Abstract

Heart rate variability (HRV) quantitatively evaluates the balance of sympathetic and parasympathetic functions, which is recognized as a promising biomarker to objectively diagnose of major depression disorder (MDD). We have found that the response of HRV indices under mental task condition (random number generation) were different in patients with MDD. Therefore, we propose an objective depression screening method by HRV analysis using neural network (NN) in this paper. Input layer of NN was the HRV indices and heart rate before, during, and after mental task, whereas the output layer represents the probabilities of MDD. To evaluate the performance of NN, we repeated training the NN with leave-one-out cross-validation scheme on 44 drug-naive patients with MDD and 47 healthy control subjects. The result showed that the patients can be detected with approximately 76% in untrained testing data.