between these cells [42]. These are two difficult issues in image segmentation where common segmentation algorithms cannot solve this problem [43]. Besides that, staining and illumination inconsistencies also act as uncertainty to the image [44]. This uncertainty makes the blood cell image segmentation a difficult and challenging task [43]. Numerous segmentation methods from peripheral blood or bone marrow smears have been proposed and most of them are region-based or edge-based schemes [42,45].

Jianhua et al. [46] developed an iterative Otsu's approach based on a circular histogram for the leukocyte segmentation. R. Sukesh Kumar et al. [47] developed two methods of color image segmentation using the RGB space as the standard processing space. These techniques might be used in blood cell image segmentation. Color images are a very rich source of information, because they provide a better description of a scene as compared to grayscale images. Hence, color segmentation becomes a very important and valuable issue [42,47]. For instance, Huang et al. [48] investigated a method based on the Otsu's method to segment and then recognize the type of leukocyte based on the characteristics of the nucleus. Willenbrock et al. [49] developed a program for image segmentation to detect both moving and stagnated cells in phase-contrast images. The program contributed to the study of the integrin LFA-1 mediation of lymphocyte arrest.

Khoo Boon et al. [50] performed comparisons between nine image segmentation methods which are gray-level thresholding, pattern matching, morphological operators, filtering operators, gradient-in method, edge detection operators, RGB color thresholding, color matching, HSL (hue, saturation, lightness) and color thresholding techniques on RBC. They concluded that there is no single method that can be considered good for RBC segmentation [42,50]. Meng Wang et al. [51] presented segmentation and online learning algorithms in acquiring, tracking and analyzing cell-cycle behaviors of a population of cells generated by time-lapse microscopy. Kan Jiang et al. [45] combined two techniques for

white blood cells (WBCs) segmentation. Two components of WBCs, nucleus and cytoplasm, are extracted respectively using different methods. First, a sub-image containing WBCs is separated from the cell image. Then, scale-space filtering is used to extract the nucleus region from the sub-image. Later, watershed clustering in a 3-D HSV (hue, saturation, value) histogram is processed to extract the cytoplasm region. Finally, morphological operations are performed to obtain the entire connective scheme successfully. Li et al. [52] developed a new method for WBCs identification. The method consists of the combination of an acousto-optic tunable filter (AOTF) adapter and a microscope for the image acquisition and an algorithm for data treatment. The results showed the high accuracy of the system. Pan et al. [53] trained a support vector machine model to simulate the human visual neuronal system and identify leukocytes from blood and bone marrow smear images.

Farnoosh et al. [54] developed a framework that consists of an integration of several digital image processing techniques, such as active contours, the snake algorithm and Zack thresholding for white blood cells, aiming to separate the nucleus and cytoplasm. Ritter et al. [55] presented an automatic method for segmentation and border identification of all objects that do not overlap the boundary [54]. Ongun et al. [56] did segmentation by morphological preprocessing followed by the snake-balloon algorithm [54]. Jiang et al. [45] proposed a WBC segmentation scheme on color space images using feature space clustering techniques for nucleus extraction [54]. Al-Dulaimi et al. [57] developed a WBC segmentation method using edge-based geometric active contours and the forces curvature, normal direction, and vector field. Maitra et al. [58] presented an approach to automatic segmentation and counting of RBCs in microscopic blood cell images using the Hough transform [54]. Another interesting investigation was carried out by Banik and colleagues [59]. They proposed an automatic WBC nucleus segmentation method, based on the HSI (hue, saturation, intensity), the L × a × b color space, and the k-means algorithm. This increases the generalization capability and evaluation result with a higher score on quality metrics. Then, to classify the localized WBC, they proposed a new convolutional neural network (CNN) model, which is the key factor to reduce the performance dependency between the proposed nucleus segmentation and classification method. In the end, they proved that segmentation performance does not affect the accuracy of the proposed classification method. Kawaguchi et al. [60] presented an image-based analytical method for time-lapse images of RBC and plasma dynamics with automatic segmentation. This method enabled the quantification of the perturbation-induced changes of the RBC and plasma passages in individual vessels and parenchymal microcirculation.

The literature has many more methods, however, most of the techniques presented previously were based in morphological analysis or in the form and constitution of the various blood constituents. Techniques developed for blood flows are still under development because there are many ways and methods for tracking movement. A good summary of object tracking methods can be found in [61] and cell tracking can be found in Miura et al. [62].

Recently other works appeared, for example, Dobbe et al. [63] presented a method applied to the sublingual microcirculation in a healthy volunteer and in a patient during cardiac surgery. Iqbal et al. [64] developed a novel method for the detection of abnormal behavior in cells through real-time images. The method was based in pixel classification using k-means and Bayesian classification. Chang et al. [32] segmented medical images through a charged fluid model. The model is divided in two steps defined by Poisson's equation. Measurements of functional microcirculatory geometry and velocity distributions using image techniques have been made, such as capillaroscopy, orthogonal polarized spectral and a side-stream dark field image [63]. Ashraf et al. [65] said that "cell mobility analysis is an essential process in many biology studies", so they have focused in developing a novel algorithm to image segmentation and tracking system conjugating the advantages of topological alignments and snakes, transforming the output of the topological alignments into the input of the active contour model to begin the analysis in the cells' boundaries and to determine cell mobility [65]. Pan et al. [66] proposed a bacterial foraging-based edge

detection (BFED) algorithm for cell image segmentation. The method was compared with the other four edge detector algorithms and showed more accurate and effective results.

In the case of Möller et al. [67], a semi-automatic tracking method with minimal user interaction was proposed. The framework was based on a topology-preserving variational segmentation approach applied to normal velocity components obtained from optical flow computations. Using the advantages of the optical flow, Kirisits et al. [68] introduced variational motion estimation for images that are defined on an evolving surface. Niazi et al. [69] studied an open-source computational method of particle tracking using MATLAB (2014 b, MathWorks, Natick, MA, US). The size and velocity of the particles are acquired from the video sequences from video-microscopic systems. The images are processed by a set of filters, selected by the user, to improve the accuracy. Park et al. [70] developed a deep learning-based super-resolution ultrasound (DL-SRU) for particle tracking. The method is based on a convolutional neural network and deep ultrasound localization microscopy. The DL-SRU was able to identify the positions of the RBCs reconstruct vessel geometry. Carboni et al. [71] used fluorescence to track blood particles flowing through a microfluidic channel. The recordings of the flow were analyzed with an algorithm developed using MATLAB to evaluate the margination parameter at relevant flows. The image processing consisted of three parts: background correction, calculation of the position and size of the particles through a gradient-based method and calculation of the displacements and velocities. Varga et al. [72] trained conventional-, deep- and convolutional neural networks to segment optical coherence tomography images to identify the number of hyperreflective foci. The networks coincide in the majority of the cases with the evaluation performed by different physicians. Chen et al. [73] studied a new approach for the segmentation of erythrocyte (red blood cell) shape. The technique was called complex local phase based subjective surfaces (CLAPSS) and presented a new variation scheme of stretching factor and was embedded with complex local phase information. The processed images were acquired by differential interference contrast (DIC) microscopy.

Some methods can also be used to track particles for diagnostic or treatments. For instance, Siegmund et al. [74] tested the use of nanoparticle labeling and magnetic resonance imaging (MRI) for in vivo tracking of adipose tissue-derived stromal cells (ASC). The labeling was stable for four months. This method has the disadvantage of not being able to identify the cell since it is an indirect method. Optimization is still required to reduce the amount of nanoparticles. Müller et al. [75] investigated the transport of magnetic particles in vessels of hen's egg models. The flow was subjected to the influence of a magnetic field in dark field reflected light and fluorescence mode. The particles were tracked by single-particle tracking (SPT). Irreversible agglomerates were visualized after stopping the magnetic field. Consequently, further studies of the interaction between cells and particles and of the particle coating are required. Also to support the diagnosis, Kucukal et al. [76] quantified the viscosity of preprocessing-free whole blood samples from the sickle cell disease patient population by using the micro-PIV technique for in vitro assessment of whole blood viscosity and RBC adhesion. More recently, Kucukal et al. [77] have been able to measure the velocity of whole blood flow in a microchannel during coagulation using a simple optical setup and processing the images using PIV and wavelet-based optical flow velocimetry. Both studies demonstrated the viability of image processing methods to obtain data with clinical relevance. Table 1 below shows the chronological progress of the studies and that, recently, the studies have been based on automatic methods with specific algorithms and particle tracking techniques.

For studies based on in vitro approaches, there are different automatic algorithms, however, most of them still under development because the results tend to overlap at high hematocrits (Hcts), and most of them are based on images that the researchers have, taking into account their aim. Therefore, to have a good method and take advantage of all its capabilities, it is ideal to develop our own algorithm for the objective that we want to achieve. In the following sections, we will discuss the application of two automatic methods.


**Table 1.** Summary of image analysis methods used for cell tracking and segmentation.

#### **3. ImageJ Manual Plugins**

ImageJ is a public domain Java image processing program. It can display, edit, analyze, process, save and print 8-bit, 16-bit, and 32-bit images. It can read many image formats including TIFF, GIF, JPEG, BMP, DICOM, FITS and "raw" data and supports "stacks", a series of images that share a single window. It is multithreaded, so time-consuming operations such as image file reading can be performed in parallel with other operations [78]. With ImageJ [78], it is possible to calculate the area and pixel value statistics of user-defined selections. It can measure distances and angles and create density histograms and line profile plots. Moreover, it supports standard image processing functions such as contrast manipulation, sharpening, smoothing, edge detection, and median filtering [78].

There are also different plugins to track RBCs, to count, or to measure the CFL thickness such as MtrackJ or ZProject. For example, in the study of the RBCs or other blood cell tracking, the plugin MtrackJ [49] is often used, facilitating the manual tracking of moving objects in image sequences and the measurement of basic track statistics. Through the MtrackJ plugin, the centroid of individual RBCs can be tracked, allowing obtaining the trajectory of each RBC. Additionally, it can be used to estimate RBC velocity, taking into consideration the x and y positions at each point (Figure 1a). To study the phenomena of CFL, manual tracking by MtrackJ can also be used or, as an alternative, the automatic function ZProject in ImageJ can be applied to process several images at once, creating a stack, and allowing observing the path of RBCs in the channel (Figure 1b). In Figure 1 is possible to see the application of the MtrackJ plugin to determine the CFL thickness in blood flow study [79].

**Figure 1.** ImageJ plugins: (**a**) MtrackJ used to obtain the RBC trajectory [79] and (**b**) application of the *plot Z-axis profile* function at the selected ROI [80].

Another tool from ImageJ used in studies of blood flow is the *Plot Z-axis profile*. This function allows determining the tonality of the pixels in a region of the interest (ROI) through time. After selecting a particular area of the video the *Plot Z-axis profile* tool measures the average of tonality of the pixels in the ROI and this tonality was used as a proxy of the local hematocrit. High tonality corresponds to low hematocrit and low tonality corresponds to high hematocrit [80]. Figure 2 represents the variation in the tonality in the ROI, and consequently the variation of the hematocrit in that region, over time.

**Figure 2.** Image of blood flow in the microchannel with labeled bright RBCs, f(x,y) and the centroid of the tracking cell.

Note that in MATLAB [27,39] there are some algorithms that researchers provide and also an application to work with ImageJ. A promising particle tracking velocimetry (PTV) plug-in for Image J is the "Particle tracker 2D and 3D" [81,82].

#### **4. Automatic Image Analysis Methods**

*4.1. Red blood Cells Trajectory in a Glass Capillary*

#### 4.1.1. Set-Up and Working Fluids

The confocal system used in this study consists of an inverted microscope (IX71; Olympus, Tokyo, Japan) combined with a confocal scanning unit (CSU22; Yokogawa Tokyo, Japan), a diode-pumped solid-state (DPSS) laser (Laser Quantum, Stockport, UK) with an excitation wavelength of 532 nm and a high-speed camera (Phantom v7.1; Vision Research, Wayne, NJ, USA). The laser beam was illuminated from below the microscope stage through a dry 40x objective lens with a numerical aperture (NA) equal to 0.9.

The light emitted from the fluorescent flowing RBCs, passes through a color filter into the scanning unit CSU22, where, by means of a dichromatic mirror, the light is reflected onto a high-speed camera to record the confocal images. The physiological fluid used was a solution of Dextran 40 (Dx40) with a Hct of 12%. Such was selected to obtain images with the best possible quality and consequently to reduce errors during cell tracking.

The RBCs were fluorescently labeled with a lipophilic carbocyanine derivative dye, chloromethylbenzamido (CM-Dil, C-7000, Molecular Probes, Eugene, OR, USA) using a procedure previously described [1,83]. This dye was well retained by the RBCs and had a strong light intensity, which allowed good visualization and tracking of labeled RBCs flowing in concentrated suspensions.

The microchannel used in this study was a 100 µm circular borosilicate glass capillary fabricated by Vitrocom (Mountain Lakes, NJ, USA). The capillary was mounted on a sliding glass with a thickness of 170 ± 20 µm and was immersed in glycerin to minimize the refraction from the walls.

#### 4.1.2. Manual Method

All confocal images were captured around the middle of the capillary with a resolution of 640 × 480 pixels, at a rate of 100 frames/second and then transferred to a computer for evaluation using Phantom camera control software (PH607). The manual method to track individual RBCs relies on the manual tracking plugin MTrackJ [84]. The bright centroid of the selected RBC was manually computed through successive images. After obtaining x and y positions, the data were exported for the determination of each individual RBC trajectory.

The output of this process is:


Figure 2 is an example of the blood flow image acquired with labeled bright RBCs and x-y coordinates.

#### 4.1.3. Automatic Method

A graphical user interface (GUI) in MATLAB was developed, for a better work environment for all users. This application must detect and track all objects that are present in a video sequence.

The algorithm is based on the steps as follows:


Firstly, the sequences of images were loaded to the GUI. Then, the region of interest (defined by the user) was cropped from the original images with the function *imcrop*; also a standard region is defined, but the user can change it for a better purpose. With this operation, we can work only with the region which needs to be analyzed (the region between the microchannel walls), making it easier to handle the images for the next steps, as presented in Figure 3.

**Figure 3.** Image sequences imported (**a**) and respective region of interest cropped (**b**).

The next operation is the image noise elimination by applying the median filter, *medfilt2*, with one 5 × 5 pixel mask. With that, the background of the images was smooth, and the objects are enhanced preserving the edges. Figure 4 presents the result of these processes.

**Figure 4.** The region of interest (**a**) and the image filtered by using the median function *medfilt2* (**b**).

In the next stage, the images were subject to a segmentation step using a threshold method. The definition of one or more values of separation is enough to divide the image into one or more regions, that is, differentiate the area of interest (the RBCs) from the not-interest area (background image). The level of threshold is calculated by default, by an iterative method, which means that for each image an adequate level of threshold is calculated. However, users can apply the value that they think to be more appropriate. After thresholding, the objects were defined with the Sobel filter (see Figure 5), which shows only the edge of the objects. The *Sobel* computes an approximation of the gradient of the image intensity. At each pixel point in the image, the result of the Sobel operator is either the corresponding gradient vector or the norm of this vector [31].

After the segmentation processing, the RBCs were tracked and sets of data (and positions) were obtained with the MATLAB function from the image processing toolbox, *regionprops* [27] (cf. Figure 6). This function measures a set of properties (area, centroid, etc.) for each connected component (RBC) in the binary image.

**Figure 6.** (**a**) Data extraction and (**b**) RBCs trajectories.

The data obtained were filtered because some of the objects are not RBC (that is, white blood cells or platelets that have higher or lower, respectively, area than the RBCs). Therefore, it is possible to filter the data by area, by imposing a minimum and maximum value. Another filter applied was the number of images where the objects are visible, because if the object has only a tracking with 10 positions, this data is not enough to be analyzed. The data with an extremely low number of tracking positions per object was eliminated.

Another approach for this type of application is underway, which is based on optical flow. Optical flow is a technique used in computer vision algorithms to measure the speed of the pixels based on comparisons of frames, creating a field that describes the displacement that occurred between two consecutive frames of a video sequence. In other words, the optical flow consists of a dense field of velocity where each pixel in the image plane is associated with a single velocity vector [85,86]. The Kalman method and the Lucas Kanade pyramidal method were applied to the same sequence of images (cf. Figure 7).

**Figure 7.** The obtained image when the Lucas Kanade pyramidal method was applied.

The Lucas Kanade pyramidal method shows a better approach to the objective, but the real dimension of the object and a continuous track along the image sequence are still under development. There is a great potential in this technique to follow moving objects, such as the RBCs flowing through a glass capillary, however, due to the complexity of the method and the need for multiple variables, further investigation is required.

#### 4.1.4. Results

Figure 8 shows the developed graphical user interface (GUI) in MATLAB performing the image processing described in the upper sections and the trajectories of individual labeled RBCs flowing in the center plane of a microchannel, determined by the manual tracking and the proposed automatic tracking method.

The present study indicates that the data obtained from the proposed automatic method significantly matches the data obtained from the manual method. This data, xy positions, can be used to calculate the means square deviation (MSD) and the radial dispersion (Dyy) to analyze the behavior of the RBC through a microchannel.

#### *4.2. Cell-Free Layer Thickness in a Bifurcation and Confluence Microchannel* 4.2.1. Set-Up and Working Fluids

The series of x-y images were captured with a resolution of 600 × 800 pixels. All images were recorded at the center plane of the microchannels at a rate of 200 frames/second, transferred to the computer and then evaluated by using an image analysis software. The microscope system used in the present study consisted of an inverted microscope (IX71, Olympus, Tokyo, Japan) combined with a high-speed camera (i-SPEED LT, Olympus, Tokyo, Japan). The blood samples used were collected from a healthy adult sheep, and ethylenediaminetetraacetic acid (EDTA) was added to prevent coagulation. The RBCs were separated from the blood by centrifugation and washed twice with physiological saline. The washed RBCs were suspended in Dextran 40 to make up the required RBCs concentration by volume.

**Figure 8.** Automatic method results: (**a**) developed graphical user interface (GUI) in MATLAB and (**b**) trajectories of individual labeled RBCs determined by the manual and automatic method.

4.2.2. Manual Methods

The MTrackJ plugin was used to automatically compute the centroid of the selected RBC. After obtaining x and y positions, the data was exported for the determination of each individual RBC trajectory (cf. Figure 9).

**Figure 9.** Manual method showing the trajectories of RBC defining the region of the CFL: (**a**) for an expansion geometry and (**b**) for a bifurcation geometry.

A semi-automatic method was also applied based on the use of the ZProject plugin [78]. This plugin projects an image stack along the axis perpendicular to the image plane (the so-called "z" axis) and has six different projection types.


**Figure 10.** (**a**) The obtained image by applying the projection average intensity and (**b**) the obtained image by applying the projection sum slices.

**Figure 11.** (**a**) Image obtained by applying the standard deviation projection and (**b**) image obtained by applying the median projection.

**Figure 12.** (**a**) The obtained image with the projection minimum intensity, and (**b**) the obtained image with the projection maximum intensity.

After applying an appropriate projection to a stack, the resulting image is obtained, and it is then converted to a binary image (see Figure 13). The thresholding in ImageJ can be done automatically or by applying the level that the user requires.

**Figure 13.** The obtained image from the ZProject method with the projection maximum intensity to extract the data. It shows a well defined CFL thickness.

This method works well for a good image quality and for simple geometry of the channels and represents the data accurately. Nevertheless, for more complex image data the method has some difficulties to get the correct data, so it will be necessary to specifically develop a method able to represent the data accurately.

To obtain the data, the tool *Wand* is used, which creates a selection by tracing objects of uniform color or thresholded objects. To trace an object with the *Wand* tool, it is necessary to click inside near the right edge, or outside to the left of the object. Once it finds the edge, it follows it until it returns to the starting point. The *Wand* takes the pixel value where you click as an initial value. Then, it selects a contiguous area under the condition that all pixel values in that area must be in the range initial value—tolerance to initial value + tolerance. Then the selected area will be analyzed to measure the CFL thickness.

#### 4.2.3. Automatic Method

The method is based on the binarization of the sequence image. The general steps of the method are:


All image sequences were processed using the image processing toolbox available in MATLAB [45]. The sequence of images was loaded (cf. Figure 14), and a median filter with a 3 × 3 pixel mask was applied to each frame to reduce the noise of the images.

**Figure 14.** An image from the original sequence of images.

Then, the intensity of each pixel in the frame sequence was evaluated to obtain an image with the maximum intensity. With this step, it was possible to identify the region with the highest concentration of blood cells and the region where blood cells do not exist, the cell-free layer (CFL). The regions that represent the CFL have the highest intensity (white) near the microchannel walls (cf. Figure 15).

**Figure 15.** Image with the maximum intensity evaluation.

As a final step, the image was converted into a binary image, the regions of interest were selected and the upper CFL trajectories were automatically measured. Figure 16 shows the image processing result for the developed method.

**Figure 16.** The obtained image from the automatic method.

The area to take the data is defined by the user selecting the wall of the channel and the limit area from the cell-free layer.

#### 4.2.4. Results

Figure 17 shows the results obtained by the manual method using the MtrackJ plugin and the automatic method presented in this work to measure the CFL thickness. A

microchannel with bifurcation and confluence shown in Figure 15 was used for the measurements. The values obtained with both methods can be seen also in Figure 15. Data was taken in the regions represented by A to F.

**Figure 17.** Comparison between the manual and the automatic data, taken in the regions A to F.

It is possible to note that the data obtained by the automatic method have similar behavior with the manual data. However, the values have some discrepancies. The quality of the image and also the level of the threshold can influence this type of measurements.

#### **5. Conclusions and Future Work**

The present work presents not only a review on blood cells tracking methods but also comparisons of a manual method and an automatic method for two different blood flow studies. Regarding the study where RBCs were tracked through a 100 µm glass capillary, the automatic method based on a threshold algorithm was used to provide an accurate and automated process to track and as a result, it measured the RBCs flowing in microchannels. The automatic results were in good agreement with the manual method. Further work aims to implement an image analysis application able to track flowing RBCs and, consequently, extract multiple features of the RBCs that can be used in other applications, such as measuring the RBC deformability. Another method based on optical flow was also tested but it is still under development, so that it can be further improved in the future for data collection.

To study the CFL phenomenon in microchannels, the method developed based in the binarization of the image with the maximum intensity evaluation presents some discrepant results when compared to the manual data. Nonetheless, a similar qualitative tendency was observed. In this type of study, the quality of the image sequence plays a crucial role. Hence, by acquiring a sequence of images with higher quality and resolution, we believe that this automatic method can be improved and as a result, it will be able to obtain more accurate results, which should be closer to the ones obtained manually.

**Author Contributions:** The authors have contributed equally to the work. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project has been funded by Portuguese national funds of FCT/MCTES (PIDDAC) through the base funding from the following research units: UIDB/00532/2020 (Transport Phenomena Research Center—CEFT), UIDB/04077/2020 (Mechanical Engineering and Resource Sustainability Center—MEtRICs), UIDB/00690/2020 (CIMO). The authors are also grateful for the partial funding of FCT through the projects, NORTE-01-0145-FEDER-029394 (PTDC/EMD-EMD/29394/2017) and NORTE-01-0145-FEDER-030171 (PTDC/EMD-EMD/30171/2017) funded by COMPETE2020, NORTE2020, PORTUGAL2020 and FEDER. D. Bento acknowledges the PhD scholarship SFRH/BD/ 91192/2012 granted by FCT.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

