Related Experiment Video
Updated: Aug 29, 2025
09:21
Optimized Management of Endovascular Treatment for Acute Ischemic Stroke
Published on: January 18, 2018
12.1K
Identifying acute ischemic stroke patients within the thrombolytic treatment window using deep learning
Jennifer S Polson1,2, Haoyue Zhang1,2, Kambiz Nael3
1Computational Diagnostics Lab, University of California, Los Angeles, Los Angeles, California, USA.
Summary
A new deep learning model automatically estimates time since stroke (TSS) from MRI scans, performing similarly to expert radiologists. This technology aids acute ischemic stroke treatment by providing rapid TSS classification without needing specialized expertise.
Area of Science:
- Neurology
- Radiology
- Artificial Intelligence
Background:
- Acute ischemic stroke treatment is time-sensitive, with outcomes directly related to the time since stroke (TSS).
- Diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) mismatch are established imaging biomarkers for estimating TSS.
- Current methods often require specialized neuroradiology expertise.
Purpose of the Study:
- To develop an automated deep learning technique for determining TSS from MR images.
- To eliminate the need for subspecialist radiology expertise in TSS assessment.
- To provide a rapid and accurate method for classifying TSS in acute ischemic stroke patients.
Main Methods:
- A deep learning network was developed and externally validated using 772 patient MR images.
- The model was trained to classify TSS within a 4.5-hour window.
- Algorithm predictions were compared against neuroradiologist assessments of DWI-FLAIR mismatch.
Main Results:
- The deep learning model achieved comparable performance to neuroradiologists in classifying TSS.
- The best model demonstrated an average accuracy of .726 (internal) and .724 (external).
- The model outperformed previously reported methods for TSS classification on external datasets.
Conclusions:
- The developed deep learning model shows potential for automatic, expert-free assessment of stroke onset time from imaging.
- This automated approach can improve the efficiency and accessibility of TSS determination in acute ischemic stroke.
- The model's strong generalization performance highlights its clinical applicability.
