Skip to main content
. 2023 Mar 25;90:104540. doi: 10.1016/j.ebiom.2023.104540

Fig. 2.

Fig. 2

ROC curves and radar plots for the unmatched sample. (A) ROC curves AUC for the bvFTD vs HC, AD vs HC, and bvFTD vs AD classification for the 3T ONLINE, 3T LAC, and 1.5T LAC datasets. First row: In the 3T ONLINE testing dataset, we obtained an AUC of 0.94 (p-value = 0.005 [permutation test, FDR corrected]), for the bvFTD vs HC classification, an AUC of 0.95 (p-value = 0.004 [permutation test, FDR corrected]) for the AD vs HC classification, and 0.90 (p-value = 0.007 [permutation test, FDR corrected]) for the bvFTD vs AD classification. Second row: In the 3T LAC testing dataset, we obtained an AUC of 0.90 (p-value = 0.007 [permutation test, FDR corrected]) for the bvFTD vs HC classification, an AUC of 0.88 (p-value = 0.009 [permutation test, FDR corrected]) for the AD vs HC classification, and 0.82 (p-value = 0.019 [permutation test, FDR corrected]) for the bvFTD vs AD classification. Third row: Lastly, in the 1.5T LAC testing dataset, we obtained an AUC of 0.78 (p-value = 0.031 [permutation test, FDR corrected]) for the bvFTD vs HC classification, an AUC of 0.79 (p-value = 0.027 [permutation test, FDR corrected]) for the AD vs HC classification, and 0.77 (p-value = 0.035 [permutation test, FDR corrected]) for the bvFTD vs AD classification. (B) Radar plots for the accuracy, sensitivity, specificity, precision, and recall performance metrics including each binary classification results. First row 3T ONLINE, second row 3T LAC, and third row 1.5T LAC. AD: Alzheimer's disease, AUC: Area Under the Curve, bvFTD: behavioral variant of fronto-temporal dementia, HCs: healthy controls, Sens: Sensitivity, Spec: Specificity. Prec: Precision. Sample size: (AD, n = 1000), (bvFTD, n = 1000), and (HC, n = 1000).