Bridging clinic and wildlife care with AI-powered pan-species computational pathology

Khalib AbdulJabbar, Simon P Castillo, Katherine Hughes, Hannah Davidson, Amy M Boddy, Lisa M Abegglen, Lucia Minoli, Selina Iussich, Elizabeth P Murchison, Trevor A Graham, Simon Spiro, Carlo C Maley, Luca Aresu, Chiara Palmieri, Yinyin Yuan

Introduction to the Pan-Species Cancer Digital Pathology Atlas 

A team of researchers has built a pan-species cancer digital pathology atlas, panspecies.ai, to gain a better understanding of cancer initiation and evolution across different species. They trained a supervised convolutional neural network algorithm on human samples for a computational comparative pathology study. The algorithm achieved high accuracy in measuring immune response through single-cell classification for two transmissible cancers, canine transmissible venereal tumor, and Tasmanian devil facial tumor disease. It was also evaluated across 18 other vertebrate species, including mammals, reptiles, birds, and amphibians.

Artificial Intelligence and Spatial Statistics in Veterinary Pathology 

A spatial immune score based on artificial intelligence and spatial statistics was associated with prognosis in canine melanoma and prostate tumors. The team developed a metric named morphospace overlap, which could guide veterinary pathologists towards deploying technology on new samples rationally, and approximates the overlap of other species pathology with the human phenotype. The study provides the foundation for transferring artificial intelligence technologies to veterinary pathology based on an understanding of morphological conservation. The research has significant implications for animal welfare and wildlife conservation by providing new tools and insights into tumorigenesis and cancer resistance. It also promises to pave the way for pan-species comparative pathology and could vastly accelerate developments in veterinary medicine and comparative oncology.

Link to Article: Bridging clinic and wildlife care with AI-powered pan-species computational pathology | Nature Communications