P163 Detection of intracranial aneurysms on 3D rotational digital subtraction angiography using neural networks

<h3>Introduction</h3> Automatic detection of intracranial aneurysms is challenging and currently available commercial software is not able to detect an intracranial aneurysm in all cases. An undetected aneurysm may rebleed and cause the death of the patient. Artificial intelligence can potentially be helpful in reducing the rate of failure. <h3>Aim of Study</h3> The purpose of the study was to configure a neural network able to detect intracranial aneurysm on 3D rotational angiography and to test its precision and recall. <h3>Methods</h3> The initial data set was created by anonymizing 11 patients with positive aneurysm findings on standard intracranial 3D rotational angiography and 2 cases with negative aneurysm findings. The data set contains a total of 1557 images. <h3>Results</h3> The average precision across the models was 93.86%, while the average recall was 29.75%. <h3>Conclusion</h3> The results of this initial study suggest that neural networks have the potential to be used in clinical practice for the detection of intracranial aneurysms.