Image-based morphometric analysis reveals sexual dimorphism in the threatened yellow-spotted mountain newt (Neurergus derjugini): applications of deep learning and automated pattern recognition

Abstract Sexual dimorphism is a widespread evolutionary phenomenon across vertebrates, shaped by the combined effects of natural and sexual selection. In this study, we investigated sex-based morphological differences in the endangered Yellow-Spotted Mountain Newt ( Neurergus derjugini ) using a combination of traditional morphometrics and modern image-based analysis. A total of 394 adults individuals (102 females and 292 males) were photographed in the field, and dorsal images were analyzed to extract six phenotypic traits: body area, yellow spot area (%), total yellow spot area, average spot size, number of yellow spots, and spot circularity. Morphometric indices were computed using OpenCV-based segmentation and a convolutional neural network (EfficientNetB0) trained to distinguish body and spot regions. Statistical analyses revealed significant sexual dimorphism across all traits. Females had significantly larger body areas, consistent with fecundity-driven size dimorphism, while males exhibited greater yellow spot area relative to body size, more numerous and circular spots, suggesting a role for sexual selection. Principal component analysis (PCA) confirmed clear multivariate separation between sexes, emphasizing the multidimensional nature of sexual dimorphism. Our findings provide quantifiable, image-derived criteria for sex identification in N. derjugini and highlight the utility of automated phenotypic analysis in amphibian conservation. These methods offer scalable, non-invasive tools for population monitoring and contribute to a deeper understanding of morphological evolution in threatened species.