AI(H)
Description
The study introduces AI(H), a deep learning tool designed to analyze liver biopsies from autoimmune hepatitis (AIH) and provide accurate, reproducible, and interpretable results directly from pathology slides. Trained on 123 pre-treatment liver biopsies, the AI models achieved high accuracy in detecting various features such as tissue, liver microanatomy, necroinflammation, and fibrosis, with accuracies ranging from 79.2% to 99.4% on different staining methods. AI(H) also successfully identified and classified immune cells and detected bile duct injury in a significant portion of AIH cases, offering a valuable tool for improving the reproducibility and detail of AIH biopsy assessments.
Creator
Aiforia
Information
Pediatrics or Adult
Adult
Speciality
Pathology
Modality
Biopsy Slides
Training
123 pre-treatment liver biopsies from 116 adult patients with AIH
Github
Publication
FDA