Diagnosis Of Neoplasms In Veterinary Medicine Samples Using Artificial Intelligence

The study investigates application of machine learning (ML) to hematoxylin and eosin (H&E)-stained histopathological brightfield and acoustic microscopy images for the diagnosis of common canine and feline neoplasms, specifically soft tissue sarcomas (STS) and mast cell tumors (MCT). We attempted to distinguish between these morphologically similar tumor types by extracting twelve morphological features and utilizing five different ML classifiers: Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), Naïve Bayes (NB), Decision Tree (DT), and k-nearest neighbor (KNN). The results showed that best performing ML models were able to classify STS and MCT as tumors from optical and acoustic microscopy images with 7585% accuracy and could distinguish healthy tissue from cancerous with 79.4-87.9% accuracy, even when the tumor cell infiltration is minimal.