Deep Learning Radiomics Based on Preoperational Ultrasound Images for Predicting Ipsilateral Ischemic Stroke in...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Stroke Prevention

Background:

• Carotid artery stenting (CAS) is a procedure to prevent ischemic stroke.
• Predicting long-term stroke risk after CAS remains challenging.
• Current methods may not fully capture complex plaque characteristics.

Purpose of the Study:

• To develop and validate an integrated model for predicting long-term ipsilateral ischemic stroke risk post-CAS.
• To incorporate clinical, radiomic, and deep learning features for enhanced prediction.
• To improve risk stratification and guide clinical decision-making.

Main Methods:

• Analysis of 802 patients undergoing CAS between 2018-2024.
• Extraction of radiomic and deep learning features from preoperative plaque ultrasound images.
• Development of a combined predictive model using Cox regression and random survival forest, presented as a nomogram.

Main Results:

• A significant proportion of patients (26.6%) experienced ipsilateral stroke over a median 62-month follow-up.
• The integrated deep learning radiomics model showed strong predictive performance (C-indices 0.800, 0.751, 0.708 across datasets).
• The model significantly outperformed conventional ultrasound methods and effectively stratified patients into high- and low-risk groups.

Conclusions:

• A preoperative deep learning radiomics model using ultrasound effectively predicts long-term ipsilateral stroke risk in CAS patients.
• This model offers a valuable tool for risk stratification.
• It can guide treatment decisions and follow-up management for improved patient outcomes.