14488
clinical study

Retrospective Detection of Missed Intracranial Aneurysms on Computed Tomography Angiography Using a Commercial Deep Learning Algorithm

Materials & Methods
This retrospective study evaluated a commercial deep learning algorithm (Aidoc, Tel Aviv, Israel) for detecting missed intracranial aneurysms (IAs) on head CT angiography (CTA) performed between Feb. 2020 and July 2022 at a tertiary referral center. Natural language processing (NLP) classified radiology reports as positive or negative for aneurysms, and a convolutional neural network (CNN) analyzed imaging data. Discordant cases between the AI output and reports were reviewed by three neuroradiologists, with majority consensus determining the reference standard.

Results
Among 2,615 CTA studies, the algorithm identified 34 suspected missed aneurysms, with 67% (23/34) confirmed as true positives — representing a 20.9% increase in detection (23/110 total aneurysms) and 0.88% of all studies. Most missed aneurysms were ≤3 mm. Overall, the algorithm achieved 96.36% sensitivity, 99.56% specificity, 90.6% positive predictive value and 99.84% negative predictive value.

Conclusions
The deep learning algorithm demonstrated high accuracy in detecting missed intracranial aneurysms, especially small lesions often overlooked in routine interpretation. Integration of such AI tools into clinical workflow may enhance diagnostic precision, ensure appropriate follow-up and improve patient outcomes through earlier detection of actionable aneurysms.

Explore the Latest AI Insights, Trends and Research