CT Perfusion improves Detection and Exclusion of Clinically Relevant Medium or Distal Vessel Occlusions - A Multi-Reader Diagnostic Accuracy Study
Year: 2026
Type: article
Study design: Observational Study
Abstract: BACKGROUND AND PURPOSE: Distal/medium vessel occlusions (DMVOs) are frequent and often considered for endovascular therapy. However, detection with non-contrast CT (NCCT) and CT angiography (CTA) is challenging. This study evaluates the impact of CT perfusion (CTP) on diagnostic accuracy, speed, and confidence in DMVO detection among physicians with varying expertise in stroke imaging. MATERIALS AND METHODS: A retrospective, multi-reader diagnostic accuracy study was conducted with paired-patient, paired-reader design. Imaging from 50 patients with DSA-confirmed solitary DMVO, who received endovascular treatment, and 50 patients without stroke, confirmed by follow-up MRI, was included. Five readers of varying training backgrounds evaluated CT images in three sessions: Session 1A included NCCT+CTA of 50 cases (with and without stroke). In Session 1B, readers re-evaluated this cohort with access to CTP. Session 2 involved the remaining 50 mixed cases evaluated with NCCT+CTA+CTP. Primary endpoints were sensitivity and specificity for DMVO detection; secondary endpoints included time to diagnosis and diagnostic confidence. Statistical analysis utilized Bayesian generalized linear mixed models. RESULTS: Sensitivity (84% vs. 100%) and specificity (94% vs. 99%) for DMVO detection improved with CTP, logarithmic odds ratio (sensitivity) of 10 (95%-CI: 3-33). This effect was consistent across all experience levels. Time to diagnosis decreased from 94±24 seconds to 49±24 seconds with CTP, and diagnostic confidence improved (median score increased from 4 to 5). Inter-reader agreement improved substantially, with Fleiss' kappa increasing from 0.62 (Session 1A) to 0.97 (Session 1B). CONCLUSIONS: Incorporating CTP into imaging protocols significantly enhances diagnostic accuracy, speeds diagnosis, and improves diagnostic confidence for detecting DMVO across varying expertise.
DOI: 10.3174/ajnr.a9564
Language: English
Subfield: Epidemiology
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7172344342
FWCI:
Cites:
Cited by:
Related to:
