Artificial Intelligence in Stroke Imaging: A Review of Current Applications and Limitations.

Abstract

Stroke is a major global health burden, requiring time-sensitive diagnosis and treatment to improve patient outcomes. This urgency has created a compelling role for artificial intelligence in the stroke imaging workflow to accelerate diagnosis and treatment. Artificial intelligence has demonstrated a significant impact across multiple aspects of stroke care, including automated detection of acute findings, expedited triage and notification of findings, quantitative assessment of infarcts, predictive prognostication of outcomes, as well as acceleration of image acquisition. However, these advances are accompanied by important limitations including introduction of biases and challenges in the real-world clinical integration of such tools. In this review, we examine the current applications of artificial intelligence in stroke imaging and evaluate the limitations and real-world implementation challenges.