Radiomics model for risk stratification of intracranial aneurysm: a high-resolution vessel wall imaging-based study

Area of Science:

• Radiology and Medical Imaging
• Cardiovascular Imaging
• Artificial Intelligence in Medicine

Background:

• High-resolution vessel wall imaging (HR-VWI) allows in vivo assessment of aneurysm wall pathology.
• Current qualitative evaluation methods for aneurysm wall pathology are limited.
• Accurate identification of symptomatic intracranial aneurysms (SIAs) is crucial for risk stratification.

Purpose of the Study:

• To develop and validate an HR-VWI-based radiomics model for identifying SIAs.
• To integrate aneurysm wall and parent artery wall features for improved risk stratification.
• To compare the novel radiomics model's performance against existing methods like the PHASES score.

Main Methods:

• Retrospective analysis of 446 intracranial aneurysms (IAs) from 410 patients across two centers.
• Extraction of 851 radiomic features from HR-VWI images of aneurysm walls and parent arteries (PAs) using Pyradiomics.
• Development of multiple radiomics models (Radscore_IA, Radscore_PA, Radscore_IA_PA) and location-integrated versions (Radscore_LOC_IA, Radscore_LOC_PA, Radscore_LOC_IA_PA), validated using AUC.

Main Results:

• Radiomics features from IA and PA walls were identified as significant for SIA identification.
• The combined location-integrated model, Radscore_LOC_IA_PA, achieved the highest AUC (0.888) in the validation cohort.
• All developed radiomics models demonstrated superior performance compared to the PHASES score (AUC 0.679) and showed robustness in calibration and decision curve analyses.

Conclusions:

• The novel HR-VWI radiomics model, Radscore_LOC_IA_PA, effectively identifies high-risk SIAs by integrating IA wall, PA wall, and location features.
• This model offers improved discrimination for SIA identification compared to traditional stratification methods.
• The findings support the clinical application of this radiomics model for more accurate risk stratification and management of patients with intracranial aneurysms.