JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Synergistic H&E and IHC image analysis by AI predicts cancer biomarkers and survival outcomes in colorectal and breast cancer

  • 0Caris Life Sciences, Phoenix, AZ, USA.
Communications Medicine +

|

August 1, 2025

|

August 1, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

An AI model accurately predicts cancer biomarkers (microsatellite instability/mismatch repair deficiency and PD-L1) from pathology slides, improving treatment selection for immunotherapy like pembrolizumab.

Area Of Science

  • Computational pathology
  • Artificial intelligence in oncology
  • Biomarker discovery

Background

  • Immunotherapy, including pembrolizumab, shows promise for metastatic colorectal cancer (CRC) and triple-negative breast cancer (TNBC).
  • Accurate detection of biomarkers like microsatellite instability (MSI)/mismatch repair deficiency (MMRd) and programmed death-ligand 1 (PD-L1) is crucial for treatment efficacy.
  • Current biomarker detection methods (IHC, NGS) are labor-intensive and require subjective interpretation.

Purpose Of The Study

  • To develop and evaluate a dual-modality transformer-based AI model for predicting MSI/MMRd and PD-L1 status.
  • To assess the model's performance in stratifying patients for immunotherapy response.
  • To compare the AI model's predictive accuracy against traditional methods.

Main Methods

  • A transformer-based AI model was developed using hematoxylin & eosin and IHC stained whole slide images.
  • Model performance was evaluated using the area under the receiver operating curve (AUROC).
  • Time-on-treatment (TOT) and overall survival (OS) were analyzed using Kaplan-Meier and Cox proportional hazard models.

Main Results

  • The AI framework achieved AUROC > 0.97 for MSI/MMRd in CRC and > 0.96 for PD-L1 in breast cancer.
  • Biomarker-positive predictions correlated with prolonged TOT and OS in patients treated with pembrolizumab.
  • The AI model demonstrated superior patient stratification for pembrolizumab outcomes compared to PD-L1 IHC in breast cancer.

Conclusions

  • The study supports integrating AI tools into clinical pathology for enhanced precision and efficiency in cancer biomarker evaluation.
  • The developed AI model offers a customizable framework for diverse clinical scenarios.
  • The AI model provides superior prognostic precision over current biomarker assessments by integrating multi-modal staining features.