Fig. 3
From: A deep learning MRI approach outperforms other biomarkers of prodromal Alzheimer’s disease
Classifying Alzheimer’s disease in its prodromal stage. By comparing the “MCI stable” to the “MCI progression” groups, ROC curves show that the deep learning MRI (DLMRI) scores were superior in classifying prodromal Alzheimer’s disease (indicated in red). The deep learning MRI scores outperformed (left panel) CSF measures of Aβ, tau, or tau/Aβ; MRI measures of hippocampal (HC) or entorhinal cortex (EC) volume or thickness; clinical measures using the modified mental status exam (MMSE) or the retention of the Rey Auditory Verbal Learning Task (RAVLT) (left panel). In a smaller subset, the deep learning MRI scores (right panel) outperformed PET measures of amyloid using the AV45 radioligand or metabolism using fluorodeoxyglucose (FDG). Specific area under the curve (AUROC) values for each measure, and statistical probability values for each comparison, are shown in the table