Skip to main content
. 2024 Jul 10;15:1418060. doi: 10.3389/fneur.2024.1418060

Table 2.

Summary table of the application of artificial intelligence in the treatment of ischemic stroke.

Author and year Imaging modality Dataset size Methodology Multi center External validation Clinical application Evaluation metrics Conclusion
Hyunna et al. (32) MRI 355 SVM, RF, LR No No Identify of Onset Time RF: TPR = 75.8%
FPR = 82.6%
ML can predict the onset time, aiding doctors in choosing treatment plans.
Liang et al. (33) MRI 433 LR, SVM No No Identify of Onset Time LR: AUC = 0.91
SVM: AUC = 0.90
Integrating different images can also improve prediction accuracy.
Zhu et al. (34) MRI 268 EfficientNet-B0U No No Identify of Onset Time ACC = 80.5%
TPR = 76.9%
FPR = 84.0%
DL shows higher accuracy than ML.
Wu et al. (38) MRI 2,770 DeepMedic Yes Yes Classification of Subtypes Precision = 0.83 AI based on imaging data can accurately classify stroke subtypes.
Zhang et al. (37) MRI 174 RF + Radiomic No No Classification of Subtypes AUC = 0.936 Combining ML with radiomics can diagnose stroke subtypes.

TPR, Sensitivity; AUC, Area Under Curve; FPR, Specificity; SVM, Support Vector Machine; LR, logistic regression; RF, Random Forest; ML, Machine Learning; DL, Deep Learning.