Rupture discrimination of multiple small (< 7 mm) intracranial aneurysms based on machine learning-based cluster...

Area of Science:

• Neurosurgery
• Medical Informatics
• Radiology

Background:

• Small multiple intracranial aneurysms (SMIAs) present a higher rupture risk than single aneurysms.
• Current guidelines lack specific recommendations for managing SMIAs.
• Risk stratification for SMIA rupture is crucial for patient management.

Purpose of the Study:

• To evaluate the feasibility of machine learning-based cluster analysis for discriminating SMIA rupture risk.
• To identify distinct patient clusters based on morphological and clinical features.
• To develop a predictive model for SMIA rupture.

Main Methods:

• A multi-institutional cross-sectional study involving 1,427 SMIAs from 660 patients.
• Hierarchical cluster analysis was used for patient classification.
• Machine learning models were constructed and compared using morphological features and identified clusters.

Main Results:

• Three distinct clusters with varying subarachnoid hemorrhage (SAH) risks were identified.
• Cluster 1 showed the highest SAH risk (75.6%) with a higher prevalence of familial aneurysms.
• Cluster 2 had a moderate SAH risk (38.2%) and more vascular risk factors; Cluster 3 had a lower risk (17.6%).
• Lasso regression indicated significantly higher rupture risk in Cluster 1 (OR 7.391) and Cluster 2 (OR 3.014) compared to Cluster 3.
• The developed model achieved an area under the curve of 0.828 for discriminating SMIA rupture.

Conclusions:

• Unsupervised machine learning effectively identified three clusters with differential SAH risk in SMIA patients.
• The proposed model, utilizing morphological factors and identified clusters, demonstrates good discrimination ability for SMIA ruptures.
• This approach offers a promising tool for risk stratification and management of SMIAs.