Predictive model based on mesorectal fat radiomics for pathological complete response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer

Abstract

Purpose

To explore the predictive value of MRI radiomics based on mesorectal fat for pathological complete response (pCR) to neoadjuvant chemoradiotherapy in locally advanced rectal cancer, and to develop a combined predictive model incorporating MRI radiomics, quantitative fat parameters and clinical features.

Materials and Methods

In this retrospective study, 235 rectal cancer patients who received neoadjuvant chemoradiotherapy followed by resection were enrolled, with their pretreatment MRI. Patients were randomly allocated into training (n = 164) and test (n = 71) cohorts. Mesorectal fat was manually segmented on T2-weighted imaging. Radiomics model to predict pCR were built through maximum Relevance Minimum Redundancy algorithm and Least Absolute Shrinkage and Selection Operator regression. Univariate and multivariate logistic regression analyses were performed to select independent predictive factors from imaging and clinical features. Then a combined radiomic-clinical predictive model and a nomogram were constructed. Model performances were evaluated using the area under the curve (AUC) and compared using the DeLong test.

Results

The radiomics model demonstrated AUCs of 0.78 in the test set. A radiomics-clinical model integrating Radscore, N stage, posterior mesorectal thickness, and mesorectal fat area, reached an AUC of 0.92 (95% CI: 0.89–0.95) in the test cohort.

Conclusion

Radiomics-clinical model based on mesorectal fat could be a useful approach for pretreatment pCR prediction in locally advanced rectal cancer.