Findings from machine learning in clinical medical imaging applications - Lessons for translation to the forensic setting

A novel approach for exploring the regional features of vaginal fluids based on microbial relative abundance and alpha diversity

Vaginal fluids are one of the most common biological samples in forensic sexual assault cases, and their characterization is vital to narrow the scope of investigation. Presently, approaches for identifying vaginal fluids in different regions are not only rare but also have certain limitations. However, the microbiome has shown the potential to identify the source of body fluids and reveal the characteristics of individuals. In this study, 16S rRNA gene high-throughput sequencing was used to characterize the vaginal microbial community from three regions, Sichuan, Hainan and Hunan. In addition, data on relative abundance and alpha diversity were used to construct a random forest model. The results revealed that the dominant genera in the three regions were Lactobacillus, followed by Gardnerella. In addition, Ureaplasma, Nitrospira, Nocardiodes, Veillonella and g-norank-f-Vicinamibacteraceae were significantly enriched genera in Sichuan, llumatobacter was enriched in Hainan, and Pseudomonas was enriched in Hunan. The random forest classifier based on combined data on relative abundance and alpha diversity had a good ability to distinguish vaginal fluids with similar dominant microbial compositions in the three regions. The study suggests that combining high-throughput sequencing data with machine learning models has good potential for application in the biogeographic inference of vaginal fluids.

Applications of artificial intelligence in forensic sciences: Current potential benefits, limitations and perspectives

In recent years, new studies based on artificial intelligence (AI) have been conducted in the forensic field, posing new challenges and demonstrating the advantages and disadvantages of using AI methodologies to solve forensic well-known problems. Specifically, AI technology has tried to overcome the human subjective bias limitations of the traditional approach of the forensic sciences, which include sex prediction and age estimation from morphometric measurements in forensic anthropology or evaluating the third molar stage of development in forensic odontology. Likewise, AI has been studied as an assisting tool in forensic pathology for a quick and easy identification of the taxonomy of diatoms. The present systematic review follows the PRISMA 2020 statements and aims to explore an emerging topic that has been poorly analyzed in the forensic literature. Benefits, limitations, and forensic implications concerning AI are therefore highlighted, by providing an extensive critical review of its current applications on forensic sciences as well as its future directions. Results are divided into 5 subsections which included forensic anthropology, forensic odontology, forensic pathology, forensic genetics, and other forensic branches. The discussion offers a useful instrument to investigate the potential benefits of AI in the forensic fields as well as to point out the existing open questions and issues concerning its application on real-life scenarios. Procedural notes and technical aspects are also provided to the readers.

Systematic Selection of Age-Associated mRNA Markers and the Development of Predicted Models for Forensic Age Inference by Three Machine Learning Methods

Aging is usually accompanied by the decline of physiological function and dysfunction of cellular processes. Genetic markers related to aging not only reveal the biological mechanism of aging but also provide age information in forensic research. In this study, we aimed to screen age-associated mRNAs based on the previously reported genome-wide expression data. In addition, predicted models for age estimations were built by three machine learning methods. We identified 283 differentially expressed mRNAs between two groups with different age ranges. Nine mRNAs out of 283 mRNAs showed different expression patterns between smokers and non-smokers and were eliminated from the following analysis. Age-associated mRNAs were further screened from the remaining mRNAs by the cross-validation error analysis of random forest. Finally, 14 mRNAs were chosen to build the model for age predictions. These 14 mRNAs showed relatively high correlations with age. Furthermore, we found that random forest showed the optimal performance for age prediction in comparison to the generalized linear model and support vector machine. To sum up, the 14 age-associated mRNAs identified in this study could be viewed as valuable markers for age estimations and studying the aging process.

RiFNet: Automated rib fracture detection in postmortem computed tomography

Imaging techniques are widely used for medical diagnostics. In some cases, a lack of medical practitioners who can manually analyze the images can lead to a bottleneck. Consequently, we developed a custom-made convolutional neural network (RiFNet = Rib Fracture Network) that can detect rib fractures in postmortem computed tomography. In a retrospective cohort study, we retrieved PMCT data from 195 postmortem cases with rib fractures from July 2017 to April 2018 from our database. The computed tomography data were prepared using a plugin in the commercial imaging software Syngo.via whereby the rib cage was unfolded on a single-in-plane image reformation. Out of the 195 cases, a total of 585 images were extracted and divided into two groups labeled "with" and "without" fractures. These two groups were subsequently divided into training, validation, and test datasets to assess the performance of RiFNet. In addition, we explored the possibility of applying transfer learning techniques on our dataset by choosing two independent noncommercial off-the-shelf convolutional neural network architectures (ResNet50 V2 and Inception V3) and compared the performances of those two with RiFNet. When using pre-trained convolutional neural networks, we achieved an F₁ score of 0.64 with Inception V3 and an F₁ score of 0.61 with ResNet50 V2. We obtained an average F₁ score of 0.91 ± 0.04 with RiFNet. RiFNet is efficient in detecting rib fractures on postmortem computed tomography. Transfer learning techniques are not necessarily well adapted to make classifications in postmortem computed tomography.

Classifying Microscopic Acute and Old Myocardial Infarction Using Convolutional Neural Networks

ABSTRACT

Convolutional neural network (CNN) has advanced in recent years and translated from research into medical practice, most notably in clinical radiology and histopathology. Research on CNNs in forensic/postmortem pathology is almost exclusive to postmortem computed tomography despite the wealth of research into CNNs in surgical/anatomical histopathology. This study was carried out to investigate whether CNNs are able to identify and age myocardial infarction (a common example of forensic/postmortem histopathology) from histology slides. As a proof of concept, this study compared 4 CNNs commonly used in surgical/anatomical histopathology to identify normal myocardium from myocardial infarction. A total of 150 images of the myocardium (50 images each for normal myocardium, acute myocardial infarction, and old myocardial infarction) were used to train and test each CNN. One of the CNNs used (InceptionResNet v2) was able to show a greater than 95% accuracy in classifying normal myocardium from acute and old myocardial infarction. The result of this study is promising and demonstrates that CNN technology has potential applications as a screening and computer-assisted diagnostics tool in forensic/postmortem histopathology.

Deep learning in forensic gunshot wound interpretation-a proof-of-concept study

While the applications of deep learning are considered revolutionary within several medical specialties, forensic applications have been scarce despite the visual nature of the field. For example, a forensic pathologist may benefit from deep learning-based tools in gunshot wound interpretation. This proof-of-concept study aimed to test the hypothesis that trained neural network architectures have potential to predict shooting distance class on the basis of a simple photograph of the gunshot wound. A dataset of 204 gunshot wound images (60 negative controls, 50 contact shots, 49 close-range shots, and 45 distant shots) was constructed on the basis of nineteen piglet carcasses fired with a .22 Long Rifle pistol. The dataset was used to train, validate, and test the ability of neural net architectures to correctly classify images on the basis of shooting distance. Deep learning was performed using the AIDeveloper open-source software. Of the explored neural network architectures, a trained multilayer perceptron based model (MLP_24_16_24) reached the highest testing accuracy of 98%. Of the testing set, the trained model was able to correctly classify all negative controls, contact shots, and close-range shots, whereas one distant shot was misclassified. Our study clearly demonstrated that in the future, forensic pathologists may benefit from deep learning-based tools in gunshot wound interpretation. With these data, we seek to provide an initial impetus for larger-scale research on deep learning approaches in forensic wound interpretation.