End-to-End Depression Recognition Based on a One-Dimensional Convolution Neural Network Model Using Two-Lead ECG Signal

Abstract

Purpose Depression is a common mental illness worldwide and has become an important public health problem. The current clinical diagnosis of depression mainly relies on the doctor’s experience and subjective diagnosis, which results in the low diagnostic efficiency and insufficient objectivity of diagnostic results. Therefore, establishing a physiological and psychological model for computer-aided diagnosis is an urgent task. In order to solve the above problems, this article uses a convolutional neural network (CNN) to identify depression based on electrocardiogram (ECG). Methods Our method uses the raw ECG signal as the input of one-dimensional CNN, and uses the automatic feature processing layer of CNN to learn and distinguish signal features without additional feature extraction and feature selection steps. In order to obtain the optimal model, ECG segments of different durations (3 s, 4 s, 5 s and 6 s) and CNNs with different layers were used for comparison. In order to obtain modeling data, the resting ECG of 37 depression patients and 37 healthy controls were collected. In the proposed network, larger convolution kernels are used to better focus on overall changes. In addition, this article focuses on the inter-patient data classification standard, where the training and test sets come from different patient data. Results Through comprehensive comparison, the 5 s ECG segment and 5-layer CNN are recommended in related applications. The proposed approach achieves high classification performance with accuracy of 93.96%, sensitivity of 89.43%, specificity of 98.49%, positive productivity of 98.34%. Conclusion The experimental results indicate that the end-to-end deep learning approach can identify depression from ECG signals, and possess high diagnostic performance. It also shows that ECG is a potential biomarker in the diagnosis of depression.

Explainable Computer-Aided Detection of Obstructive Sleep Apnea and Depression

“…The C5.0 model yielded the best classification performance in terms of accuracy at 91.0 %, showing promise for the use of similar methodology in our work [37]. VOLUME 10, 2022 Works regarding detecting depression also vary in methodology; works commonly use EEG [39], [40], ECG [41], speech audio data [42], scoring systems, such as the Depression, Anxiety, and Stress Scale questionnaire (DASS 21) [43], or magnetic resonance imaging (MRI) data to train machine learning models in their efforts to diagnose depression [44]. Davies-Bouldin Index-selected features, of which LR yielded the best performance with an accuracy of 80.00 %, a sensitivity of 54.00 %, a specificity of 93 %, and an AUC of 0.75 in diagnosing internalizing disorders (anxiety and depression) [42].…”

Automatic Diagnosis of Cardiovascular Diseases from ECG Signals using Convolutional Neural Network and Soft Computing Methods

“…• The existing automatic ECG-based heart disease detection techniques [34][35][36][37][38][39][40][41][42][43][44][45][46][47] directly perform the automatic feature extraction and classification. As ECG signals are vulnerable to different types of noises, a robust and effective ECG denoising technique is required before applying automatic feature extraction and classification.…”

“…• The lack of appropriate mechanisms to handle the 2D high-dimensional deep learning features in existing methods [34][35][36][37][38][39][40][41][42][43][44][45] leads to classification unreliability and higher training time, i.e., high computational complexity.…”

“…shows the performance improvement using LSTM with FRN features compared to CNN-2019 [41], CNN-2020 [37], CNN-2019 [42], CNN-2022 [45], and CNN-2022 [46]. The models in [41][42][43] [43][44][45][46], the 1D-CNN approaches [45] [46]and the proposed model show a significant reduction in computational complexities. We are further able to improve the performance of the proposed model compared to [45] [46]methods due to the use of appropriate mechanism for ECG denoising, dynamic segmentation, and feature optimization.…”

“…It was the first method to combine linear and nonlinear properties using ECG and HRV data. As demonstrated in [45], using raw ECG as an input to a 1D CNN, the automated feature processing layer of CNN was used to learn and recognize signal attributes without the requirement for additional approaches to feature extraction and selection methods. The authors [46] created a CAD for the automated classification of arrhythmias from raw ECG data, and they used 1D-CNN and SVM classifiers to obtain their results.…”