1. Introduction
Pulmonary nodules are defined as well- or poorly defined radiographic opacities with a diameter ≤ 3 cm, surrounded by lung tissue [
1,
2]. Incidental pulmonary nodules are frequently found on routinely performed computed tomography (CT) scans of the chest. In the United States, they are detected in approximately 30% of chest CT [
3]. In addition, lung cancer screening programs using low-dose chest CT (LDCT) have become more widespread, with detection rates as high as 51% [
4,
5]. Most studies report that less than 10% of nodules are malignant [
6]. The challenge is to distinguish between the more common benign nodules that do not require follow-up and the much rarer malignant nodules that need immediate treatment to achieve survival benefits.
Increased attention has also been given to the use of magnetic resonance imaging (MRI) in the detection and screening of malignant nodules in recent years [
7,
8,
9]. Turbo spin echo (SE)-based and gradient echo (GRE)-based techniques are important MRI sequences that are widely used to detect pulmonary nodules [
10]. Motion artifacts due to respiration and heartbeat pose a particular challenge. The problem is currently addressed using a respiratory-navigated sequence with radial k-space acquisition, commonly known as MultiVane (Philips, Best, The Netherlands), fBLADE (Siemens, Erlangen, Germany), or PROPELLER (GE, Milwaukee, WI, USA), which provides excellent T2 contrast as it is insensitive to cardiac motion [
11]. Besides the obvious advantage of radiation absence, studies concluded that MRI has similar sensitivity for nodule detection, with sensitivity and specificity even higher in malignant than in benign lesions compared to CT [
12,
13,
14,
15]. However, it remains a question of whether an MRI of the lungs could be used as a stand-alone diagnostic tool or in combination with CT to classify pulmonary nodules.
Artificial intelligence has a long history in the field of pulmonary nodule detection and classification dating back to the 1960s [
16,
17]. In 2012, Lambin and colleagues coined the term radiomics to describe quantitative imaging feature extraction to achieve better diagnostic performance [
18]. Pathological studies have demonstrated increased heterogeneity within malignant pulmonary nodules, which are not visible to the naked eye on radiological examination but can be quantified with radiomics [
19,
20,
21]. Therefore, the purpose of this study was to compare the performance of radiomic features analysis and radiologists in classifying pulmonary nodules based on stand-alone and complementary CT and MRI imaging.
2. Materials and Methods
2.1. Study Design and Sample
The study was approved by the medical–ethical committee, and informed consent was waived because of the retrospective collection of study data (RWTH Aachen University Hospital, Aachen, Germany). The study was performed in accordance with relevant guidelines and regulations and contemporary data protection laws. The entire cohort dataset was acquired from April 2019 to February 2022, including institutional picture archiving and communication system records (IntelliSpace PACS; Philips, Best, The Netherlands), using a standardized query for patients with pulmonary nodules who underwent standard-dose contrast-enhanced or non-enhanced chest CT scans and non-enhanced lung MRI. A radiologist with 5 years of experience in thoracal imaging screened patients with at least one pulmonary nodule who had undergone surgical resection and histopathologic examination of the lesion to determine benignity/malignancy. No distinction was made between primary or secondary lung malignancy. Exclusion criteria were as follows: (a) CTs with a slice thickness of >3 mm; (b) MRIs with incomplete or missing axial diffusion-(DWI) and T2-weighted (T2w) sequences; (c) pulmonary nodules with unclear histopathological results. If multiple pulmonary nodes were detected on CT, only histopathologic examined nodules were included in the study. For each patient, characteristics such as age, sex, average diameter, and pulmonary lobe were determined.
Figure 1 provides an overview of the inclusion and exclusion criteria.
2.2. CT Parameters
All chest CTs were performed with 128-row spiral CT scanners (Somatom x.Site or Somatom Force, Siemens Medical Systems, Erlangen, Germany). The scans were acquired in a craniocaudal direction during a single-breath-hold either contrast-enhanced or non-enhanced. If contrast-enhanced CT was performed, a 1.0 mL/kg body weight bolus of iopromide 370 mg/mL (Ultravist, Bayer, Leverkusen, Germany) was injected intravenously by a power injector with an acquisition time of 75 s.
Table 1 shows further technical details.
2.3. MRI Parameters
All chest MRIs were performed according to a standardized protocol using a 1.5 Tesla MRI system (Ambition or Ingenia, Philips, Best, The Netherlands) with a 32-element body surface coil. The standardized protocol contained axial and coronal T2w MultiVane-XD (MVXD), axial diffusion-weighted spin echo (SE), and axial T2w turbo spin echo (TSE) with and without fat saturation.
Table 1 describes the detailed parameters of the pulse sequences used for further analysis in this study.
2.4. Image Segmentation
Two-dimensional manual segmentation in axial orientation was performed by a radiologist with 5 years of experience in thoracal imaging, using ITK-SNAP 3.6.0 (
www.itksnap.org accessed on 26 July 2021) [
22]. The segmentations delineated the visible borders of each pulmonary nodule in the pulmonary window of CT images, in T2w MVXD sequences, and in DWI SE sequences at a b-value of 800 s/mm
2. In DWI SE sequences, a second segmentation of similar size was drawn in the dorsal subcutaneous fat tissue for harmonization purposes. All segmentations were validated by a senior radiologist with 12 years of experience in thoracal CTs.
2.5. Radiomic Features
Radiomic features were extracted using a PyRadiomics 3.0.1 framework [
23] and based on feature definitions described by the Imaging Biomarker Standardization Initiative (IBSI) [
24]. They included first-order statistical features, shape-based features, and texture features (gray level co-occurrence matrix, gray level run length matrix, gray level size zone matrix, neighboring gray-tone difference matrix, and gray level dependence matrix). A total of 105 radiomic features were extracted from each segmentation.
2.6. Development of a Prediction Model
Firstly, the radiomic features of CT imaging were evaluated for their value by differentiating pulmonary nodules into benign or malignant. To reduce the number of features, backward selection was employed, i.e., the prediction model was tasked to reduce the 105 features one by one, iteratively eliminating features with the lowest discriminatory value until only six features remained. To account for data scarcity, patient-by-patient leave-one-out cross-validation (LOOCV) was used. That is, the prediction model was trained repeatedly by leaving out one patient from the training set and training with the remaining set of patients until an independent prediction could be obtained for each patient. The backward selection process was then repeated for radiomic features in T2w MVXD sequences and again for DWI SE sequences at a b-value of 800 s/mm2. The three features with the highest discriminatory value resulting from separate training with T2w MVXD and DWI SE segmentations were then combined. The total accuracy of the six features was evaluated in a test set consisting of images corresponding to T2w and DWI MRI scans. Lastly, the two features with the highest discrimination value resulting from separate training with segmentations from T2w MVXD, DWI SE, and CT imaging were combined, and the total accuracy of the six features was evaluated in a test set consisting of images corresponding to CT, T2w, and DWI scans.
2.7. Human Reader Analysis
For comparative purposes, the same 2D image slices used for radiomic feature extraction were presented to three radiologists with 4, 8, and 12 years of experience in evaluating pulmonary nodules. Consistent with the radiomic analysis, the radiologists were presented with the standalone CT, T2w, and DWI datasets, followed by image sets corresponding to T2w and DWI as well as CT, T2w, and DWI, each in random order. They were requested to classify each nodule as benign or malignant. In the event of an interrater discrepancy, the majority vote was recorded.
2.8. Statistical Analysis
All the statistical analyses were performed using the Python packages SciPy 1.7.0 [
25] and NumPy 1.21.0 [
26]. Confusion matrices were calculated for each model. The performance of radiomics and human readers was compared by calculating accuracy, sensitivity, specificity, positive predictive value, and negative predictive value with Clopper–Pearson confidence intervals.
3. Results
In this study, 57 patients with ≥1 pulmonary nodule, chest CT and MRI scans, and a histopathologic workup of the nodule were screened for eligibility. The final group comprised 50 patients with a mean age of 63 years with a standard deviation of 10 years. Female patients totaled 18/50 (36%). Within this group, 66 pulmonary nodules were found with an average diameter of 0.9 cm. There was a slightly higher incidence in the right lung with 17 (26%) in the right upper lobe, 7 (11%) in the right middle lobe, and 12 (18%) in the right lower lobe compared to 18 (27%) in the left upper lobe and 12 (18%) in the left lower lobe. Of the 66 pulmonary nodules, 26 (39%) were benign and 40 (61%) were malignant, including 16 (24%) primary lung carcinomas and 24 (36%) solitary lung metastases.
Table 2 summarizes the epidemiologic, clinical, and histological characteristics.
All 66 pulmonary nodules were visible to radiologists on CT. By contrast, only 61 (92.4%) of the nodules were visible in the T2w-weighted axial MultiVane XD sequence, and the five nodules that were not detectable turned out to be benign. Only 42 (63.6%) of the nodules were recovered in the axial diffusion-weighted spin echo sequence, with nodule detection in DWI associated with an increased likelihood of malignancy. The detection rate is summarized in
Table 3.
Figure 2 shows illustrative examples of CT images and T2-weighted and DWI MRI sequences of a benign nodule, primary lung cancer, and solitary lung metastasis.
The backward selection process redacted the initial 105 radiomic features down to six features with the highest discriminatory values for benign and malignant in each image dataset (CT, T2w sequence, and DWI sequence), as shown in
Table 4. For the combined analysis of T2w and DWI datasets, the top three radiomic features from the T2w and DWI datasets alone were used. For the combined analysis of CT, T2w, and DWI datasets, the top two radiomic features from the datasets of CT alone, T2w alone, and DWI alone were used, respectively.
A description of radiomic features with the highest discriminatory values found in each dataset is shown in
Table 5.
The radiomic features analysis of the separate modalities and sequences showed the highest accuracy for the CT dataset (ACC 0.68; 95% CI: 0.56, 0.79), followed by the T2w (ACC 0.65; 95% CI: 0.52, 0.77) and DWI datasets (ACC 0.61; 95% CI: 0.48, 0.72). Human readers achieved the highest accuracy based on DWI (ACC 0.73; 95% CI: 0.60, 0.83), followed by T2w (ACC 0.68; 95% CI: 0.54, 0.78) and CT (ACC 0.59; 95% CI: 0.46, 0.71). When the T2w and DWI datasets were available for combined radiomic features analysis, the accuracy increased slightly (ACC 0.73; 95% CI: 0.60, 0.83) and when CT and MRI image data were included, the accuracy of radiomic features analysis further increased (ACC 0.83; 95% CI: 0.72, 0.91). By contrast, radiologists’ accuracy remained essentially the same for the combined image information of T2w and DWI (ACC 0.70; 95% CI: 0.57, 0.80) compared to T2w or DWI alone. Given the combined image information from CT, T2W, and DWI, radiologists’ accuracy displayed no improvement (ACC 0.64; 95% CI: 0.51–0.75).
Table 6 provides a detailed comparison between radiomic analysis and radiologists’ results.
4. Discussion
Differentiating malignant from benign pulmonary nodules is a common diagnostic challenge for radiologists. Recent advances in MRI for evaluating pulmonary nodules have been increasingly used as a complementary or even stand-alone imaging modality to computed tomography. At the same time, machine-learning tools have also been presented as supporting tools for radiologists. Therefore, our objective was to compare the performance of human readers to radiomic feature analysis based on stand-alone and complementary CT and MRI imaging in classifying pulmonary nodules.
The study results show that the accuracy of radiomic feature analysis can increase if a combination of CT, T2w, and DWI is used (ACC 0.83) instead of CT (ACC 0.68), T2w (ACC 0.65), or DWI (ACC 0.61) alone. Interestingly, combining CT and MRI image datasets (ACC 0.64) did not significantly improve accuracy in human readers compared to CT (ACC 0.59), T2w (ACC 0.68), or DWI (ACC 0.73) alone. Each dataset consisted of images depicting the same 66 lung lesions of which 26 were benign and 40 were malignant (16 primary lung cancer and 24 solitary lung metastases). The mean diameter of the pulmonary nodules was 0.9 cm (SD 0.4 cm), and the patients were 65 years old (SD 10 years) on average. Based on these results, the supportive use of radiomic analysis in multimodal CT and MRI assessment of pulmonary nodules should be considered and further investigated to potentially improve radiologists’ assessments and decrease unnecessary biopsy rates or resections.
Many studies have already demonstrated the value of analyzing radiomic features in CT and MRI datasets in evaluating pulmonary nodules [
27,
28,
29,
30,
31,
32]. This study is the first to show that by combining quantitative image information from CT, T2w, and DWI datasets in a backward selection process, assessing pulmonary nodules through radiomics analysis is superior to that of one modality alone, even exceeding human readers’ performance.
A decisive factor for the comparatively low performance of human readers in this study is probably the small size of the nodules, with a mean of 0.9 cm. Since small nodules generally have a regular and compact shape, shape features that are easier for the human eye to recognize play a subordinate role, whereas texture features are of greater importance [
33]. These observations are also reflected in the final six feature radiomics set found in this study. In this set, there are four first-order features concerning intensities and texture, one GLCM feature, one GLRLM feature, and not a single shape feature. Therefore, the transferability of the study results to larger nodules is limited, and results should be viewed in the context of small pulmonary nodules.
Another important limitation of this study is that we used data from only one institution. In future studies, the true predictive power of the current method should be assessed with an independent test dataset. Further limitations include the retrospective nature of the study and the small dataset of pulmonary nodules.
5. Conclusions
Quantitative image information from axial CT datasets or the DWI- or T2-weighted MRI datasets alone allows the assessment of pulmonary nodules by radiomics analysis compared to human readers’ performance. By providing image information from CT and MRI sequences, radiomics analysis is better than using a single modality and even surpasses the performance of human readers. Therefore, complementary CT and MRI assessment by radiomic features analysis can potentially reduce unnecessary biopsies or resections.
Author Contributions
Conceptualization, E.T., M.Z., C.K.K., S.K. and D.T.; methodology, E.T., S.N. and D.T.; software, G.M.-F.; validation, E.T. and D.T.; formal analysis, E.T. and G.M.-F.; investigation, E.T.; resources, C.K.K., S.N. and D.T.; data curation, E.T., M.Z., S.K. and D.T.; writing—original draft preparation, E.T.; writing—review and editing, E.T. and D.T.; visualization, E.T.; supervision, C.K.K.; project administration, S.N. and D.T.; funding acquisition, D.T. All authors have read and agreed to the published version of the manuscript.
Funding
This research received no external funding.
Institutional Review Board Statement
The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of RWTH Aachen University Hospital (EK 028-19; 3 May 2019).
Informed Consent Statement
Patient consent was waived due to the retrospective nature of this study.
Data Availability Statement
The data presented in this study are available upon request from the corresponding author.
Conflicts of Interest
The authors declare no conflicts of interest.
References
- MacMahon, H.; Naidich, D.P.; Goo, J.M.; Lee, K.S.; Leung, A.N.C.; Mayo, J.R.; Mehta, A.C.; Ohno, Y.; Powell, C.A.; Prokop, M.; et al. Guidelines for Management of Incidental Pulmonary Nodules Detected on CT Images: From the Fleischner Society 2017. Radiology 2017, 284, 228–243. [Google Scholar] [CrossRef]
- Ost, D.; Fein, A.M.; Feinsilver, S.H. Clinical practice. The solitary pulmonary nodule. N. Engl. J. Med. 2003, 348, 2535–2542. [Google Scholar] [CrossRef]
- Gould, M.K.; Tang, T.; Liu, I.-L.A.; Lee, J.; Zheng, C.; Danforth, K.N.; Kosco, A.E.; Di Fiore, J.L.; Suh, D.E. Recent Trends in the Identification of Incidental Pulmonary Nodules. Am. J. Respir. Crit. Care Med. 2015, 192, 1208–1214. [Google Scholar] [CrossRef]
- Aberle, D.R.; Adams, A.M.; Berg, C.D.; Black, W.C.; Clapp, J.D.; Fagerstrom, R.M.; Gareen, I.F.; Gatsonis, C.; Marcus, P.M.; Sicks, J.D. Reduced lung-cancer mortality with low-dose computed tomographic screening. N. Engl. J. Med. 2011, 365, 395–409. [Google Scholar] [CrossRef]
- Swensen, S.J.; Jett, J.R.; Hartman, T.E.; Midthun, D.E.; Sloan, J.A.; Sykes, A.-M.; Aughenbaugh, G.L.; Clemens, M.A. Lung cancer screening with CT: Mayo Clinic experience. Radiology 2003, 226, 756–761. [Google Scholar] [CrossRef] [PubMed]
- Bach, P.B.; Mirkin, J.N.; Oliver, T.K.; Azzoli, C.G.; Berry, D.A.; Brawley, O.W.; Byers, T.; Colditz, G.A.; Gould, M.K.; Jett, J.R.; et al. Benefits and harms of CT screening for lung cancer: A systematic review. JAMA 2012, 307, 2418–2429. [Google Scholar] [CrossRef] [PubMed]
- Biederer, J.; Ohno, Y.; Hatabu, H.; Schiebler, M.L.; van Beek, E.J.R.; Vogel-Claussen, J.; Kauczor, H.-U. Screening for lung cancer: Does MRI have a role? Eur. J. Radiol. 2017, 86, 353–360. [Google Scholar] [CrossRef]
- Meier-Schroers, M.; Homsi, R.; Skowasch, D.; Buermann, J.; Zipfel, M.; Schild, H.H.; Thomas, D. Lung cancer screening with MRI: Results of the first screening round. J. Cancer Res. Clin. Oncol. 2018, 144, 117–125. [Google Scholar] [CrossRef] [PubMed]
- Sanchez, F.; Tyrrell, P.N.; Cheung, P.; Heyn, C.; Graham, S.; Poon, I.; Ung, Y.; Louie, A.; Tsao, M.; Oikonomou, A. Detection of solid and subsolid pulmonary nodules with lung MRI: Performance of UTE, T1 gradient-echo, and single-shot T2 fast spin echo. Cancer Imaging 2023, 23, 17. [Google Scholar] [CrossRef] [PubMed]
- Ohno, Y.; Kauczor, H.-U.; Hatabu, H.; Seo, J.B.; van Beek, E.J.R. MRI for solitary pulmonary nodule and mass assessment: Current state of the art. J. Magn. Reson. Imaging 2018, 47, 1437–1458. [Google Scholar] [CrossRef] [PubMed]
- Raptis, C.A.; Ludwig, D.R.; Hammer, M.M.; Luna, A.; Broncano, J.; Henry, T.S.; Bhalla, S.; Ackman, J.B. Building blocks for thoracic MRI: Challenges, sequences, and protocol design. J. Magn. Reson. Imaging 2019, 50, 682–701. [Google Scholar] [CrossRef] [PubMed]
- Koyama, H.; Ohno, Y.; Aoyama, N.; Onishi, Y.; Matsumoto, K.; Nogami, M.; Takenaka, D.; Nishio, W.; Ohbayashi, C.; Sugimura, K. Comparison of STIR turbo SE imaging and diffusion-weighted imaging of the lung: Capability for detection and subtype classification of pulmonary adenocarcinomas. Eur. Radiol. 2010, 20, 790–800. [Google Scholar] [CrossRef] [PubMed]
- Koyama, H.; Ohno, Y.; Kono, A.; Takenaka, D.; Maniwa, Y.; Nishimura, Y.; Ohbayashi, C.; Sugimura, K. Quantitative and qualitative assessment of non-contrast-enhanced pulmonary MR imaging for management of pulmonary nodules in 161 subjects. Eur. Radiol. 2008, 18, 2120–2131. [Google Scholar] [CrossRef] [PubMed]
- Sommer, G.; Tremper, J.; Koenigkam-Santos, M.; Delorme, S.; Becker, N.; Biederer, J.; Kauczor, H.-U.; Heussel, C.P.; Schlemmer, H.-P.; Puderbach, M. Lung nodule detection in a high-risk population: Comparison of magnetic resonance imaging and low-dose computed tomography. Eur. J. Radiol. 2014, 83, 600–605. [Google Scholar] [CrossRef]
- Yang, S.; Shan, F.; Shi, Y.; Liu, T.; Wang, Q.; Zhang, H.; Zhang, X.; Yang, S.; Zhang, Z. Sensitivity and specificity of magnetic resonance imaging in routine diagnosis of pulmonary lesions: A comparison with computed tomography. J. Thorac. Dis. 2022, 14, 3762–3772. [Google Scholar] [CrossRef]
- Lodwick, G.S.; Keats, T.E.; Dorst, J.P. The Coding of Roentgen Images for Computer Analysis as Applied to Lung Cancer. Radiology 1963, 81, 185–200. [Google Scholar] [CrossRef]
- Zheng, X.; He, B.; Hu, Y.; Ren, M.; Chen, Z.; Zhang, Z.; Ma, J.; Ouyang, L.; Chu, H.; Gao, H.; et al. Diagnostic Accuracy of Deep Learning and Radiomics in Lung Cancer Staging: A Systematic Review and Meta-Analysis. Front. Public Health 2022, 10, 938113. [Google Scholar] [CrossRef]
- Lambin, P.; Rios-Velazquez, E.; Leijenaar, R.; Carvalho, S.; van Stiphout, R.G.P.M.; Granton, P.; Zegers, C.M.L.; Gillies, R.; Boellard, R.; Dekker, A.; et al. Radiomics: Extracting more information from medical images using advanced feature analysis. Eur. J. Cancer 2012, 48, 441–446. [Google Scholar] [CrossRef] [PubMed]
- Mitra, S.; Shankar, B.U. Integrating Radio Imaging with Gene Expressions Toward a Personalized Management of Cancer. IEEE Trans. Hum.-Mach. Syst. 2014, 44, 664–677. [Google Scholar] [CrossRef]
- Wilson, R.; Devaraj, A. Radiomics of pulmonary nodules and lung cancer. Transl. Lung Cancer Res. 2017, 6, 86–91. [Google Scholar] [CrossRef] [PubMed]
- Wan, Q.; Deng, Y.-S.; Lei, Q.; Bao, Y.-Y.; Wang, Y.-Z.; Zhou, J.-X.; Zou, Q.; Li, X.-C. Differentiating between malignant and benign solid solitary pulmonary lesions: Are intravoxel incoherent motion and diffusion kurtosis imaging superior to conventional diffusion-weighted imaging? Eur. Radiol. 2019, 29, 1607–1615. [Google Scholar] [CrossRef]
- Yushkevich, P.A.; Piven, J.; Hazlett, H.C.; Smith, R.G.; Ho, S.; Gee, J.C.; Gerig, G. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. Neuroimage 2006, 31, 1116–1128. [Google Scholar] [CrossRef] [PubMed]
- van Griethuysen, J.J.M.; Fedorov, A.; Parmar, C.; Hosny, A.; Aucoin, N.; Narayan, V.; Beets-Tan, R.G.H.; Fillion-Robin, J.-C.; Pieper, S.; Aerts, H.J.W.L. Computational Radiomics System to Decode the Radiographic Phenotype. Cancer Res. 2017, 77, e104–e107. [Google Scholar] [CrossRef]
- Zwanenburg, A.; Vallières, M.; Abdalah, M.A.; Aerts, H.J.W.L.; Andrearczyk, V.; Apte, A.; Ashrafinia, S.; Bakas, S.; Beukinga, R.J.; Boellaard, R.; et al. The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping. Radiology 2020, 295, 328–338. [Google Scholar] [CrossRef] [PubMed]
- Virtanen, P.; Gommers, R.; Oliphant, T.E.; Haberland, M.; Reddy, T.; Cournapeau, D.; Burovski, E.; Peterson, P.; Weckesser, W.; Bright, J.; et al. SciPy 1.0: Fundamental algorithms for scientific computing in Python. Nat. Methods 2020, 17, 261–272. [Google Scholar] [CrossRef] [PubMed]
- Harris, C.R.; Millman, K.J.; van der Walt, S.J.; Gommers, R.; Virtanen, P.; Cournapeau, D.; Wieser, E.; Taylor, J.; Berg, S.; Smith, N.J.; et al. Array programming with NumPy. Nature 2020, 585, 357–362. [Google Scholar] [CrossRef]
- Xu, Q.-Q.; Shan, W.-L.; Zhu, Y.; Huang, C.-C.; Bao, S.-Y.; Guo, L.-L. Prediction efficacy of feature classification of solitary pulmonary nodules based on CT radiomics. Eur. J. Radiol. 2021, 139, 109667. [Google Scholar] [CrossRef]
- Liu, A.; Wang, Z.; Yang, Y.; Wang, J.; Dai, X.; Wang, L.; Lu, Y.; Xue, F. Preoperative diagnosis of malignant pulmonary nodules in lung cancer screening with a radiomics nomogram. Cancer Commun. 2020, 40, 16–24. [Google Scholar] [CrossRef]
- Koo, C.W.; Lu, A.; Takahashi, E.A.; Simmons, C.L.; Geske, J.R.; Wigle, D.; Peikert, T. Can MRI contribute to pulmonary nodule analysis? J. Magn. Reson. Imaging 2019, 49, e256–e264. [Google Scholar] [CrossRef]
- Li, B.; Li, Q.; Chen, C.; Guan, Y.; Liu, S. A systematic review and meta-analysis of the accuracy of diffusion-weighted MRI in the detection of malignant pulmonary nodules and masses. Acad. Radiol. 2014, 21, 21–29. [Google Scholar] [CrossRef]
- Koo, C.W.; Kline, T.L.; Yoon, J.H.; Vercnocke, A.J.; Johnson, M.P.; Suman, G.; Lu, A.; Larson, N.B. Magnetic resonance radiomic feature performance in pulmonary nodule classification and impact of segmentation variability on radiomics. Br. J. Radiol. 2022, 95, 20220230. [Google Scholar] [CrossRef] [PubMed]
- Chen, C.-H.; Chang, C.-K.; Tu, C.-Y.; Liao, W.-C.; Wu, B.-R.; Chou, K.-T.; Chiou, Y.-R.; Yang, S.-N.; Zhang, G.; Huang, T.-C. Radiomic features analysis in computed tomography images of lung nodule classification. PLoS ONE 2018, 13, e0192002. [Google Scholar] [CrossRef] [PubMed]
- Xu, Y.; Lu, L.; E, L.-N.; Lian, W.; Yang, H.; Schwartz, L.H.; Yang, Z.-H.; Zhao, B. Application of Radiomics in Predicting the Malignancy of Pulmonary Nodules in Different Sizes. Am. J. Roentgenol. 2019, 213, 1213–1220. [Google Scholar] [CrossRef] [PubMed]
| Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 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/).