CT Reconstruction Levels Affect Automated and Reader-Based ASPECTS Ratings in Acute Ischemic Stroke

Background and Purpose: This study investigated whether automated and reader-based ASPECTS in acute stroke patients are affected by different CT image reconstruction algorithms.

Methods: ASPECTS were assessed by commercial software and four independent blinded readers (two residents and two consultants) from different CT reconstructions (filtered back projection and two iterative reconstruction [IR] levels) in 43 acute stroke patients with proximal MCA occlusion. Ground truth was provided by an expert with unrestricted data access.

Results: The residents showed significant variations between IR levels and lower internal consistency across reconstructions compared to the software, which performed similarly to the consultants. The consultant and the software showed different deviations from ground truth across IR levels, least at IR strength level 2.

Conclusions: CT image postprocessing affects both automated and human ASPECTS in acute stroke patients. This effect was most pronounced in less experienced readers, while the software had the most robust performance.