Skip to main content


Fig. 4 | BMC Geriatrics

Fig. 4

From: Deep learning based low-cost high-accuracy diagnostic framework for dementia using comprehensive neuropsychological assessment profiles

Fig. 4

Dependency on the sweeping first classification threshold. Two-stage classification performance trends as function of a sweeping threshold of deep neural networks (DNNs) with MMSE for the second-stage diagnosis with Korean Longitudinal Study on Cognitive Aging and Dementia Neuropsychological Battery. a Equal error rate (EER) curve on DNNs for MMSE. b Empirically estimated performance and cost on test dataset. When first-stage classification threshold values is 0.0362, cost is minimized without any loss on performance (f1 score)

Back to article page