The Clinical Utility Of AI Models In Diagnostic And Endovascular Radiology

Abstract

Background: Artificial intelligence (AI) has emerged as a powerful tool in radiology, offering enhanced diagnostic accuracy, improved workflow efficiency, and support for complex interventional procedures. Despite rapid technological advancement, evidence comparing AI-assisted interpretation and procedural guidance with conventional radiology remains limited. This study evaluated the clinical utility of AI models in diagnostic and endovascular radiology using real clinical data.

Methods: An analytical cross-sectional study was conducted on 240 patients, including 160 diagnostic radiology cases and 80 endovascular radiology cases. Diagnostic images (CT, MRI, and radiography) were interpreted by both radiologists and an AI model, with performance compared against a consensus reference standard. Endovascular data were analyzed to assess AI-generated vessel segmentation, catheter pathway prediction, and procedure optimization. Primary outcomes included sensitivity, specificity, accuracy, interpretation time, fluoroscopy time, contrast use, and total procedure duration. Statistical analyses were performed using SPSS version 27.

Results: AI-assisted diagnostic interpretation demonstrated higher sensitivity (90.6%), specificity (86.3%), and overall accuracy (88.8%) compared with conventional interpretation. AI also reduced interpretation time, with 81.3% of cases completed in under five minutes versus 42.5% conventionally. In endovascular radiology, AI improved procedural efficiency, reducing fluoroscopy time in 72.5% of cases and lowering contrast volume in 65%. Shorter procedure duration was observed in 75% of AI-assisted evaluations. Catheter path prediction achieved high accuracy in 77.5% of cases. Radiologist satisfaction was favorable, with 87.5% expressing satisfaction or strong satisfaction with AI outputs.

Conclusion: AI models significantly enhanced diagnostic accuracy, workflow efficiency, and procedural performance across both diagnostic and endovascular radiology. These findings support the integration of AI into clinical radiology practice as a means of augmenting clinician performance and improving patient outcomes. Continued validation and responsible implementation are necessary to ensure safe and effective adoption of AI technologies in routine radiological workflows.