Clinical evaluation of an automated Alberta Stroke Program Early Computed Tomography Score (ASPECTS)-scoring system

Area of Science:

• Neurology
• Radiology
• Medical Imaging Analysis

Background:

• The Alberta Stroke Program Early Computed Tomography Score (ASPECTS) is crucial for evaluating early ischemic changes on non-contrast computed tomography (NCCT) in acute stroke.
• Automated ASPECTS tools require independent validation to understand their impact on human interpretation and clinical decision-making.

Purpose of the Study:

• To evaluate the agreement between an automated ASPECTS scoring system and expert human readers.
• To determine if software assistance influences human ASPECTS scoring in a systematic way.

Main Methods:

• An automated ASPECTS system using image normalization, registration, and net water uptake (NWU) analysis was developed.
• 224 retrospective cases were scored by unassisted expert readers and software-assisted readers.
• Agreement was quantified using intraclass correlation coefficient (ICC), mean difference (MD), and mean absolute difference (MAD).

Main Results:

• High inter-rater agreement was observed within both unassisted (ICC 0.91-0.96) and assisted (ICC 0.88-0.95) groups.
• Agreement between unassisted and assisted readings was also high (ICC 0.90-0.96), with no significant systematic shift in scores.
• The automated system showed good-to-moderate agreement with human ratings (ICC 0.76-0.89), performing slightly lower than inter-reader agreement.

Conclusions:

• The automated ASPECTS system demonstrates acceptable agreement with expert human readers.
• Software assistance did not introduce a systematic bias in human scoring, supporting its potential use.
• This automated tool could promote standardized ASPECTS evaluation, especially where expert access is limited, while preserving clinical judgment.