Longitudinal dynamic MRI radiomic models for early prediction of prognosis in locally advanced cervical cancer treated with concurrent chemoradiotherapy.

Purpose: To investigate the early predictive value of dynamic magnetic resonance imaging (MRI)-based radiomics for progression and prognosis in locally advanced cervical cancer (LACC) patients treated with concurrent chemoradiotherapy (CCRT).

Methods and materials: A total of 111 LACC patients (training set: 88; test set: 23) were retrospectively enrolled. Dynamic MR images were acquired at baseline (MRI_pre), before brachytherapy delivery (MRI_mid) and at each follow-up visit. Clinical characteristics, 2-year progression-free survival (PFS), and 2-year overall survival (OS) were evaluated. The least absolute shrinkage and selection operator (LASSO) method was applied to extract features from MR images as well as from clinical characteristics. The support vector machine (SVM) model was trained on the training set and then evaluated on the test set.

Results: Compared with single-sequence models, multisequence models exhibited superior performance. MRI_mid-based radiomics models performed better in predicting the prognosis of LACC patients than the post-treatment did. The MRI_pre-, MRI_mid- and the ΔMRI_mid (variations in radiomics features from MRI_pre and MRI_mid) -based radiomics models achieve AUC scores of 0.723, 0.750 and 0.759 for 2-year PFS and 0.711, 0.737 and 0.789 for 2-year OS in the test set. When combined with the clinical characteristics, the ΔMRI_mid-based predictive model also performed better than the other models did, with an AUC of 0.812 for progression and 0.868 for survival.

Conclusion: We built machine learning models from dynamic features in longitudinal images and found that the ΔMRI_mid-based model can serve as a non-invasive indicator for the early prediction of prognosis in LACC patients receiving CCRT. The integrated models with clinical characteristics further enhanced the predictive performance.