Artificial intelligence and deep learning to map immune cell types in inflamed human tissue

Area of Science:

• Immunology
• Computational Pathology
• Dermatology

Background:

• Inflammatory tissue biopsies contain complex immune cell networks crucial for immune and autoimmune responses.
• Standard histological examination has limitations in quantitating and categorizing the vast data within biopsy slides.
• Unbiased and comprehensive analysis of cellular interactions in inflammatory lesions is needed.

Purpose of the Study:

• To develop a deep learning algorithm for unbiased identification and classification of immune cells in inflammatory tissue biopsies.
• To specifically detect and classify T follicular helper (Tfh) cell subsets and B cells in dermatomyositis biopsy images.
• To enable quantitative analysis of cellular composition and spatial relationships in diseased tissues.

Main Methods:

• Development of a deep learning algorithm for image analysis of biopsy tissues.
• Application of the algorithm to identify and classify immune cell populations, including Tfh and B cells.
• Focus on images from dermatomyositis patient biopsies.

Main Results:

• The deep learning algorithm demonstrated strong performance in detecting and classifying immune cells.
• The algorithm successfully identified rare Tfh cell subsets within the tissue microenvironment.
• High accuracy was achieved in distinguishing between different immune cell types.

Conclusions:

• Deep learning offers a powerful tool for comprehensive and unbiased analysis of immune cells in inflammatory biopsies.
• The developed algorithm can quantify cellular data, providing insights beyond standard histological examination.
• This approach has potential applications in various disease states, including autoimmune and inflammatory conditions, and facilitates spatial mapping of cell types.