Abstract
Background/Objectives: Computed tomography (CT)-based radiomic analysis is an emerging technique that enables non-invasive assessment of tumor characteristics. In gastrointestinal stromal tumors (GISTs), radiomics may reflect biological behavior such as proliferative activity, often indicated by Ki-67 expression. To our knowledge, this is the first systematic review and meta-analysis synthesizing evidence on the ability of CT radiomics to predict the Ki-67 index in GISTs, addressing an important gap in the literature. Methods: A systematic review and meta-analysis were conducted following PRISMA guidelines to evaluate the predictive performance of CT radiomics for Ki-67 expression in GISTs. A literature search of PubMed, Scopus, Science Direct, and the Cochrane Library was performed up to December 2024 using predefined terms. Extracted data included study design, patient demographics, imaging protocols, radiomic features, and diagnostic performance. Study quality was assessed using the QUADAS-2 tool. A random-effects meta-analysis summarized the pooled area under the ROC curve (AUC), sensitivity, and specificity. Subgroup and sensitivity analyses explored heterogeneity sources. Publication bias was assessed using Egger's test and funnel plots. Results: Six studies involving 1632 patients were included. The pooled sensitivity and specificity for predicting Ki-67 expression were 0.71 and 0.76, respectively, with a summary AUC of 0.79. Subgroup analyses showed consistent results across different imaging protocols and radiomic feature sets, though the Ki-67 cutoff (8% vs. 10%) affected diagnostic performance. Moderate heterogeneity and potential publication bias in specificity were observed. Conclusions: CT-based radiomics demonstrates moderate accuracy for non-invasively predicting Ki-67 index in GISTs. While not a substitute for histology, it may support personalized preoperative planning and guide future immunotherapy strategies. In the future, radiomic signatures-particularly when integrated with molecular or immune-related biomarkers-could help refine patient selection and monitoring strategies for emerging therapies, including immunotherapy.