Deep Learning Algorithm‑Based MRI Radiomics and Pathomics for Predicting Microsatellite Instability Status in Rectal...

Area of Science:

• Oncology
• Radiology
• Pathology

Background:

• Microsatellite instability (MSI) is a crucial biomarker in rectal cancer.
• Accurate prediction of MSI status is vital for treatment selection.
• Current prediction methods have limitations.

Purpose of the Study:

• To develop and validate multimodal deep-learning models for predicting MSI status in rectal cancer.
• To integrate clinical variables, multiparametric MRI (mp-MRI), and hematoxylin and eosin (HE) stained pathology slides.
• To assess the predictive performance of a combined nomogram.

Main Methods:

• Multicenter study with 467 rectal cancer patients.
• Development of clinical, deep learning radiomics score (DLRS), and deep learning pathomics score (DLPS) models.
• Construction of a nomogram integrating clinical data, mp-MRI, and HE images.

Main Results:

• The nomogram achieved high AUC values (0.974-0.987) across training and validation sets.
• The nomogram demonstrated superior predictive performance compared to individual models.
• High sensitivity and specificity were observed in predicting MSI status.

Conclusions:

• The multimodal nomogram effectively predicts MSI status in rectal cancer.
• The model integrates diverse data sources to capture tumor heterogeneity.
• This approach offers high accuracy and generalizability for MSI prediction.