Utilization of a Deep Learning Algorithm for Microscope-Based Fatty Vacuole Quantification in a Fatty Liver Model in...

Area of Science:

• Pathology
• Artificial Intelligence
• Computational Biology

Background:

• Traditional semiquantitative pathology assessment faces challenges in accurately quantifying features like fatty vacuoles in liver tissue.
• Differentiating fatty vacuoles from liver blood vessels and bile ducts requires precise analysis.

Purpose of the Study:

• To develop and validate a deep learning artificial intelligence (AI) algorithm for precise quantification of fatty vacuoles in liver tissue.
• To compare the AI algorithm's quantitative output with traditional semiquantitative pathology assessment methods.

Main Methods:

• A deep learning AI algorithm utilizing a segmentation framework was developed using glass slides of mouse liver as a model for nonalcoholic fatty liver disease.
• The algorithm was designed for real-time analysis of histopathology fields during microscope-based assessment.
• Manual semiquantitative assessment was compared against the AI algorithm's quantitative output.

Main Results:

• The deep learning AI algorithm demonstrated high accuracy in recognizing and quantifying the percentage of fatty vacuoles.
• A strong and significant positive correlation (r = 0.87, P < .001) was observed between the AI's quantitative results and the manual semiquantitative assessment.
• The AI algorithm proved effective in differentiating fatty vacuoles from lumina of liver blood vessels and bile ducts.

Conclusions:

• Deep learning AI algorithms can significantly enhance the accuracy and efficiency of pathology assessments, particularly for complex quantifications.
• This AI approach offers a valuable tool to improve the outputs of toxicologic pathology workflows.
• The developed AI algorithm shows promise for real-time, quantitative analysis in histopathology.