A Comprehensive Prediction Model Based on MRI Radiomics and Clinical Factors to Predict Tumor Response After
Hao Jiang1, Wei Guo2, Zhuo Yu3
1Department of Radiology, the Second Affiliated Hospital of Harbin Medical University, Harbin 150086, China (H.J., X.L., H.J., Z.S., J.L., S.Z., H.H.).
Academic Radiology
|July 2, 2023
Summary
A new model using pretreatment MRI and clinical data accurately predicts neoadjuvant chemoradiotherapy (nCRT) efficacy in locally advanced rectal cancer (LARC) patients. This noninvasive tool aids in predicting treatment outcomes for LARC.
Area of Science:
- Oncology
- Radiology
- Medical Imaging
Background:
- Locally advanced rectal cancer (LARC) requires effective neoadjuvant chemoradiotherapy (nCRT).
- Predicting nCRT efficacy is crucial for tailoring treatment strategies.
- Pretreatment imaging features and clinical data hold potential for predictive modeling.
Purpose of the Study:
- To develop and validate a prediction model for nCRT efficacy in LARC patients.
- To integrate magnetic resonance imaging (MRI) multisequence features with clinical parameters.
- To assess the predictive performance of combined clinical and imaging models.
Main Methods:
- Retrospective analysis of LARC patients (n=127) with clinicopathological confirmation.
- Extraction of pretreatment MRI multisequence imaging features and clinical data.
- Construction of clinical, imaging, and comprehensive prediction models; evaluation using AUC and decision curve analysis.
Main Results:
- The comprehensive model integrating MRI features and clinical data achieved high predictive accuracy (AUC 0.99 training, 0.94 validation).
- Radiomic nomograms demonstrated good resolution and superior calibration/discrimination compared to single models.
- The model effectively predicted treatment response based on tumor regression grading (TRG).
Conclusions:
- A nomogram based on pretreatment MRI characteristics and clinical factors can noninvasively predict nCRT outcomes in LARC.
- This predictive tool offers potential for optimizing patient management and treatment planning.
- The comprehensive model significantly outperforms models based solely on clinical or imaging data.
