AI-Assisted Risk Stratification in Stage II Colorectal Cancer: Multi-Institutional Validation of...

Area of Science:

• Computational pathology
• Artificial intelligence in oncology
• Colorectal cancer research

Background:

• Accurate risk stratification is crucial for stage II colorectal cancer (CRC) treatment decisions.
• Current guidelines recommend adjuvant chemotherapy for high-risk patients, necessitating improved prognostic tools.

Purpose of the Study:

• To develop and validate an AI-based approach for automated invasive front assessment.
• To enhance prognostic stratification in stage II colorectal cancer patients.

Main Methods:

• Developed SÉMIL (Semantically-Enhanced Multiple Instance Learning), integrating vision-language models with attention-based MIL.
• Trained and validated SÉMIL on 1,608 H&E-stained whole slide images from three independent cohorts.
• Compared SÉMIL performance against manual pathologist assessment and non-semantic MIL methods.

Main Results:

• SÉMIL demonstrated superior performance in binary classification and survival prediction across validation cohorts.
• The AI tool successfully stratified outcomes within National Comprehensive Cancer Network (NCCN) guideline-defined high-risk stage II CRC patients.
• SÉMIL retained independent prognostic significance in multivariate analysis, outperforming conventional clinicopathological features.

Conclusions:

• SÉMIL provides validated prognostic stratification for stage II colorectal cancer.
• The AI tool shows potential for refining risk assessment, particularly within challenging high-risk categories.
• Automated invasive front assessment using SÉMIL can improve clinical decision-making for CRC patients.