Novel deep learning algorithm based MRI radiomics for predicting lymph node metastases in rectal cancer.

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Lymph node metastasis (LNM) is a critical prognostic factor in rectal cancer (RC).
• Accurate prediction of LNM is essential for optimal treatment planning and patient management in RC.
• Current prediction methods may not fully leverage the potential of advanced imaging techniques.

Purpose of the Study:

• To evaluate the efficacy of an MRI-based radiomic nomogram for predicting LNM in RC.
• To compare the predictive performance of the radiomic nomogram with traditional physician-based assessments and deep learning models.

Main Methods:

• Retrospective analysis of 430 rectal cancer patients from two medical centers.
• Development of a physician model based on clinical predictors.
• Extraction of deep features from multiparametric MRI (mpMRI) to create deep learning radscore (DLRS) models.
• Construction of a nomogram model combining physician and DLRS models.

Main Results:

• Out of 430 patients, 192 (44.65%) had LNM.
• The radiomic nomogram and DLRS models achieved superior predictive performance, with AUC values ranging from 0.83 to 0.99.
• The nomogram and DLRS models demonstrated significantly higher accuracy in predicting LNM status compared to the physician model (AUC 0.7-0.79).

Conclusions:

• An MRI-based radiomic nomogram, particularly when incorporating DLRS, is a highly accurate tool for predicting LNM in RC.
• This advanced imaging-based approach surpasses traditional clinical assessments in predicting LNM status.
• The findings support the integration of radiomic nomograms into clinical practice for improved RC management.