Lymphocyte-Infiltrated Periportal Region Detection With Structurally-Refined Deep Portal Segmentation and...

Area of Science:

• Medical Imaging
• Computational Pathology
• Hepatology

Background:

• Early hepatitis diagnosis and treatment are crucial for preventing liver damage and mortality.
• Accurate assessment of lymphocyte infiltration in periportal regions is key for hepatitis grading using systems like Ishak.
• Automated detection of infiltrated periportal regions is challenging due to irregular boundaries.

Purpose of the Study:

• To develop a deep-learning-based automatic detection framework for identifying lymphocyte-infiltrated periportal regions in liver Whole Slide Images.
• To assist pathologists in the early diagnosis and grading of hepatitis.

Main Methods:

• A framework combining a Structurally-REfined Deep Portal Segmentation module and an Infiltrated Periportal Region Detection module was developed.
• Heterogeneous infiltration features were utilized for accurate detection of infiltrated periportal regions.
• The method was applied to liver Whole Slide Images.

Main Results:

• The proposed method achieved an F1-score of 0.725 for detecting lymphocyte-infiltrated periportal regions.
• Detected infiltrated portal boundary ratios showed high correlation with the Ishak grade (Spearman's correlation > 0.87, p < 0.001).
• Medium correlations were observed with liver function indices aspartate aminotransferase (correlation > 0.63, p < 0.001) and alanine aminotransferase (correlation > 0.57, p < 0.001).

Conclusions:

• The ratio of infiltrated portal boundary, as determined by the framework, correlates with hepatitis severity (Ishak grade) and liver function.
• The developed deep learning framework offers a reliable tool for pathologists in hepatitis diagnosis.
• Automated detection of infiltrated regions can improve the accuracy and efficiency of hepatitis assessment.