ASPECTS Software aids non-neuroradiologists & improves ASPECT scoring

A study my colleagues and I conducted showed that Rapid ASPECTS can help non-neuroradiologists improve their ASPECT scoring of non-contrast CT (NCCT) scans. Our findings were published in the Journal of Stroke and Cerebrovascular Diseases.

Key takeaways:

• With Rapid ASPECTS assistance, scores assigned by non-neuroradiologists were more comparable with the expert reference standard .
•  Software assistance increased agreement across non-neuroradiologist readers.
• The study found that Rapid ASPECTS can help improve consistency and agreement between readers.

The Alberta Stroke Program Early CT Score (ASPECTS) is determined by evaluating 10 predefined regions of the middle cerebral artery territory for focal hypoattenuation. It helps physicians assess ischemic stroke severity and support treatment decisions.

Reaching consensus between readers can be challenging, with previous studies reporting suboptimal agreement for both total ASPECT scores and individual affected regions. Rapid ASPECTS is an FDA-cleared CADx device designed to assist clinicians with ASPECTS assessment.

Setting up the reference standard

The study included 50 patients with middle cerebral artery or internal carotid occlusions whose NCCT and follow-up MRI scans were available. Three expert neuroradiologists assessed the NCCT scans and used the follow-up MRI scans to establish the reference standard ASPECTS scores.

ASPECT scoring by test readers

The test readers included neuroradiologists and non-neuroradiologists from several specialties. Each reader assessed the NCCT scans twice across two reading sessions, once with Rapid ASPECTS assistance and once without it. For 80% of scans, readers were told which hemisphere was affected to mimic clinical practice.

With 50 cases and 10 ASPECTS regions, each test reader evaluated 500 regions in each condition. Agreement between the test-reader scores and the reference standard was then assessed.

Rapid ASPECTS improved agreement across readers

Among non-neuroradiologists, agreement with the reference standard was 72% without software and increased to 78% with Rapid ASPECTS. When the affected hemisphere was shared before review, agreement reached 80% with the software.

Rapid ASPECTS alone agreed with the reference standard in 77% of regions. The test neuroradiologists agreed with the reference standard in 81% of regions with no change between assisted and unassisted reads. Software assistance benefited readers most in the insula, caudate and M2 regions.

Rapid ASPECTS can support more consistent ASPECTS assessment, particularly where neuroradiologists are not immediately available.