Delta‐radiomics‐based models for toxicity prediction in radiotherapy: A systematic review and meta‐analysis

Abstract

AbstractIntroductionDelta‐radiomics models are potentially able to improve the treatment assessment than single‐time point features. The purpose of this study is to systematically synthesize the performance of delta‐radiomics‐based models for radiotherapy (RT)‐induced toxicity.MethodsA literature search was performed following the PRISMA guidelines. Systematic searches were performed in PubMed, Scopus, Cochrane and Embase databases in October 2022. Retrospective and prospective studies on the delta‐radiomics model for RT‐induced toxicity were included based on predefined PICOS criteria. A random‐effect meta‐analysis of AUC was performed on the performance of delta‐radiomics models, and a comparison with non‐delta radiomics models was included.ResultsOf the 563 articles retrieved, 13 selected studies of RT‐treated patients on different types of cancer (HNC = 571, NPC = 186, NSCLC = 165, oesophagus = 106, prostate = 33, OPC = 21) were eligible for inclusion in the systematic review. Included studies show that morphological and dosimetric features may improve the predictive model performance for the selected toxicity. Four studies that reported both delta and non‐delta radiomics features with AUC were included in the meta‐analysis. The AUC random effects estimate for delta and non‐delta radiomics models were 0.80 and 0.78 with heterogeneity, I2 of 73% and 27% respectively.ConclusionDelta‐radiomics‐based models were found to be promising predictors of predefined end points. Future studies should consider using standardized methods and radiomics features and external validation to the reviewed delta‐radiomics model.