Machine Learning Application for Rupture Risk Assessment in Small-Sized Intracranial Aneurysm

Area of Science:

• Neurosurgery
• Medical Imaging
• Artificial Intelligence

Background:

• Intracranial aneurysms (IA) pose a rupture risk, necessitating accurate probability assessment, especially in patients with multiple aneurysms.
• Identifying high-risk small-sized aneurysms is critical for timely intervention and patient management.

Purpose of the Study:

• To develop and evaluate a computer-assisted detection system for predicting small-sized intracranial aneurysm ruptures.
• To utilize a convolutional neural network (CNN) trained on 3D digital subtraction angiography images for rupture risk assessment.

Main Methods:

• A CNN model was developed and trained on retrospective data from 368 patients using the TensorFlow platform.
• Aneurysm images from six directions were processed, and regions of interest were extracted.
• The CNN system was prospectively tested on 272 patients and compared against a human evaluator.

Main Results:

• The CNN system achieved a sensitivity of 78.76%, specificity of 72.15%, and overall accuracy of 76.84% in predicting aneurysm rupture.
• The area under the ROC curve (AUROC) for the CNN was 0.755, significantly outperforming the human evaluator's AUROC of 0.537 (p < 0.001).

Conclusions:

• A CNN-based prediction system is feasible for assessing rupture risk in small-sized intracranial aneurysms.
• The developed system demonstrated diagnostic accuracy superior to human evaluators, suggesting potential for clinical application.
• Further validation with larger datasets is recommended to enhance accuracy and clinical utility.