Prediction of immunochemotherapy response for diffuse large B-cell lymphoma using artificial intelligence digital...

Area of Science:

• Oncology
• Digital Pathology
• Artificial Intelligence

Background:

• Diffuse Large B-cell Lymphoma (DLBCL) is an aggressive non-Hodgkin lymphoma with variable drug response.
• Predicting treatment efficacy in DLBCL remains a significant clinical challenge.

Purpose of the Study:

• To develop a predictive model for immunochemotherapy response in DLBCL using digital pathology and deep learning.
• To identify key histological features associated with treatment responsiveness.

Main Methods:

• Retrospective analysis of 251 DLBCL patient slide images treated with R-CHOP.
• Feature extraction using contrastive learning and development of a multi-modal prediction model integrating clinical and image data.
• Application of knowledge distillation to mitigate overfitting and enable pathology-only predictions.

Main Results:

• The multi-modal model achieved an AUC of 0.856, correlating with clinical prognostic factors.
• Survival analyses demonstrated the model's effectiveness in predicting relapse-free survival.
• External validation confirmed the model's prognostic capabilities, with pathology-based predictions showing independent prognostic value.

Conclusions:

• Digital pathology combined with deep learning offers a novel, objective, and reproducible approach for predicting DLBCL treatment response.
• The developed model shows potential as a diagnostic and prognostic tool to enhance DLBCL clinical management.
• Further integration with genomic data could refine predictive accuracy and improve patient outcomes.