Sex estimation from coxal bones using deep learning in a population balanced by sex and age.

In the field of forensic anthropology, researchers aim to identify anonymous human remains and determine the cause and circumstances of death from skeletonized human remains. Sex determination is a fundamental step of this procedure because it influences the estimation of other traits, such as age and stature. Pelvic bones are especially dimorphic, and are thus the most useful bones for sex identification. Sex estimation methods are usually based on morphologic traits, measurements, or landmarks on the bones. However, these methods are time-consuming and can be subject to inter- or intra-observer bias. Sex determination can be done using dry bones or CT scans. Recently, artificial neural networks (ANN) have attracted attention in forensic anthropology. Here we tested a fully automated and data-driven machine learning method for sex estimation using CT-scan reconstructions of coxal bones. We studied 580 CT scans of living individuals. Sex was predicted by two networks trained on an independent sample: a disentangled variational auto-encoder (DVAE) alone, and the same DVAE associated with another classifier (Crecon). The DVAE alone exhibited an accuracy of 97.9%, and the DVAE + Crecon showed an accuracy of 99.8%. Sensibility and precision were also high for both sexes. These results are better than those reported from previous studies. These data-driven algorithms are easy to implement, since the pre-processing step is also entirely automatic. Fully automated methods save time, as it only takes a few minutes to pre-process the images and predict sex, and does not require strong experience in forensic anthropology.

Sex estimation from long bones: a machine learning approach.

Sex estimation from skeletal remains is one of the crucial issues in forensic anthropology. Long bones can be a valid alternative to skeletal remains for sex estimation when more dimorphic bones are absent or degraded, preventing any estimation from the first intention methods. The purpose of this study was to generate and compare classification models for sex estimation based on combined measurement of long bones using machine learning classifiers. Eighteen measurements from four long bones (radius, humerus, femur, and tibia) were taken from a total of 2141 individuals. Five machine learning methods were employed to predict the sex: a linear discriminant analysis (LDA), penalized logistic regression (PLR), random forest (RF), support vector machine (SVM), and artificial neural network (ANN). The different classification algorithms using all bones generated highly accuracy models with cross-validation, ranging from 90 to 92% on the validation sample. The classification with isolated bones ranked between 83.3 and 90.3% on the validation sample. In both cases, random forest stands out with the highest accuracy and seems to be the best model for our investigation. This study upholds the value of combined long bones for sex estimation and provides models that can be applied with high accuracy to different populations.

Morphometric analysis of dry atlas vertebrae in a northeastern Thai population and possible correlation with sex.

PURPOSE: The uppermost segment of the cervical vertebra or atlas (C1) is a critically important anatomical structure, housing the medulla oblongata and containing the grooves for the C1 spinal nerve and the vertebral vessels. Variations of the C1 vertebra can affect upper spine stability, and morphometric parameters have been reported to differ by population. However, there are few data regarding these parameters in Thais. The use of this bone to predict sex and age has never been reported. METHODS: This study aimed to examine C1 morphometry and determine its ability to predict sex. Twelve diameter parameters were taken from the C1 vertebrae of identified skeletons (n = 104, males [n, 54], females [n, 50]). Correlation analysis was also performed for sex and age, which were predicted using machine learning algorithms. RESULTS: The results showed that 8 of the 12 measured parameters were significantly longer in the male atlas (p < 0.05), while the remaining 4 (distance between both medial-most edges of the transverse foramen, transverse dimension of the superior articular surface, frontal plane passing through the canal's midpoint, and anteroposterior dimension of the inferior articular surface) did not differ significantly by sex. There was no statistically significant difference in these parameters on the lateral side. The decision stump classifier was trained on C1 parameters, and the resulting model could predict sex with 82.6% accuracy (root mean square error = 0.38). CONCLUSION: Assertation of the morphometric parameters of the atlas is important for preoperative assessment, especially for the treatment of atlas dislocation. Our findings also highlighted the potential use of atlas measurements for sex prediction.

Machine learning and discriminant function analysis in the formulation of generic models for sex prediction using patella measurements.

Sex prediction from bone measurements that display sexual dimorphism is one of the most important aspects of forensic anthropology. Some bones like the skull and pelvis display distinct morphological traits that are based on shape. These morphological traits which are sexually dimorphic across different population groups have been shown to provide an acceptably high degree of accuracy in the prediction of sex. A sample of 100 patella of Mixed Ancestry South Africans (MASA) was collected from the Dart collection. Six parameters: maximum height (maxh), maximum breadth (maxw), maximum thickness (maxt), the height of articular facet (haf), lateral articular facet breadth (lafb), and medial articular facet breath (mafb) were used in this study. Stepwise and direct discriminant function analyses were performed for measurements that exhibited significant differences between male and female mean measurements, and the "leave-one-out" approach was used for validation. Moreover, we have used eight classical machine learning techniques along with feature ranking techniques to identify the best feature combinations for sex prediction. A stacking machine learning technique was trained and validated to classify the sex of the subject. Here, we have used the top performing three ML classifiers as base learners and the predictions of these models were used as inputs to different machine learning classifiers as meta learners to make the final decision. The measurements of the patella of South Africans are sexually dimorphic and this observation is consistent with previous studies on the patella of different countries. The range of average accuracies obtained for pooled multivariate discriminant function equations is 81.9-84.2%, while the stacking ML technique provides 90.8% accuracy which compares well with those presented for previous studies in other parts of the world. In conclusion, the models proposed in this study from measurements of the patella of different population groups in South Africa are useful resent with reasonably high average accuracies.

Sex prediction with morphometric measurements of first and fifth metatarsal and phalanx obtained from X-ray images by using machine learning algorithms.

BACKGROUND: The aim of this study is to predict sex with machine learning (ML) algorithms by making morphometric measurements on radiological images of the first and fifth metatarsal and phalanx bones. MATERIALS AND METHODS: In this study, radiologic images of 263 individuals (135 female, 128 male) between the ages of 27 and 60 were analysed retrospectively. The images in digital imaging and communications in medicine (DICOM) format were transferred to personal workstation Radiant DICOM Viewer programme. Length and width measurements of the first and fifth metatarsal and foot phalanx bones were performed on the transferred images. In addition, the ratios of the total length of the first proximal and distal phalanx and length of the first metatarsal and total length of fifth proximal, middle, and distal phalanx and maximum length of fifth metatarsal were calculated. RESULTS: As a result of machine learning algorithms, highest accuracy, specificity, sensitivity, and Matthews correlation coefficient values were found as 0.85, 0.86, 0.85, and 0.71, respectively with decision tree algorithm. It was found that accuracy rates of other algorithms varied between 0.74 and 0.83. CONCLUSIONS: As a result of our study, it was found that sex estimation was made with high accuracy rate by using machine learning algorithms on X-ray images of the first and fifth metatarsal and foot phalanx. We think that in cases when pelvis, cranium and long bones are harmed and examination is difficult, bones of the first and fifth metatarsal and foot phalanx can be used for sex estimation.

Morphology of the antegonial notch and its utility in the determination of sex on skeletal materials.

A number of criteria for the assessment of biological sex, which are applied to living or recently deceased individuals, have been developed, such as genetic, chromosomal, gonadal, hormonal, and phenotypic criteria. Features of a metric and descriptive nature are used to assess the sex of skeletal materials. The diagnostic features of the skull are concentrated in the craniofacial region and around the eye sockets. The mandible is a diagnostically important part of the skull, on which a complex of features is visible. These features develop up to the third decade of life. The goal of the research was to assess the suitability of the parameters of the preangular notch, in other words, the length, height, and surface area for sex prediction applied to skeletal materials. The study included computed tomography images of the masticatory system of 194 patients, consisting of 83 females and 111 males, aged from 16 to 93 years. The three straight lines which correspond to the sides of the triangle representing the notch, in addition to its height, were determined and measured digitally. The receiver operating characteristic method was used to assess the usefulness of the studied features for the purposes of sex prediction. The sensitivity of the test ranged from 51.4% to 67% for the parameters of the preangular notch on the right-hand side, and from 44.4% to 80.2% on the left-hand side. The most reliable predictive models were obtained for two features (shown in the graphical abstract). However, when taking into account, the specificity and sensitivity of the tests presented here, only the length of side AB (a basal length of the notch) can be regarded as a feature that supports the assessment of sex on the basis of other diagnostic features of the mandible. The size and shape of the preangular notch should be treated as auxiliary features in the assessment of sex. For this reason, it is recommended that they should be applied simultaneously to the set of features described in the section on anthropological standards.

Deep learning prediction of sex on chest radiographs: a potential contributor to biased algorithms.

BACKGROUND: Deep convolutional neural networks (DCNNs) for diagnosis of disease on chest radiographs (CXR) have been shown to be biased against males or females if the datasets used to train them have unbalanced sex representation. Prior work has suggested that DCNNs can predict sex on CXR, which could aid forensic evaluations, but also be a source of bias. OBJECTIVE: To (1) evaluate the performance of DCNNs for predicting sex across different datasets and architectures and (2) evaluate visual biomarkers used by DCNNs to predict sex on CXRs. MATERIALS AND METHODS: Chest radiographs were obtained from the Stanford CheXPert and NIH Chest XRay14 datasets which comprised of 224,316 and 112,120 CXRs, respectively. To control for dataset size and class imbalance, random undersampling was used to reduce each dataset to 97,560 images that were balanced for sex. Each dataset was randomly split into training (70%), validation (10%), and test (20%) sets. Four DCNN architectures pre-trained on ImageNet were used for transfer learning. DCNNs were externally validated using a test set from the opposing dataset. Performance was evaluated using area under the receiver operating characteristic curve (AUC). Class activation mapping (CAM) was used to generate heatmaps visualizing the regions contributing to the DCNN's prediction. RESULTS: On the internal test set, DCNNs achieved AUROCs ranging from 0.98 to 0.99. On external validation, the models reached peak cross-dataset performance of 0.94 for the VGG19-Stanford model and 0.95 for the InceptionV3-NIH model. Heatmaps highlighted similar regions of attention between model architectures and datasets, localizing to the mediastinal and upper rib regions, as well as to the lower chest/diaphragmatic regions. CONCLUSION: DCNNs trained on two large CXR datasets accurately predicted sex on internal and external test data with similar heatmap localizations across DCNN architectures and datasets. These findings support the notion that DCNNs can leverage imaging biomarkers to predict sex and potentially confound the accurate prediction of disease on CXRs and contribute to biased models. On the other hand, these DCNNs can be beneficial to emergency radiologists for forensic evaluations and identifying patient sex for patients whose identities are unknown, such as in acute trauma.

Using Deep Neural Networks for Predicting Age and Sex in Healthy Adult Chest Radiographs.

BACKGROUND: The performance of chest radiography-based age and sex prediction has not been well validated. We used a deep learning model to predict the age and sex of healthy adults based on chest radiographs (CXRs). METHODS: In this retrospective study, 66,643 CXRs of 47,060 healthy adults were used for model training and testing. In total, 47,060 individuals (mean age ± standard deviation, 38.7 ± 11.9 years; 22,144 males) were included. By using chronological ages as references, mean absolute error (MAE), root mean square error (RMSE), and Pearson's correlation coefficient were used to assess the model performance. Summarized class activation maps were used to highlight the activated anatomical regions. The area under the curve (AUC) was used to examine the validity for sex prediction. RESULTS: When model predictions were compared with the chronological ages, the MAE was 2.1 years, RMSE was 2.8 years, and Pearson's correlation coefficient was 0.97 (p < 0.001). Cervical, thoracic spines, first ribs, aortic arch, heart, rib cage, and soft tissue of thorax and flank seemed to be the most crucial activated regions in the age prediction model. The sex prediction model demonstrated an AUC of >0.99. CONCLUSION: Deep learning can accurately estimate age and sex based on CXRs.

DNA methylation-based sex classifier to predict sex and identify sex chromosome aneuploidy.

BACKGROUND: Sex is an important covariate of epigenome-wide association studies due to its strong influence on DNA methylation patterns across numerous genomic positions. Nevertheless, many samples on the Gene Expression Omnibus (GEO) frequently lack a sex annotation or are incorrectly labelled. Considering the influence that sex imposes on DNA methylation patterns, it is necessary to ensure that methods for filtering poor samples and checking of sex assignment are accurate and widely applicable. RESULTS: Here we presented a novel method to predict sex using only DNA methylation beta values, which can be readily applied to almost all DNA methylation datasets of different formats (raw IDATs or text files with only signal intensities) uploaded to GEO. We identified 4345 significantly (p<0.01) sex-associated CpG sites present on both 450K and EPIC arrays, and constructed a sex classifier based on the two first principal components of the DNA methylation data of sex-associated probes mapped on sex chromosomes. The proposed method is constructed using whole blood samples and exhibits good performance across a wide range of tissues. We further demonstrated that our method can be used to identify samples with sex chromosome aneuploidy, this function is validated by five Turner syndrome cases and one Klinefelter syndrome case. CONCLUSIONS: This proposed sex classifier not only can be used for sex predictions but also applied to identify samples with sex chromosome aneuploidy, and it is freely and easily accessible by calling the 'estimateSex' function from the newest wateRmelon Bioconductor package ( https://github.com/schalkwyk/wateRmelon ).

Sex prediction by metric and non-metric analysis of the hard palate and the pyriform aperture.

BACKGROUND: Analysis of the bones and bone fragments of the cranium may be a useful tool for sex diagnosis in the identification of human remains which have been exposed to adverse conditions. The object of the present study was to evaluate sex prediction through metric and non-metric analysis of the hard palate (HP) and the pyriform aperture (PA), using macerated skulls of adult individuals. MATERIALS AND METHODS: We analysed 312 dry skulls of adult individuals of both sexes, studying the metric and non-metric characteristics of the HP and PA. The accuracy, sensitivity, specificity and positive and negative predictive values were evaluated. A binary logistic regression and a linear regression were performed. The receiver operating characteristic curve was constructed to analyse the perfor- mance of sex diagnosis. Measurements of the HP and the PA were analysed by ANOVA and Tukey's test. The SPSS v. 20.0 software was used, with a significance threshold of 5%. RESULTS: The shape of the PA presented 61.9% accuracy, 54.4% sensitivity and 65.7% specificity. The shape of the HP presented 51.5% accuracy, 65.6% sen- sitivity and 44.7% specificity. Only the height of the PA functioned as a good predictor of sex. CONCLUSIONS: The height of the PA produced good diagnostic performance (area under curve = 0.764). The height of the PA was the most reliable indicator for sex prediction, and could be used by forensic scientists to identify sex.

A mini-multiplex SNaPshot assay for the triage of degraded human DNA.

Short Tandem Repeat (STR) genotyping is currently the primary DNA-based method for human identification, however it can have limited success when applied to degraded human remains. Massively parallel sequencing (MPS) provides new opportunities to obtain genetic data for hundreds of loci in a single assay with higher success from degraded samples. However, due to the extra requirement for specialised equipment, expertise and resources, routine use of MPS may not be feasible or necessary for many forensic cases. Here we describe the development of a mini-multiplex SNaPshot screening tool (Miniplex) for human samples which allows the qualitative comparison of short mitochondrial and nuclear DNA targets, as well as the interrogation of biogeographic ancestry, lineage, and phenotype single nucleotide polymorphisms (SNPs). This tool is useful to triage samples based on sample quality prior to downstream identification workflows and provides broad biological profile data for intelligence purposes.

Crime scene reconstruction-Sex prediction from blood stained foot sole impressions.

It is often difficult to predict the sex of an individual based on bloody incomplete footprints. However, such prints/impressions are particularly common in a crime scene. Again variability in the texture, color of the target surface has an impact on the bloodstained impression formed. The study of bare foot, footprint, footwear (i.e. shoe, canvas etc.) within the legal context is referred to as forensic podiatry. Based on the fact that it is possible to predict the sex of an individual from footprint impressions, an automated model has been proposed in this paper for analyzing the sex of an individual from his/her broken/incomplete footprint impressions based on morphological features alone. Five male and female volunteers aged between 20 to 65 years participated in dataset development. Keeping the blood volume constant and having stepped on differently shaped porcine blood pools, the individuals were asked to walk on herbarium sheets. The footprints were recorded and documented in accordance with the guidelines in place for physical evidence documentation within the forensic domain. The morphological features that were extracted from each of the footprint impressions are footprint length, footprint breadth, angle of walking, approximated heel radius etc. Using exhaustive cross validation technique, the dataset was divided into training and test set. Non-redundant, relevant features that are particularly effective at sex prediction were marked out using the relief algorithm in coherence with the correlation metric. Supervised learning techniques were used on the dataset to predict the sex of the owner of an unknown footprint. The study concentrates on morphological features in order to deal with bloodstain footprint transfer stains formed on any non-porous/non-absorbent surfaces such as cemented floor, glass, mosaic floor space, colored and designed tiled floor spaces. Features such as the angle of walking and foot breadth were found to be particularly influential in sex prediction from incomplete bloodstained foot sole impressions. In comparison to a system for sex prediction from complete footprint impressions (82.2%), the automated system developed on incomplete foot impressions recorded an accuracy level of 83.47%.

A radio-odontometric analysis of sexual dimorphism in deciduous dentition.

OBJECTIVE: It has been suggested that the level of sexual dimorphism in deciduous dentition is relatively lower than that in permanent dentition. However, in sub adult skeletal remains whose osseous morphological traits of the sex have not defined yet, predicting sex on the basis of odontometric features may be the most precise technique. The aim of the present study was to assess the degree of sexual dimorphism in marginal enamel, dentin and pulp dimensions of second molar deciduous teeth in a pediatric population. METHOD AND MATERIALS: The present study was conducted on bitewing radiographs of 64 males and 60 females. The greatest width of enamel, dentin and pulp on mandibular and maxillary second molar deciduous teeth were measured. Student's t-test and discriminant analysis were used to compare the differences in the odontometric parameters between females and males. RESULTS: Among the second molar measurements, only the maxillary pulp width did significantly discriminate the sex groups. The accuracy of sex identification of a case based on deciduous second molar tooth was approximately 68%. CONCLUSIONS: The application of second molar deciduous teeth in sex prediction showed moderate level of sexual dimorphism. In this respect, the maxillary pulp width had the greatest amount of contribution in sex discrimination. Therefore, these odontometric traits, in conjunction with other skeletal features, can be used as a supplementary sexing tool for gender prediction in forensic anthropology.

Odontometric sex variation in Malaysians with application to sex prediction.

Information about the sex of individuals is important for human identification. This study was conducted to quantify classification rates of sex prediction models for Malaysians using odontometric profiles. Mesiodistal (MD) and buccolingual (BL) crown dimensions of the permanent dentition were studied in 400 young adult Malaysians, giving a total of 28 tooth size variables. The sample consisted of three major ethnic groups, the Malays, Chinese and Tamils, since the aim was to assess sex dimorphism in Malaysians as a whole. Results showed that the mesiodistal diameter of the lower canine was the most sexually dimorphic dimension in Malaysian Malays and Tamils. Univariate analyses showed that the magnitude and pattern of sex dimorphism varies between these three ethnic groups, with Malaysian Chinese and Tamils being more dimorphic than the Malaysian Malays. Stepwise discriminant functions were generated bearing in mind their application in practical forensic situations. The range of classification rates was from 70.2% to 78.5% for the composite Malaysian group, and 83.8%, 77.9%, 72.4% for Malaysian Chinese, Malays and Tamils, respectively. The 'Area Under the Receiver Operating Characteristic Curve statistics' indicated good classification rates for three prediction models obtained using a combination of all tooth size variables, mandibular teeth, and mesiodistal dimensions in the composite Malaysian group, and for all tooth size variables in each ethnic group. The present study provides strong support for the value of odontometry as an adjunct scientific method for sex prediction in human identification.