Prediction of Microsatellite Instability in Colorectal Cancer Using Two Internally Validated Radiomic Models.

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Microsatellite instability (MSI) is a key biomarker in colorectal cancer (CRC).
• Accurate preoperative prediction of MSI status is crucial for treatment selection.
• Current methods often rely on invasive testing.

Purpose of the Study:

• To develop and validate radiomic models using preoperative contrast-enhanced computed tomography (PP CT) to predict MSI in CRC patients.
• To compare the performance of models developed using multi-scanner versus single-scanner data.

Main Methods:

• 115 CRC patients' PP CT scans were analyzed.
• Two radiomic models were developed: Model I (multi-scanner) and Model II (single-scanner).
• Radiomic features (RFs) and clinical data were extracted and analyzed using LASSO regression for feature selection.

Main Results:

• Model I (2 RFs + 1 clinical feature) achieved AUCs of 0.76 (training) and 0.74 (validation).
• Model II (3 RFs) achieved AUCs of 0.85 (training) and 0.72 (validation).
• Both models demonstrated good performance in distinguishing MSI from non-MSI tumors.

Conclusions:

• Radiomic models show potential as non-invasive preoperative tools for predicting MSI status in CRC.
• Model I, incorporating clinical features, showed better generalizability and less overfitting compared to Model II.
• Further validation on larger, diverse datasets is recommended to enhance model generalizability.