Shape-Based Classification and Anatomical Localization of Phalanges in Hamdani Sheep and Hair Goats

Reliable identification of isolated sheep and goat phalanges is challenging because of the close skeletal similarity between these species and the repeated arrangement of homologous digital bones across the forelimbs, hindlimbs, and medial-lateral positions. This study evaluated three-dimensional phalangeal shape for discrimination between 12-month-old male Hamdani sheep and Hair goats and for localization among the sampled anatomical positions. Proximal phalanx (P1), middle phalanx (P2), and distal phalanx (P3) from 30 male Hamdani sheep (Ovis aries) and 30 male Hair goats (Capra hircus), all 12 months of age, were analysed. Bone-specific configurations of 60, 77, and 60 surface pseudolandmarks were used for P1, P2, and P3, respectively. Three-dimensional shape provided substantially greater diagnostic information than centroid size. Shape-only species classification reached 98.3% for P1, 98.3% for P2, and 100% for P3, whereas forelimb-hindlimb classification reached 100%, 98.3%, and 96.2%, respectively. P1 provided the strongest discrimination among the four sampled compound anatomical positions, achieving 91.1% internal classification accuracy. Size-only classification accuracy ranged from 48.5% to 58.6%, compared with 98.3-100.0% for shape-only models. P2 showed the largest adjusted breed-species effect in the animal-mean analysis (partial R² = 0.421), whereas P3 yielded the highest internal species-classification accuracy (100.0%) and the largest descriptive sheep-goat mean-shape distance within its own phalanx-specific shape space. Importantly, P3 showed no significant interspecific centroid-size difference despite perfect shape-based species classification, demonstrating that its diagnostic signal was primarily geometric rather than size-dependent. Overall, the results revealed a phalanx-specific diagnostic pattern within the present reference sample: P3 yielded the highest observed breed-species classification accuracy, whereas P1 provided the strongest discrimination of limb region and the four sampled compound anatomical positions. These findings support discrimination within the present modern reference sample; independent validation across other breeds, both sexes, different ages, and archaeological specimens of molecularly confirmed taxon is required before routine application.