Diagnostic performance of contrast-enhanced CT-based radiomics models for Ki-67 status prediction in gastrointestinal stromal tumors: a systematic review and meta-analysis

Ki-67 is a key proliferation marker and prognostic indicator in gastrointestinal stromal tumors (GISTs), but its assessment requires invasive tissue sampling. Contrast-enhanced CT (CE CT)–based radiomics offers a non-invasive alternative. We systematically evaluated the diagnostic performance of CE CT–based radiomics models for predicting Ki-67 status in GISTs. Following PRISMA-DTA, we searched PubMed, Embase, Scopus, Web of Science, and CNKI to October 20, 2025. Eligible studies applied CT-based radiomics to predict Ki-67 status in GISTs and reported diagnostic accuracy metrics. Risk of bias and radiomics methodology were assessed with QUADAS-2 and METRICS. A bivariate random-effects model was used to pool sensitivity, specificity, and prespecified subgroup analyses examined key methodological moderators, including Ki-67 cut-off (5–6% vs. 8–10%). Eleven retrospective studies, enrolling 2,868 patients (11 validation cohorts), met the inclusion criteria. Four studies reported external validation. Pooled sensitivity and specificity for distinguishing high from low Ki-67 status were 0.67 and 0.82, with an AUC of 0.75. Deeks’ test did not indicate significant publication bias ( p = 0.63). Exploratory subgroup analyses suggested lower specificity among studies using PyRadiomics, whereas differences between Ki-67 cut-offs were small; however, these comparisons were based on few studies and should be interpreted cautiously. CE CT–based radiomics models demonstrated moderate diagnostic accuracy for preoperative prediction of Ki-67 status in GISTs and may serve as adjuncts to risk stratification. However, substantial heterogeneity in specificity, limited external validation, and the exclusively retrospective evidence base preclude routine clinical implementation and warrant confirmation in prospective multinational studies.