Scoring of tumor-infiltrating lymphocytes: From visual estimation to machine learning

Area of Science:

• Computational pathology
• Digital pathology
• Cancer immunology

Background:

• Tumor-infiltrating lymphocytes (TILs) are key biomarkers for predicting response to immune-checkpoint therapy.
• Manual TIL quantification is time-consuming, imprecise, and lacks detailed pattern analysis.

Purpose of the Study:

• To review and discuss automated computational image analysis approaches for TIL quantification.
• To highlight the advantages of explainable machine learning in this context.

Main Methods:

• Classical image segmentation and classification based on cell properties.
• Machine learning (ML) approaches, including explainable ML, for direct cell classification.
• Generation of high-resolution heatmaps for result interpretation.

Main Results:

• Automated methods offer standardized and efficient TIL quantification compared to manual estimation.
• Explainable ML provides interpretable results, aiding in plausibility checks for diagnostics.

Conclusions:

• Computational image analysis, particularly explainable ML, is crucial for advancing TIL quantification in cancer research and diagnostics.
• These methods enhance precision, efficiency, and interpretability in predicting therapy response.