Multicenter Validation of Artificial Intelligence Predicting Anterior Circulation Large Vessel Occlusion Using...

Area of Science:

• Neurology
• Radiology
• Artificial Intelligence

Background:

• Large vessel occlusion (LVO) is a critical condition in stroke requiring rapid diagnosis.
• Noncontrast computed tomography (NCCT) is a primary imaging modality for stroke assessment.
• Artificial intelligence (AI) offers potential for improving diagnostic accuracy and efficiency.

Purpose of the Study:

• To validate an AI software (JLK CTL) for predicting anterior circulation LVO on NCCT.
• To assess the clinical implications of the AI's LVO predictions concerning infarct volume and patient outcomes.

Main Methods:

• Retrospective analysis of 534 patients from 6 stroke centers who underwent NCCT and CT angiography.
• Inclusion of 274 non-stroke subjects to evaluate AI specificity.
• Evaluation of AI performance using ROC curve, sensitivity, and specificity.
• Correlation analysis between AI LVO scores, infarct volumes, and functional outcomes.

Main Results:

• The AI software achieved an Area Under the ROC Curve of 0.859, with a sensitivity of 0.787 and specificity of 0.832 for LVO detection.
• The AI demonstrated high specificity (0.898) in non-stroke subjects.
• JLK CTL LVO scores correlated significantly with infarct volumes (r=0.54, P<0.001).
• Each 1-point increase in JLK CTL LVO score was associated with a 2% increase in unfavorable 3-month outcomes (P=0.011).

Conclusions:

• The AI software demonstrates validated performance in predicting LVO on NCCT.
• The AI's LVO score is associated with infarct volume and functional outcomes, supporting its clinical utility.
• The AI tool can aid in screening patients and informing rapid clinical decision-making for stroke management.