Development of artificial intelligence based computational pathology to assess the histopathological toxicity of antibiotics to marine mussels

Antibiotics are emerging contaminants of concern worldwide, especially in coastal areas. However, no conclusion can confirm their ecotoxic effects based on popular employed biomarkers. Based on our previous exploration on the histopathological changes on marine model organism-marine mussels, convincing results indicated sulfamethoxazole (SMX) can induce tissue damage but the examination procedure was labor- and time-consuming. In this study, a systematic working flow of histopathological assessment was developed, including qualitative, semi-quantitative, quantitative and artificial intelligence (AI)-based quantitative methods, forming the computational pathology on functional tissues in marine mussels. The exposure of mussels to a serial concentration of SMX was conducted, gill and digestive gland of mussels were stained by H&E to perform the developed working flow. The results confirmed that SMX exposure indeed cause significant histopathological alterations in both tissues. The manual semi-quantitative, quantitative and AI-based quantitative indicators all showed a well dose-response relationship with SMX exposure. In particular, AI-based quantitative methods can identify and segment biological pathological images, and screen quantitative pathological indicators and significantly reduce the time-cost. This study confirmed the valuable application of quantitative and AI-based quantitative histopathological indicators in marine ecotoxicology, and promotes the study of computational pathology of marine organisms in emerging marine pollutants.