A review published in the Journal of NeuroInterventional Surgery in early October shared hopeful news in the ability of artificial intelligence (AI) to help quickly diagnose ischemic strokes and whether they have been caused by large vessel occlusions. The review covered studies published over the last five years up until February.
The research team was led by Stanford’s Nick Murray, and the review was completed in collaboration with Johns Hopkins. The team pulled studies from PubMed and two other literature databases, picking 20 total that wove different types of AI into stroke research that incorporated large vessel occlusion.
They found that in terms of types of machine learning, convolutional neural networks (CNN) was more sensitive than random forest learning (RFL): the former clocked in at 85%, while the latter at 68%. Additionally, the review discovered that physicians generally make use of CNN to find large vessel occlusions, and RFL to detect strokes early.
“Acute stroke caused by large vessel occlusion requires emergent detection and treatment by endovascular thrombectomy,” Murray and his team wrote. “However, radiologic large vessel occlusion detection and treatment is subject to variable delays and human expertise, resulting in morbidity. Imaging software using artificial intelligence and machine learning… may improve rapid frontline detection of large vessel occlusion strokes.”
The team also remarked that AI has the potential to help detect strokes that are caused by large vessel occlusions in a major way, because the technology is able to carry out quick interventions that are necessary to prevent long-term disabilities or perhaps even death. They did, however, also say that experts should create a standard of metrics for future studies, because the studies they used in the review often used different metrics to measure the performances of their AIs. Nonetheless, the potential for AI to aid in stroke detection and harm prevention is incredible.
