The Role of Artificial Intelligence in Diagnosing Acute Ischemic Stroke Using Diffusion MRI: A Multicenter External...

Area of Science:

• Medical Imaging and Diagnostics
• Artificial Intelligence in Healthcare
• Neurology

Background:

• Acute ischemic stroke diagnosis is time-sensitive, necessitating rapid and accurate methods.
• Diffusion-Weighted Imaging (DWI) MRI is sensitive but can be enhanced by AI for speed and accuracy.

Purpose of the Study:

• To evaluate and externally validate a DWI-MRI-based deep learning model for automated stroke detection.
• To compare the model's multicenter diagnostic performance against expert radiologists.

Main Methods:

• Retrospective study of 732 patients across three centers.
• Development and internal validation of a deep convolutional neural network (CNN) on 532 cases.
• External validation on 200 independent cases from two additional centers.

Main Results:

• The deep learning model achieved 100% sensitivity, 100% specificity, and 100% accuracy in internal validation (AUC=1.000).
• External validation showed high performance: 100% sensitivity, 98% specificity, 99% accuracy (AUCs ≈ 0.987).
• AI model performance was comparable to expert radiologists across all datasets.

Conclusions:

• The DWI-MRI deep learning model offers high, reliable diagnostic accuracy for acute ischemic stroke.
• Its multicenter generalizability suggests potential as a decision-support tool in emergency settings.
• The AI facilitates faster, more consistent stroke diagnoses, especially where specialists are limited.