AI Enables Accurate Sex Determination of Sheep

A recent study has paved the way for more accurate and efficient sex determination of archaeological animal finds, particularly sheep. Utilizing AI-based algorithms, researchers demonstrated that the sex of sheep can be identified swiftly and easily by analyzing just four distinct measurements of their ankle bone, known scientifically as the talus (plural: tali). This approach provides a significant advantage for archaeology, where the ankle bones of sheep or cattle are often well-preserved due to their small and compact nature.

An interdisciplinary team from the Bavarian State Collections of Natural History (SNSB) and LMU München conducted the study, testing various machine-learning algorithms to achieve high accuracy rates. Nadine Schüler, first author of the study and a scientist at the State Collection for Palaeonanatomy Munich (SNSB-SPM) and LMU München, noted, “The accuracy rate of the AI algorithms is high. Most algorithms correctly determine the sex—up to 70 percent; some variants even manage up to 90 percent.”

Traditionally, zoologists relied on morphological characteristics to ascertain animal sex, assessing skeletal elements specific to each species, such as the pelvic bones in sheep and cattle. However, this method presents challenges in archaeological contexts where only bone fragments are available, making it difficult even for experts to determine the sex of animals with certainty. While DNA analysis could provide a solution, it is often cost-prohibitive and risks damaging the archaeological remains.

“Machine learning could be the solution, but it is rarely applied in archaeozoology,” Schüler remarked. “In medicine, AI is already used to classify human bones. Our study represents a first step toward applying this methodology to archaeozoological data, providing a quick initial assessment of finds for archaeozoologists.”

For their study, researchers trained various AI algorithms using over 240 known sheep ankle bones. The AI program learned to differentiate between male and female tali. This learned method was then successfully applied to 170 unidentified sheep ankle bones from an archaeological excavation site in Mongolia.

This advancement highlights AI’s potential to facilitate new discoveries and streamline processes in the field of archaeology, augmenting traditional methods and addressing specific challenges in sex determination of animal remains.