A Machine Learning Approach to Predict Acute Ischemic Stroke Thrombectomy Reperfusion using Discriminative MR Image
Haoyue Zhang1, Jennifer Polson1, Kambiz Nael2
1Department of Bioengineering, University of California, Los Angeles, Los Angeles, CA, USA.
... IEEE-EMBS International Conference on Biomedical and Health Informatics. IEEE-EMBS International Conference on Biomedical and Health Informatics
|July 11, 2022
Summary
This study developed an automated machine learning algorithm using pretreatment MRI scans to predict reperfusion quality after mechanical thrombectomy for acute ischemic stroke. The model shows promise in guiding treatment decisions by forecasting patient response to intervention.
Area of Science:
- Neurology
- Radiology
- Artificial Intelligence
Background:
- Mechanical thrombectomy (MTB) is a standard treatment for acute ischemic stroke (AIS).
- Predicting treatment success relies on pretreatment imaging to assess cerebrovascular flow.
- Automated methods are needed to leverage imaging data for treatment guidance.
Purpose of the Study:
- To develop and validate a fully automated machine learning algorithm.
- To predict the final modified thrombolysis in cerebral infarction (mTICI) score post-MTB.
- To utilize pretreatment MRI scans for automated treatment response prediction.
Main Methods:
- Computed 321 radiomics features from segmented pretreatment MRI scans of 141 AIS patients.
- Defined successful recanalization as mTICI score >= 2c.
- Evaluated various feature selection methods and classification models.
Main Results:
- The best performing model achieved an AUC of 74.42±2.52%.
- Sensitivity and specificity for predicting reperfusion quality were 75.56±4.44% and 76.75±4.55%, respectively.
- Demonstrated good prediction of reperfusion quality using pretreatment MRI.
Conclusions:
- Pretreatment MR images are informative for predicting patient response to mechanical thrombectomy.
- The developed algorithm shows potential for guiding treatment decisions in AIS.
- Further validation in larger cohorts is necessary to establish clinical utility.
