*Article* **Age-at-Death Estimation of Fetuses and Infants in Forensic Anthropology: A New "Coupling" Method to Detect Biases Due to Altered Growth Trajectories**

**Mélissa Niel 1,\* , Kathia Chaumoître 1,2 and Pascal Adalian <sup>1</sup>**


**Simple Summary:** In forensic anthropology, estimating the age-at-death of young juvenile skeletons is crucial as a direct determinant of legal issues in many countries. Most methods published for this purpose are based on either maturation or growth processes (two essential components of development) and focus on "normal" (i.e., nonpathological) growth. However, when the osseous remains available for study are from an individual that experienced an altered growth process, age estimation may be biased, and accounting for this would be helpful for potentially avoiding inaccuracies in estimation. In this research, we developed a method based on the combined evaluation of both maturation and growth. Maturation is evaluated by the conformation of the *pars basilaris*, a bone at the skull base that provides an indirect estimate of brain maturation, while growth is assessed using femoral biometry. The method was tested on two medical validation samples of normal and pathological individuals. The results show that it was possible to identify "uncoupling" between maturation and growth in 22.8% of the pathological individuals. Highlighting potential uncoupling is therefore an essential step in assessing the confidence of an age estimate, and its presence should lead experts to be cautious in their conclusions in court.

**Abstract:** The coupling between maturation and growth in the age estimation of young individuals with altered growth processes was analyzed in this study, whereby the age was determined using a geometric morphometrics method. A medical sample comprising 223 fetuses and infants was used to establish the method. The *pars basilaris* shapes, quantified by elliptic Fourier analysis, were grouped into consensus stages to characterize the maturation process along increasing age groups. Each *pars basilaris* maturation stage was "coupled" to biometry by defining an associated femur length range. The method was tested on a validation sample of 42 normal individuals and a pathological sample of 114 individuals whose pathologies were medically assessed. Couplings were present in 90.48% of the normal sample and 77.19% of the pathological sample. The method was able to detect "uncoupling" (i.e., possibly altered growth) in more than 22.8% of samples, even if there was no visible traces of pathology on bones in most cases. In conclusion, experts should be warned that living conditions may cause alterations in the development of young individuals in terms of uncoupling, and that the age-at-death estimation based on long bone biometry could be biased. In a forensic context, when age has been estimated in cases where uncoupling is present, experts should be careful to take potential inaccuracies into account when forming their conclusions.

**Keywords:** forensic anthropology; age estimation; femur length; *pars basilaris* shape; inverse Fourier transform; geometric morphometrics

**Citation:** Niel, M.; Chaumoître, K.; Adalian, P. Age-at-Death Estimation of Fetuses and Infants in Forensic Anthropology: A New "Coupling" Method to Detect Biases Due to Altered Growth Trajectories. *Biology* **2022**, *11*, 200. https://doi.org/ 10.3390/biology11020200

Academic Editors: Ann H. Ross and Eugénia Cunha

Received: 23 December 2021 Accepted: 25 January 2022 Published: 27 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/).

#### **1. Introduction**

Estimating an individual's age-at-death from skeletal remains is one of the major issues in biological and forensic anthropology when assessing a biological profile. In the case of young individual skeletons, age-at-death is crucial to any analysis of biological remains. In forensic anthropology, a fetus's legal personality is dependent on fetal age estimation, with the resulting social, ethical, and economic consequences [1], and the assessment of fetal viability and legislation on abortion and infanticide are also directly dependent on fetal and infant age estimation—hence contributing to the need and importance of developing reliable and accurate methods.

Several fetal and infant age-at-death estimation methods have been established. Most of these are osteometric, radiographic, or ultrasound methods [2–20]. They can be development based, which aim to estimate physiological age based on maturation processes (e.g., skeletal morphology, appearance and maturation of secondary ossification centers, maturation of dental germs), or biometric based, which rely on growth processes (e.g., crown–rump length, cranial and abdominal perimeters, and the maximum length of long bones).

However, the question of living conditions and, therefore, the context in which the development of a young juvenile took place can remain unanswered. Most methods assume that these conditions are "favorable" or "normal", though these can obviously be disturbed by any pathological conditions experienced by the mother or child. In other words, the ontogenetic trajectory—the child's developmental trajectory—is likely to be altered.

It is generally accepted that brain maturation is the best criterion to establish physiological age during early development, regardless of the environmental or socioeconomic conditions, even in cases of fetal or maternal pathologies [21–23]. The brain unfortunately undergoes rapid autolysis after death (within approximately 48 h) and can no longer be studied, but it has an influence on skull base osseous structures [24–31]. Therefore, these structures can be considered to be indirect and taphonomically resistant testaments of brain maturation.

To establish a biometric age, it is accepted that femoral length is the most reliable and accurate estimation indicator [3,7–9,32]. Nevertheless, growth-based age estimation may be biased in cases of growth delay or growth advancement caused by pathological conditions. These conditions are difficult to detect because most pathologies leave little or no trace on fetal and infant bones. Sherwood et al. [32] demonstrated that diseases causing abnormally short femurs (such as trisomy 21 or Turner syndrome) or abnormally long femurs (such as *spina bifida*) can lead to inaccuracies of up to almost four weeks in fetuses when estimating age at death.

Therefore, when only using femoral length without considering possible alterations in developmental conditions, one cannot know whether the age at death will be underestimated, correct, or overestimated with respect to the chronological age (real age).

Our biological hypothesis is that the physiological age (maturation) is more reliable and stable than the biometric age (growth), and that these two "different kinds of ages" are coupled for nonpathological individuals. Accepting this hypothesis, it can be argued that living conditions, whether they are simply "changing" or truly "unfavorable" to development, influence biometric growth more than maturation.

This "coupling" or agreement between maturation and growth processes could be used to assess and control fetal and infant age-at-death estimation, targeting individuals with growth variation due to possible pathological conditions. As a consequence, the demonstration of the "uncoupling" of these two processes would be an indication, or even serve as an alert, that the accuracy of the age-at-death estimation of a young juvenile skeleton must be considered with great caution.

As a direct indicator of skull base maturation and therefore an indirect indicator of brain (and thus general) maturation, we chose to use the *pars basilaris* of the future occipital bone [33–36]. We quantified its degree of maturation with geometric morphometric analyses from its outline shape. The estimation of biometric age (growth) was based on the maximum diaphyseal length of the femur.

These two bones are both dense and compact [11,37,38], and they are generally found to be in good preservation states considering forensic and archaeological contexts [11,37].

Using computerized tomography (CT) scan imaging of fetuses and infants with nonpathological conditions, the aim of our study was to develop a method based on the expected coupling between maturation and growth to detect possible growth variation.

Once established on a medical imaging sample (learning sample) of nonpathological individuals, the method was applied to a separate validation medical sample of nonpathological individuals and another sample of individuals whose pathologies were fully documented.

If an individual presents the "normal" (i.e., nonpathological) coupling variability established by the learning sample, the hypothesis of an alteration of his ontogenetic trajectory can be proposed. It is then necessary to discuss the potential reason for this alteration (growth delay or advancement in connection or otherwise with an identified pathology). Regardless, this study shows that estimated age must be considered with caution.

#### **2. Materials and Methods**

#### *2.1. Sample*

An anonymized database composed of 1136 individuals aged between 11 weeks in utero and 20 years old was compiled within UMR 7268 ADES (AMU-CNRS-EFS). From this, a medical imaging sample of 379 individuals aged 16 weeks in utero to approximately one and a half years (17.7 months) was derived.

#### 2.1.1. Normal and Pathological Development

The studied population was divided into three samples. A learning sample (A) comprising 223 fetuses and infants with nonpathological conditions (77 girls, 115 boys, and 31 of unknown sex) ranging from 16 fetal weeks to 77 postnatal weeks (mean age: 33.28 fetal weeks; Figure 1) was used to establish the method. A second sample (B) comprising 42 fetuses and infants ranging between 18 fetal weeks and 61 postnatal weeks (mean age: 34.69 fetal weeks; Figure 1) was used as a separate validation sample. Given that the available age classes were not homogeneous for normal individuals, random selection by age classes was conducted to ensure a good representation of age; the selection comprised approximately 85% for the learning sample and 15% for the validation sample.

For our analyses, the ages of fetuses (based on accurate reports of the mother's last normal menstrual period and ultrasound data obtained at 10 weeks of gestation, which is an obligatory examination under French law) and infants were expressed in weeks: from 16 to 38 weeks for fetuses and from 39 to 115 weeks for postnatal individuals. This means that a "45-week-old" individual is actually an individual aged 45 weeks minus 38 weeks (average length of pregnancy), which corresponds to 7 postnatal weeks.

Nonpathological conditions were essential for sample A and B individuals. The conditions considered for mothers were the absence of congenital disease, diabetes, or arterial hypertension. The nonpathological conditions of fetuses (such as the absence of external or visceral malformation, the absence of bone anomaly on a CT scan, the absence of cerebral anomaly on MRI, and normal karyotypes) were established based on multidisciplinary ante mortem and post mortem examinations conducted by medical experts of the prenatal diagnosis center. Concerning infants, CT scans allowed us to verify developmental normality. Examinations were performed in cases of road accidents, sudden or unexpected infant death syndrome, and forensic investigations.

Fetuses and infants with identified pathological conditions were included in a third sample (C) comprising 114 fetuses and infants (61 girls, 47 boys, and 6 of unknown sex) ranging from 16 fetal weeks to 47 postnatal weeks (mean age: 27.24 fetal weeks) (Figure 1C). 2 3

18

21

24

27

30

33

36

39

42

45

48

51

54

57

**Validation sample (B)**

60

63

66

69

72

75

78

81

84

87

90

93

96

99

16

19

22

25

28

31

34

37

40

43

46

49

52

55

58

61

64

Age (weeks)

67

70

73

76

79

82

85

88

91

94

97

100

103

106

Girl Boy

109

Unknown sex

112

115

*Biology* **2022**, *10*, x FOR PEER REVIEW 4 of 20

**Learning sample (A)**

of the prenatal diagnosis center. Concerning infants, CT scans allowed us to verify developmental normality. Examinations were performed in cases of road accidents, sudden or unexpected infant death syndrome, and forensic investigations. **Figure 2.** Age (in weeks) and sex distribution of the pathological sample (C) comprising 114 individuals. **Figure 1.** Age (in weeks) and sex distribution of the learning sample (**A**) comprising 223 individuals and the validation sample (**B**) comprising 42 individuals. Age (in weeks) and sex distribution of the pathological sample (**C**) comprising 114 individuals.

disciplinary ante mortem and post mortem examinations conducted by medical experts

#### Fetuses and infants with identified pathological conditions were included in a third sample (C) comprising 114 fetuses and infants (61 girls, 47 boys, and 6 of unknown sex) 2.1.2. Pathologies Groups 2.1.2. Pathologies Groups

ranging from 16 fetal weeks to 47 postnatal weeks (mean age: 27.24 fetal weeks) (Figure Depending on the pathological conditions, the following subgroups were established: Depending on the pathological conditions, the following subgroups were established:


Girl


Age (weeks)

16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 The same individual could be classified in several types of pathologies, such as a localized anomaly and a cerebral anomaly.

#### *2.2. Data Acquisition*

> **Figure 2.** Age (in weeks) and sex distribution of the pathological sample (C) comprising 114 individuals. 2.1.2. Pathologies Groups Depending on the pathological conditions, the following subgroups were established: The ante mortem and post mortem CT scans of sample A, B, and C individuals were collected from the Picture Archiving and Communication System (PACS) in the hospital of Marseilles (Assistance Publique—Hôpitaux de Marseille, France). Individuals were scanned using a helical CT scanner (Somatom Sensation Cardiac 64; Siemens, Erlangen, Germany). The scanning parameters were as follows: voltage of 100–140 kVp, amperage of 50–180 mAs, 512 × 512 pixels, resolution of 0.25–4.87 pixels per mm, voxel size of approximatively 0.5 <sup>×</sup> 0.5 <sup>×</sup> 0.6 or 1 mm<sup>3</sup> , and a slice thickness of 0.6–1 mm. These highresolution native slices recorded in the Digital Imaging and Communications in Medicine

(DICOM) format were anonymized before being used in the study, in accordance with the standards of the French National Consultative Ethics Committee for health and life sciences (CCNE) and the Helsinki Declaration of 1975 concerning the privacy and confidentiality of personal data.

#### *2.3. Bone Reconstruction*

Before reconstructing the femur and *pars basilaris* in three dimensions (3D), region of interest (ROI) segmentation on the DICOM slices was performed with the ImageJ®v1.51 software (National Institutes of Health, Bethesda, MD, USA) to separate the bone from adjacent tissues. The threshold value was obtained by calculating a threshold mean value (TMV) [38], which is an average of the half-maximum height (HMH) values [39]. The TMV was used in Avizo Standard Edition (v.7.0.0®, Visualization Sciences Group, SAS, Berlin, Germany) to reconstruct the 3D bone surfaces.

Since there are no significant differences between the right and left femur in young juveniles [3,9,10,40–42] and convention suggests that the left femur is preferred, we only measured the right femur when the left was not available.

#### *2.4. Maturation Criterion: Elliptic Fourier Analysis of the Pars Basilaris*

The complete protocol was described by Niel et al. [43] and was used in this study.

#### 2.4.1. Outline Process

Briefly, we defined a homologous reference plane for all the *pars basilaris* in the inferior (external) view. This was defined thanks to two type II and one type III landmarks [44]. Type II landmarks are the most posterior point of the left and right horns, and a type III landmark is the central point of the anterior surface. All landmarks were digitized on 3D reconstructed surfaces using Avizo Standard Edition® software.

This step allowed us to project all reconstructions in the same 2D plane and with the same orientation. Then, outline shapes were quantified according to 150 equally linearly spaced points digitized along the *pars basilaris* contour with the tpsDig2 v.2.17® digitization program [45]. Finally, the contour data of the *pars basilaris* were normalized using generalized Procrustes analysis (GPA) [46–49] based on four type II and III homologous landmarks [44] called control points [46].

#### 2.4.2. Measurement Error

Repeatability (intra-observer error) and reproducibility tests (inter-observer error) were realized to validate the protocol on 30 randomly selected individuals in sample A. Repeatability was tested by the same observer repeating the protocol twice several weeks apart; for reproducibility, a second observer applied the protocol once.

#### 2.4.3. Harmonics Number

With EFA, one may wonder what the appropriate number of harmonics is, since this number determines the accuracy of the contour reconstruction. The following two paragraphs of text is the explanation as reproduced from Niel et al. pp. 37–38 [43]:

According to the Nyquist theorem [50], the harmonic number must be less than half the number of sampled outline points. Consequently, on the 150 points sampled for EFA, only the first 74 harmonics were retained for analysis. Given that we cannot retain all the Fourier coefficients for our analysis" (74 harmonics × 4 coefficients = 296 coefficients), because the measurement error is expected to increase with harmonic ranks, the percentage of error on harmonic coefficients was calculated using a Procrustes analysis of variance (ANOVA) on the three sessions [51]. This procedure calculated the mean sums of squares for the four coefficients of each harmonic to observe the evolution of error according to the rank of the harmonics (in percentage). Only the first harmonics, showing an acceptable digitization error rate, were retained for further analyses. An error rate under 35% is considered to be reasonable in an outline analysis using EFA [51].

The assessment of the total percentage of measurement error was then performed using a Procrustes ANOVA [51–55] adapted to elliptic Fourier coefficients [51]. The Fourier coefficients of the coupled series are used in the Procrustes ANOVA with the number of harmonics previously defined. The intra- and interindividual variances were directly calculated from the means of the sums of squares and crossed products corresponding to individuals and residual sources of variation [51]. These residuals, representing the variability between the two sessions, correspond to the measurement error [55].
