Diffusion-weighted imaging showing 'pearls' predicts large-vessel disease as stroke etiology

The patterns on diffusion-weighted (DWI) MRI may be predictive of stroke etiology. In this retrospective study, we assessed whether DWI bright lesions termed 'pearls' predicted the presence of large-vessel arterial stenosis as the etiology of stroke.

All stroke and transient ischemic attack admissions to an academic hospital over a 2-year period were reviewed. Patients without DWI, with hemorrhagic strokes or with nonvascular disease were excluded. Two vascular neurologists reviewed medical records and classified the probable stroke etiology by modified TOAST (Trial of Org 10172 in Acute Stroke Treatment) criteria. Another investigator, blinded to the clinical diagnoses, reviewed MRI images only. A 'pearl' was defined as a DWI bright lesion 20 mm or less in diameter. A 'string of pearls' was defined as 3 or more pearls in a line found unilaterally in the anterior circulation. 'Scattered pearls' were classified as 3 or more pearls distributed such that no single line could connect them, also found unilaterally in the anterior circulation.

Fifty-six percent of the patients with either 'pearls' sign were classified as TOAST 1 or 2, with this classification only found in 33% of the patients with other DWI patterns (p = 0.0009). If either 'pearl' sign was seen on MRI, a patient was 2.65 times (95% CI = 1.32-5.32) more likely to have either extracranial or intracranial large-vessel stenosis.

The presence of either 'string of pearls' or 'scattered pearls' on MRI was associated with an independently determined mechanism of intracranial or extracranial arterial stenosis. In patients with acute stroke distributions consistent with either of these patterns, additional vascular imaging should be considered.