Keywords:
Artificial Intelligence, Neuroradiology brain, CT, CAD, Computer Applications-Detection, diagnosis, Ischaemia / Infarction
Authors:
A. Ayobi, P. Chang, D. Chow, C. Filippi, S. Quenet, M. Tassy, Y. Chaibi
DOI:
10.26044/ecr2023/C-19206
Purpose
Stroke is a major disease of the 21st century. Despite significant improvements in primary prevention and treatment, stroke remains a devastating disease and it is considered an absolute emergency [1, 2]. Ischemic stroke (IS), the most common type of stroke, consists of a brain blood vessel occlusion that creates a lack of blood flow and results in the death of brain cells within the first few minutes [3]. In the treatment and diagnosis of IS "time is brain" [4]. Hence, non-contrast CT (NCCT) remains the primary imaging modality for the initial assessment of suspected stroke patients due to its quick availability [5].
A standardization of this procedure has been developed through the Alberta Stroke Program Early Computed Tomography Score (ASPECTS) by dividing the territory of the Middle Cerebral Artery (MCA) and/or Internal Carotid Artery (ICA) into 10 regions (internal capsule (IC), caudate (C), lentiform (L), insula (I), M1, M2, M3, M4, M5 and M6) and subtracting one point for each region where early ischemic change (i.e. hypodense regions, low attenuation and/or sulcal effacement) is present [6, 7]. ASPECTS is especially useful in patient management: in the event an ASPECT score is 6 or more, patients are more likely to benefit from mechanical thrombectomy (MT). Patients with a lower ASPECT score are usually treated with intravenous thrombolysis, as reperfusion procedures would be potentially futile [8, 9]. However, the ASPECTS methodology has certain limitations: artifacts, poor image quality and head tilt may induce some errors [5]. In addition, ASPECTS in clinical practice is performed manually by physicians, thus, interobserver variabilities are always present [10].
Nowadays, artificial intelligence (AI) applications are proposed as tools offering a faster, consistent and precise assessment of the ASPECTS methodology [11]. Thus, this study aims to evaluate the performance of an AI-based automated software designed to identify early ischemic change (EIC) on NCCT per the ASPECTS methodology. In addition, the effect of the software use on the accuracy of physicians’ interpretations is also analyzed.