Fibrosis severity scoring on Sirius red histology with multiple-instance deep learning

Area of Science:

• Digital pathology
• Artificial intelligence in medicine
• Liver disease research

Background:

• Non-alcoholic fatty liver disease (NAFLD) is the leading cause of chronic liver disease globally, impacting 30% of the population.
• Accurate histopathological assessment of fibrosis is critical for NAFLD diagnosis and patient outcome prediction.
• Current histopathology reading suffers from significant inter- and intra-rater variability.

Purpose of the Study:

• To develop and validate a deep learning solution for automated fibrosis severity scoring in NAFLD.
• To leverage digitized whole-slide images (WSIs) for objective and reproducible fibrosis staging.
• To establish new state-of-the-art performance benchmarks for AI-driven NAFLD fibrosis assessment.

Main Methods:

• A retrospective cohort of 152 Sirius-Red stained WSIs from NAFLD patients was analyzed.
• Fibrosis stage was annotated at the slide level by an expert pathologist.
• Multiple instance learning and multiple-inference techniques were employed to handle sparse pathological features.

Main Results:

• The developed deep learning model achieved high performance metrics.
• Accuracy: [insert accuracy value]
• F1 Score: [insert F1 score value]
• AUC: [insert AUC value]

Conclusions:

• The novel deep learning solution demonstrates significant potential for accurate and consistent fibrosis severity scoring in NAFLD.
• This AI-driven approach can mitigate variability in histopathological diagnoses.
• The achieved results represent a new state-of-the-art for automated fibrosis assessment in NAFLD.