Machine Learning-Enabled Automated Large Vessel Occlusion Detection Improves Transfer Times at Primary Stroke Centers

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Optimizing door-in-door-out (DIDO) times is crucial for acute ischemic stroke patients with large vessel occlusion (LVO) requiring transfer for endovascular stroke therapy (EVT).
• Primary stroke centers (PSCs) play a key role in the initial management of these critical cases.

Purpose of the Study:

• To evaluate the impact of automated LVO detection via artificial intelligence (AI) on DIDO times at PSCs.
• To determine if AI implementation increases the proportion of patients receiving EVT after transfer.

Main Methods:

• A prospective multicenter registry identified LVO acute ischemic stroke patients at 7 Greater Houston PSCs.
• A machine learning (AI) algorithm (Viz.AI) for LVO detection was implemented across all participating hospitals.
• DIDO time was the primary outcome, analyzed using multivariable linear regression; EVT likelihood was a secondary outcome.

Main Results:

• Among 115 patients (80 pre-AI, 35 post-AI), automated LVO detection was associated with a significant reduction in DIDO time (106 minutes).
• The most common occlusions were in the middle cerebral artery (51.3%) and internal carotid artery (25.2%).
• No significant difference was observed in the likelihood of receiving EVT post-transfer (OR, 2.13).

Conclusions:

• Implementing AI for automated LVO detection significantly decreases DIDO times at PSCs.
• This technology streamlines patient evaluation, potentially improving outcomes for stroke patients requiring transfer for advanced care.