Artificial intelligence-based assessment of PD-L1 expression in diffuse large B cell lymphoma.

Area of Science:

• Hematology
• Oncology
• Medical Imaging Analysis

Background:

• Diffuse large B cell lymphoma (DLBCL) is an aggressive hematologic malignancy.
• Immunohistochemistry (IHC) is vital for DLBCL cell characterization and treatment guidance.
• PD-L1 expression is a key biomarker for predicting response to immunotherapy.

Purpose of the Study:

• To develop and validate an AI-based image analysis approach for quantifying PD-L1 expression in DLBCL.
• To assess the utility of AI in improving objectivity and interpretability of PD-L1 assessment.
• To evaluate the performance of AI compared to pathologists in analyzing IHC slides from fine needle biopsies and surgical specimens.

Main Methods:

• Development of an AI-powered image analysis tool for PD-L1 assessment in DLBCL IHC slides.
• Large-scale cell annotation of 5101 tissue regions and 146,439 cells.
• Validation of the AI approach in primary and validation cohorts, comparing results from fine needle biopsies and surgical specimens.

Main Results:

• The AI approach demonstrated effectiveness in quantitative PD-L1 assessment, overcoming challenges in cell type identification.
• Higher agreement was observed between AI and pathologists, and among pathologists, for fine needle biopsy data compared to surgical specimens.
• The AI-enabled analytics enhanced the objectivity and interpretability of PD-L1 quantification.

Conclusions:

• AI-based image analysis offers a robust method for PD-L1 quantification in DLBCL.
• This technology has the potential to improve patient selection for targeted immunotherapy.
• AI enhances diagnostic accuracy and consistency in assessing biomarkers for blood cancer treatment.