DWI scrolling artery sign for the diagnosis of giant cell arteritis: a pattern recognition approach.

Objectives: To investigate the diagnostic accuracy of a pattern recognition approach for the evaluation of MRI scans of the head with diffusion-weighted imaging (DWI) in suspected giant cell arteritis (GCA).

Methods: Retrospectively, 156 patients with suspected GCA were included. The 'DWI-Scrolling-Artery-Sign' (DSAS) was defined as hyperintense DWI signals in the cranial subcutaneous tissue that gives the impression of a blood vessel when scrolling through a stack of images. The DSAS was rated by experts and a novice in four regions (frontotemporal and occipital, bilaterally). The temporal, occipital and posterior auricular arteries were assessed in the T1-weighted black-blood sequence (T1-BB). The diagnostic reference was the clinical diagnosis after ≥6 months of follow-up.

Results: The population consisted of 87 patients with and 69 without GCA; median age was 71 years and 59% were women. The DSAS showed a sensitivity of 73.6% and specificity of 94.2% (experts) and 59.8% and 95.7% (novice), respectively. Agreement between DSAS and T1-BB was 80% for the region level (499/624; kappa(κ)=0.59) and 86.5% for the patient level (135/156; κ=0.73). Inter-reader agreement was 95% (19/20; κ=0.90) for DSAS on the patient level and 91.3% (73/80; κ=0.81) on the region level for experts. For expert versus novice, inter-reader agreement for DSAS was 87.8% on the patient level (137/156; κ=0.75) and 91.2% on the region level (569/624; κ=0.77).

Conclusions: The DSAS can be assessed in less than 1 min and has a good diagnostic accuracy and reliability for the diagnosis of GCA. The DSAS can be used immediately in clinical practice.