Predicting recurrence in osteosarcoma via a quantitative histological image classifier derived from tumour nuclear morphological features

Abstract Recurrence is the key factor affecting prognosis of osteosarcoma. Currently, there a lack clinically useful tools to predict osteosarcoma recurrence. The application pathological images for artificial intelligence‐assisted accurate prediction tumour outcomes increasing. Thus, present study constructed quantitative histological image classifier with nuclear features using haematoxylin and eosin (H&E)‐stained whole‐slide (WSIs) from 150 patients. We first segmented eight distinct tissues in H&E‐stained WSIs, an average accuracy 90.63% on testing set. areas were automatically accurately acquired, facilitating cell feature extraction process. Based six selected features, we developed (OSHIC) recurrence survival following standard treatment. OSHIC derived independently predicted patients, thereby contributing precision oncology. Moreover, fully automated workflow extract evaluate diagnostic values sets build classifiers outcomes. provides novel tool predicting outcomes, which has broad prospect clinical practice.