Fig. 5: Classification uncertainty analysis.

a, b Uncertainty analysis for our synthesis-empowered classification framework for normal cognition (NC) vs. Alzheimer’s disease (AD) and static mild cognitive impairment (sMCI) vs. progressive MCI (pMCI); (c)-(d): Uncertainty analysis for the commonly used single-modal variant. a and c Plots of confusion matrix with the corresponding average uncertainty on test data. The estimated uncertainty of all the test subjects are categorized according to the confusion matrix, i.e., true positive (TP), true negative (TN), false positive (FP), and false negative (FN). The average uncertainty of each category is shown along with the ratio of covered data. b and d Plots of average accuracy (ACC) curve and covered data ratio over normalized uncertainty. We compute the average ACC of the subjects, which have the uncertainty above the given threshold. We also show the covered data ratio above this given uncertainty threshold. We denote the two-sided p-value p < 0.05 as ^*p < 0.01 as ^**p < 0.001 as and ^***p < 0.0001 as ^****.