ABSTRACT
In acute ischemic stroke, identifying brain tissue at high risk of infarction is important for clinical decision-making. This tissue may be identified with suitable classification methods from magnetic resonance imaging (MRI) data. The aim of the present study was to assess comparatively the performance of five popular classification methods (Adaptive Boosting (ADA), Logistic Regression (LR), Artificial Neural Networks (ANN), Random Forest (RF), and Support Vector Machine (SVM)) in identifying tissue at high risk of infarction on human voxel-based brain imaging data. The classification methods were used with eight MRI parameters including diffusion-weighted imaging (DWI) and perfusion-weighted imaging (PWI) obtained in 55 patients. Sensitivity, specificity, the area under the receiver operating curve (ROC) as well as the area under the precision-recall curve criteria were used to compare the method performances. The methods performed equally in terms of sensitivity and specificity while the results of the area under the ROC were significantly better for ADA, LR, ANN and RF. However, there was no statistically significant difference between the performances of these five classification methods regarding the area under the precision-recall curve, which was the main comparison metric.
- Abbreviations
- ADA
- Adaptative Boosting
- ADC
- Apparent diffusion coefficient
- AUCprc
- Area under the precision-recall curve
- AUCroc
- Area under the receiver operating curve
- ANN
- Artificial Neural Networks
- CBV
- Cerebral blood volume
- CBF
- Cerebral blood flow
- DWI
- Diffusion-weighted imaging
- LR
- Logistic Regression
- MRI
- Magnetic Resonance Imaging
- MTT
- Mean transit time
- PWI
- Perfusion-weighted imaging
- ROC
- Receiver operating curve
- RF
- Random Forest
- SVM
- Support Vector Machine
- TTP
- Time to peak
- TMAX
- Time-to-maximum
Footnotes
Sources of funding that require acknowledgment: I-KNOW consortium was funded by the European Commission’s Sixth Framework Programme (Grant 027294).