Comprehensive Summary
This study used computed tomography angiography (CTA) scans to see whether an automated artificial intelligence (AI) framework could predict rupture risk in cerebral aneurysms. An external dataset scanned with 3D rotational angiography was used to validate the model after researchers retrospectively examined 335 individuals with 500 aneurysms, evenly split between ruptured and unruptured instances. For vascular and aneurysm segmentation, the platform employed deep learning models (nnU-net and PointNet++), which extracted intricate shape features in addition to standard variables like patient demographics and aneurysm location. With a 78.2% accuracy rate and an area under the receiver operating characteristic curve (AUROC) of 0.851, the random forest classifier proved to be the most successful model. The authors discovered that vascular form features had less of an impact on rupture prediction than aneurysm shape features. With an AUROC of 0.805, external validation validated the model's resilience.
Outcomes and Implications
Since cerebral aneurysm rupture frequently causes potentially fatal subarachnoid hemorrhage, the ability to accurately evaluate rupture risk is essential. Physician judgment is a major component of current evaluation methodologies. In order to support treatment decisions for patients with unruptured aneurysms, this study demonstrates that AI-based analysis of routine CTA scans may offer reliable, accurate, and non-invasive risk assessment. The work's performance across various imaging modalities indicates high clinical potential, even though it is currently in the validation stage. Such AI techniques have the potential to standardize risk assessment and potentially improve outcomes for individuals at risk of aneurysm rupture with additional prospective research and incorporation into clinical procedures.