Fig. 5

Results of ten-fold cross-validation using our method and other benchmark methods on FDG features. Panel A is the working pipeline of our OT TL model. We combine linear and polynomial kernelized logistic regression classifier with different OT mapping strategies. In B, we represent the accuracy score of different OT and kernel combinations. The blue and red horizontal lines represent the average accuracy of our best model and the logistic regression model respectively. In panel C, we demonstrate the performance of two baseline methods, e.g. logistic regression and SVM, and the rMLTFL model. In D and E, we visualize the performance of TL benchmarks and Multi-kernel learning strategies. In F, we plot the AUC curve of our model across ten folds