Performance of Automated ASPECTS Software and Value as a Computer-Aided Detection Tool

Area of Science:

• Neuroradiology
• Stroke Imaging
• Quantitative Analysis

Background:

• The Alberta Stroke Program Early CT Score (ASPECTS) assesses early ischemic changes on non-contrast CT (NCCT) in anterior circulation stroke.
• Interrater variability in conventional ASPECTS scoring can impact clinical decision-making, particularly for thrombectomy eligibility.

Purpose of the Study:

• To evaluate the agreement of automated ASPECTS software compared to conventional scoring by multiple raters.
• To assess the impact of computer-aided detection on improving scoring accuracy and consistency in acute ischemic stroke.

Main Methods:

• Retrospective analysis of NCCT scans from 175 acute ischemic stroke patients with large-vessel occlusion undergoing thrombectomy.
• Five raters independently scored conventional ASPECTS; automated ASPECTS was generated using RAPID software.
• Agreement was assessed against a criterion standard, both for full ASPECTS and dichotomized scores (0-5 vs. 6-10) relevant to thrombectomy guidelines.

Main Results:

• Automated ASPECTS showed moderate agreement (weighted κ = 0.59) with the criterion standard, comparable to individual raters.
• Software assistance significantly improved both individual rater agreement with the criterion standard and interrater agreement (Fleiss κ increased).
• Improved agreement was observed for both full and dichotomized ASPECTS scoring when using automated outputs.

Conclusions:

• Automated ASPECTS provides a level of agreement with the criterion standard similar to conventional scoring.
• Integrating automated ASPECTS into the NCCT evaluation process enhances agreement and reduces interrater variability.
• This computer-aided approach holds potential for more standardized and reliable ASPECTS scoring in acute stroke management.