Evaluation of the novel medical imaging software e-ASPECTS for patient selection in stroke

Introduction: The interpretation of a CT scan of acute ischaemic stroke patients in the acute setting requires experience and carries significant variability. The Alberta Stroke Program Early CT score (ASPECTS) is an established 10-point quantitative topographic CT scan score to assess early ischaemic changes on plain CTs of acute stroke patients. We compared the performance of the standardised and fully automated software, e-ASPECTS (Brainomix, www.brainomix.com) with 3 expert neuroradiologists. Method: The baseline non-contrast enhanced CT scans of 132 acute stroke patients were evaluated by 3 expert neuroradiologists and e-ASPECTS using the ASPECTS method. Ground truth was determined by an independent core lab with access to follow-up CT/MR scans. e-ASPECTS is a standardised, fully automated ASPECTS scoring software tool that carries out a 3D registration/segmentation of ASPECTS regions and uses machine learning techniques for scoring. The sensitivity and specificity of e-ASPECTS to the 3 experts was compared by noninferiority analysis.Results: On average, e-ASPECTS deviated from the ground truth by less
than one point (+0.8). For the 3 experts the deviation from the ground truth was +1.2, −0.7 and +1.1, respectively. Receiver operating characteristic (ROC) analysis on per region and per score basis showed that e-ASPECTS sensitivity and specificity is equivalent to the 3 experts. Discussion: e-ASPECTS is statistically non-inferior and equivalent to expert neuroradiologists. e-ASPECTS is a valuable tool to assist patient selection for both intravenous and endovascular stroke treatment.