IVIM-DWI and MRI-based radiomics in cervical cancer: Prediction of concurrent chemoradiotherapy sensitivity in combination with clinical prognostic factors

Abstract

PurposeTo identify the feasibility and value of intravoxel incoherent motion diffusion weighted imaging (IVIM-DWI) and magnetic resonance imaging (MRI)-based radiomics combined with clinical prognostic factors (CPF) in predicting concurrent chemoradiotherapy (CCRT) sensitivity of locally advanced cervical cancer (LACC). MethodsA retrospective analysis of 163 patients (assigned to training or test groups) who underwent conventional MRI and IVIM-DWI before CCRT were divided into sensitive and resistant groups according to their efficacy at 6 months after CCRT. Per-treatment IVIM-DWI parameters (ADC, D, D⁎ and f value), 3D texture features (from axial T2WI) and CPF were measured, analyzed and screened. The prediction model and its nomogram were developed by combining screened parameters and then validated internally and externally. ResultsClinical stage, f value, D value, InverseVariance, SizeZoneNonUniformity, and Minimum were selected to construct prediction model. All parameters except D value showed independent diagnostic value in multivariate Logistic regression analysis and composed prediction model, with AUCs of 0.987 and 0.984 for training and test groups, respectively. The calibration curve (Brier score of 0.042, C-index of 0.987), decision curve and clinical impact curve further demonstrated the reliability and clinical value of prediction model. ConclusionIVIM-DWI, MRI-based radiomics and CPF showed high clinical value in predicting CCRT sensitivity for LACC with better predictive performance when combined.