By Asakanya Pato
Imagine investigators find unidentified human remains, but there is no identification card, no usable fingerprints and no obvious way to determine who the person was. One important clue could be their age. Estimating age at death can help forensic scientists narrow the search and contribute to the identification process. However, estimating age is not always straightforward. Our tissues change as we grow and age, but these changes can also be influenced by factors such as health, lifestyle and other biological differences.
So, could artificial intelligence help forensic scientists make better use of these tiny biological clues?
A team of researchers set out to investigate whether machine learning could classify a person’s age group using microscopic characteristics of bone and cartilage. Their study used histomorphometric data from 294 male corpses aged between 10 and 93 years. The researchers examined 76 microscopic measurements describing changes in bone and cartilage tissue.
Turning tiny tissue clues into useful information
The researchers first needed to make sense of a large and complex dataset. They used several machine learning approaches, including random forest, CatBoost, support vector machines, k-nearest neighbours and logistic regression.
But more data does not necessarily mean better predictions. The researchers therefore selected the features that contained the most useful information and removed less informative ones. They also used techniques called t-SNE and uMAP to visualise the data. Before feature selection, the different age groups overlapped considerably. After selecting the most informative features, clearer clusters corresponding to age groups emerged.
The results were promising. Among the models tested, random forest performed best, achieving an F1-score of 0.91. Overall, the researchers reported that the machine learning approaches could establish certain age groups with approximately 90% accuracy.
Why does this matter?
The study shows that microscopic changes in bone and cartilage contain valuable information about age, information that can be difficult to interpret using simple mathematical models. By combining histomorphometry with machine learning, forensic scientists may be able to extract patterns from complex biological data more effectively.
This does not mean that artificial intelligence can simply “tell you someone’s age” Rather, it can help classify an unknown individual into an age group, providing another piece of evidence for forensic investigations.
The researchers suggest that these approaches could contribute to the development of digital databases of age-related tissue characteristics, potentially improving the consistency and efficiency of future forensic age estimation.
Reference
Zolotenkova G.V., Rogachev A.I., Pigolkin Y.I., Edelev I.S., Borshchevskaya V.N., Cameriere R. (2022). Age classification in forensic medicine using machine learning techniques. Sovremennye tehnologii v medicine, 14(1), 15. https://doi.org/10.17691/stm2022.14.1.02
Leave a Reply