Artificial intelligence/machine learning for neuroimaging to predict hemorrhagic transformation: Systematic
Richard Dagher1, Burak Berksu Ozkara1, Mert Karabacak2
1Department of Neuroradiology, MD Anderson Cancer Center, Houston, Texas, USA.
Summary
Artificial intelligence and machine learning models show reliable prediction of hemorrhagic transformation in acute ischemic stroke patients. Further prospective studies are recommended for enhanced clinical certainty and subgroup analysis.
Area of Science:
- Neuroimaging and artificial intelligence
- Stroke research
- Medical diagnostics
Background:
- Hemorrhagic transformation (HT) prediction in acute ischemic stroke (AIS) is critical for patient management.
- Artificial intelligence (AI) and machine learning (ML) offer potential for accurate HT prediction using neuroimaging.
- This study systematically reviews AI/ML model performance in predicting HT.
Purpose of the Study:
- To conduct a systematic review and meta-analysis.
- To evaluate the predictive performance of AI/ML models utilizing neuroimaging for HT in AIS patients.
Main Methods:
- Systematic literature search of PubMed, EMBASE, and Web of Science.
- Inclusion of studies on AIS patients with reperfusion therapy using AI/ML imaging algorithms for HT prediction.
- Quality assessment using QUADAS-2 and CAIMI; pooled sensitivity, specificity, and DOR calculated via random-effects model.
Main Results:
- Six studies including 1640 patients were identified.
- Most studies demonstrated low risk of bias and applicability concerns.
- Pooled sensitivity was 0.849, specificity 0.878, and diagnostic odds ratio (DOR) 45.598.
Conclusions:
- AI/ML models demonstrate reliable predictive capability for HT in AIS patients.
- Further prospective research is needed for subgroup analyses and to increase clinical utility.
