Evaluation of the Accuracy Equivalence of Head CT Interpretations in Acute Stroke Patients Using a Smartphone, a Laptop, or a Medical Workstation

Abstract

Purpose: To evaluate and compare the clinical performance of observers interpreting head CT images from patients with symptoms of acute stroke with a medical workstation or a smartphone or laptop reading system.

Materials and methods: Our institutional review board approved this retrospective study and waived the requirement for informed consent. We employed a factorial design including 2,256 interpretations (188 patients × 4 neuroradiologists × 3 reading systems). Accuracy equivalence tests, at a threshold of 5% and 10%, were performed for the following variables: detection of imaging contraindications for intravenous recombinant tissue-type plasminogen activator administration (eg, hemorrhagic lesions), ischemic lesions, hyperdense arteries, and acute ischemic lesions. For each clinical variable, the sensitivity, specificity, and receiver operating characteristic (ROC) curves were evaluated.

Results: For each variable, the shapes of the ROC curves were very similar for all of the reading systems, indicating similar observer performance with different specificities and sensitivities. For all the clinical variables, the areas under the ROC curves were equivalent for all of the reading systems at a 10% threshold and were equivalent at a 5% threshold for hemorrhagic lesions, hyperdense middle cerebral artery, and acute ischemic lesion in the middle cerebral artery territory. There was no more than 30 seconds of difference between the reading time of the mobile devices compared with that for the medical workstation.

Conclusion: The results of this pilot study showed equivalent diagnostic accuracy when using smartphone and laptops compared with medical monitors to interpret head CT images of patients with symptoms of acute stroke.

Keywords: Head CT; stroke; telestroke.