Deep Learning-Driven Analysis and Quantification of Histopathologic Features in a Dextran Sulfate Sodium-Induced...

Area of Science:

• Veterinary Pathology
• Digital Pathology
• Artificial Intelligence in Medicine

Background:

• Dextran sodium sulfate (DSS)-induced colitis models are crucial for studying inflammatory bowel disease (IBD) pathophysiology.
• Histological scoring of colon sections in these models is vital but often subjective and time-consuming.
• Digital pathology offers potential solutions to enhance objectivity and efficiency in histological assessments.

Purpose of the Study:

• To develop and validate an AI-based deep learning classifier for quantifying histological features in a DSS-induced colitis mouse model.
• To assess the utility of AI in identifying and quantifying tissue damage, cell infiltration, and goblet cell numbers.
• To compare the sensitivity and accuracy of AI-driven assessments against traditional pathologist evaluations.

Main Methods:

• A deep learning classifier was trained using the HALO image analysis platform on colon sections from a DSS-induced colitis mouse model.
• The AI sequentially identified tissue layers (tissues, mucosa, submucosa, lymphoid tissues).
• AI quantified tissue damage, cell infiltration, goblet cell counts, and nuclear phenotypes.

Main Results:

• AI-based classifiers accurately identified key pathological features and disease progression in the colitis model.
• AI assessments demonstrated high correlation and superior sensitivity in detecting histopathological changes compared to pathologist assessments.
• The AI approach successfully detected the alleviation of tissue damage.

Conclusions:

• Machine learning-based classifiers are efficient for characterizing disease pathology in preclinical models.
• AI-driven digital pathology provides a quantitative, sensitive, and unbiased tool for histological assessment.
• This technology shows significant utility in the early stages of drug discovery for inflammatory bowel disease.