Accuracy of Radiomics-Based Models in Distinguishing Between Ruptured and Unruptured Intracranial Aneurysms: A Systematic Review and Meta-Analysis Publisher Pubmed



Sohrabiashlaghi A¹ ; Azizi N¹ ; Abbastabar H¹ ; Shakiba M¹ ; Zebardast J¹ ; Firouznia K¹ Show Affiliations
Source: European Journal of Radiology Published:2024

Abstract

Introduction: Intracranial aneurysms (IAs) pose a severe health risk due to the potential for subarachnoid hemorrhage upon rupture. This study aims to conduct a systematic review and meta-analysis on the accuracy of radiomics features derived from computed tomography angiography (CTA) in differentiating ruptured from unruptured IAs. Materials and Methods: A systematic search was performed across multiple databases for articles published up to January 2024. Observational studies analyzing CTA using radiomics features were included. The area under the curve (AUC) for classifying ruptured vs. unruptured IAs was pooled using a random-effects model. Subgroup analyses were conducted based on the use of radiomics-only features versus radiomics plus additional image-based features, as well as the type of filters used for image processing. Results: Six studies with 4,408 patients were included. The overall pooled AUC for radiomics features in differentiating ruptured from unruptured IAs was 0.86 (95% CI: 0.84–0.88). The AUC was 0.85 (95% CI: 0.82–0.88) for studies using only radiomics features and 0.87 (95% CI: 0.83–0.91) for studies incorporating radiomics plus additional image-based features. Subgroup analysis based on filter type showed an AUC of 0.87 (95% CI: 0.83–0.90) for original filters and 0.86 (95% CI: 0.81–0.90) for studies using additional filters. Conclusion: Radiomics-based models demonstrate very good diagnostic accuracy in classifying ruptured and unruptured IAs, with AUC values exceeding 0.8. This highlights the potential of radiomics as a useful tool in the non-invasive assessment of aneurysm rupture risk, particularly in the management of patients with multiple aneurysms. © 2024 Elsevier B.V.