MRI-based AI framework for predicting lymph node metastasis and prognosis in rectal cancer after neoadjuvant...

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Early prediction of lymph node metastasis (LNM) is critical for rectal cancer (RC) treatment planning and prognosis.
• Neoadjuvant chemoradiotherapy (nCRT) is a standard treatment for RC, but predicting treatment response and outcomes remains challenging.

Purpose of the Study:

• To develop and validate an AI framework for predicting LNM and prognosis in RC patients after nCRT.
• To integrate clinical, radiomics, and autoencoder features for enhanced predictive accuracy.

Main Methods:

• A multi-center dataset of 577 RC patients treated with nCRT and surgery was utilized.
• A 3D U-Net model performed lesion segmentation, followed by feature extraction (clinical, radiomics, autoencoder).
• Machine learning models (XGBoost) were trained to classify LNM, and SHAP values analyzed feature importance.

Main Results:

• The AI framework achieved strong performance in LNM classification, with XGBoost showing an AUROC of 0.73 on the external test set.
• The model demonstrated superior diagnostic accuracy, specificity, and sensitivity compared to radiologists.
• Specific MRI features (T2WI_feat_16, DWI_feat_164) were identified as indicators of poor prognosis.

Conclusions:

• The developed AI framework serves as a valuable tool for predicting LNM and prognosis in RC patients post-nCRT.
• This AI-driven approach can significantly enhance clinical decision-making and prognostic assessment for rectal cancer patients.