|
MRI-based radiomics models for pre-operative prediction of microsatellite instability in rectal cancer: A systematic review and meta-analysis
|
B. Wang   , J. Gu , B. Wu |
|
|
|
Abstract: (6 Views) |
Background: This study aimed to comprehensively assess the diagnostic capability of radiomics models derived from magnetic resonance imaging in determining microsatellite instability (MSI) status among patients with rectal cancer, while exploring variables that may affect predictive accuracy. Materials and Methods: We performed a comprehensive literature search across PubMed, EMBASE, Cochrane Library, and Web of Science databases through August 2025, targeting studies employing MRI-derived radiomics for MSI status prediction in rectal malignancies. Summary estimates for sensitivity, specificity, and the area under the summary receiver operating characteristic curve (SROC AUC) were derived through random-effects meta-analysis. Heterogeneity sources were investigated via subgroup stratification. Results: Thirteen eligible investigations encompassing 3,292 individuals with rectal cancer were incorporated. For validation cohorts, pooled estimates demonstrated a sensitivity of 0.88, specificity of 0.82, and SROC AUC of 0.91. According to Fagan nomogram interpretation, positive radiomics findings elevated MSI post-test probability from an initial 18% to 52%, whereas negative findings lowered it to 3%. Meta-regression revealed classifier selection as the predominant determinant of performance variability. Robustness was substantiated through sensitivity testing, with Deeks' funnel plot asymmetry test showing no substantial publication bias (P=0.93). Conclusion: Radiomics models based on MRI exhibit robust diagnostic performance for preoperative MSI status prediction in rectal cancer, demonstrating promise as non-invasive screening instruments to support treatment decision-making. Nevertheless, considering the methodological constraints and inter-study variability observed, rigorously designed multicenter prospective investigations are imperative to establish their clinical utility and economic feasibility. |
|
| Keywords: Rectal neoplasms, microsatellite instability, radiomics, magnetic resonance imaging, machine learning. |
|
|
Full-Text [PDF 1319 kb]
(1 Downloads)
|
|
Type of Study: Review article |
Subject:
Radiation Biology
|
|
|
|
|
|
|
| Add your comments about this article |
|
|
|
