Prediction of Stroke Infarct Growth Rates by Baseline Perfusion Imaging.

Area of Science:

• Neuroimaging
• Artificial Intelligence in Medicine
• Stroke Imaging

Background:

• Computed tomography perfusion (CTP) imaging is crucial for assessing tissue status in acute ischemic stroke.
• Accurate prediction of final infarct volume and growth rate is essential for effective stroke management.

Purpose of the Study:

• To develop and validate a deep learning (DL) approach for improved prediction of final infarct volume and individual infarct growth rates in acute ischemic stroke patients.
• To compare the DL model's performance against standard CTP processing software.

Main Methods:

• A deep neural network was trained to predict final infarct volume using CTP images, time to reperfusion, and reperfusion status.
• The model was trained on the MR CLEAN trial cohort and validated on the independent CRISP study cohort.
• Performance was assessed by comparing the mean absolute difference in infarct volume prediction against RAPID software.

Main Results:

• The DL model demonstrated superior performance in predicting final infarct volume compared to RAPID software in both derivation (34.5 mL vs. 52.4 mL) and validation cohorts (41.2 mL vs. 52.4 mL).
• The study successfully obtained individual infarct growth rates, allowing for final infarct volume estimation based on reperfusion parameters.

Conclusions:

• A validated deep learning method significantly enhances the accuracy of final infarct volume estimation in acute ischemic stroke compared to traditional CTP processing.
• The DL model's ability to predict individual infarct growth rates may facilitate the implementation of 'tissue clocks' in acute stroke care.