Creating an Artificial Intelligence-Driven Chatbot to Triage Patients for Mechanical Thrombectomy

Introduction: Mechanical thrombectomy (MT) is the standard of care in patients presenting with acute ischemic stroke (AIS) due to large vessel occlusion (LVO). Early recanalization is critical to restoring blood flow to ischemic brain tissue. Previous literature has reported disparities in access to MT capable centers in patients with AIS. Prompt identification of MT eligibility could help stratify access to MT capable centers. Our study created an artificial intelligence (AI)-driven chatbot to help determine eligibility for MT.

Methods: Guidelines established by the American Heart Association (AHA) and evidence from randomized clinical trials were used to determine MT eligibility based on presence of LVO, absence of intracranial hemorrhage (ICH), National Institute of Health Stroke Scale (NIHSS) and ASPECTS score. These criteria were utilized to develop a custom chatbot using the Python library for Open AI (San Francisco, California). The chatbot was used to determine eligibility and report reasons for each decision. Responses on MT eligibility were compared to the surgeon’s decision on MT eligibility at our institution between May 2024- August 2024.

Results: 34 patients presented with AIS and 52.9% (n= 18) underwent MT at the discretion of the treating surgeon. The chatbot’s response agreed with the surgeon’s decision in 61.8% patients (n= 21). There were 13 cases of disagreement (38.2%) where the chatbot deemed patients to be ineligible for MT, however, the surgeon chose to perform MT. The chatbot was made available in the form of an open access web application. https://huggingface.co/spaces/thrombectomypredictor/MT_eligibility

Conclusion : Our chatbot provides a user-friendly interface to identify MT eligibility. This could be incorporated into AI-based imaging software to streamline the process of referrals to MT-capable centers.