The stroke outcome optimization project: Acute ischemic strokes from a comprehensive stroke center.

Area of Science:

• Neurology
• Radiology
• Medical Imaging
• Machine Learning

Background:

• Stroke is a major cause of disability, necessitating advanced diagnostic tools.
• Magnetic Resonance Imaging (MRI) is crucial for acute stroke management.
• Publicly available, diverse datasets are vital for developing robust machine learning models in healthcare.

Purpose of the Study:

• To create and share a comprehensive dataset of acute ischemic stroke brain MRI scans.
• To enable the development of machine learning algorithms for stroke lesion identification, brain health assessment, and prognosis.
• To provide a benchmark dataset and reproducible methods for stroke research.

Main Methods:

• Acquisition of clinical MRI data (diffusion-weighted, FLAIR, T1-weighted) from 1715 individuals, including 1461 with acute ischemic stroke.
• Collection of demographic and impairment data (NIH Stroke Scale/Score - NIHSS) for 1106 stroke survivors.
• Development and validation of machine learning models to predict stroke severity using imaging data.

Main Results:

• A large, diverse dataset of acute ischemic stroke MRI scans and associated clinical data is now publicly available.
• Machine learning models successfully predicted stroke severity (NIHSS) using the provided imaging data.
• The study provides reproducible scripts, facilitating further research and validation.

Conclusions:

• The shared dataset and methods support the advancement of AI-driven stroke diagnosis and management.
• This resource can accelerate the development of more accurate and generalizable machine learning tools for stroke prognosis.
• Public data sharing and reproducible research are essential for progress in computational neuroimaging.