Anxiety Detection on ECG Signal using Fuzzy Deep Learning

Anxiety is indicated by physiological changes which can be monitored non-invasively by utilizing affordable and wearable sensors. For clinical practitioners, these alternations facilitate an reliable and language-free assessment of anxiety-related symptoms that can supplement therapy programs. Recently, as a larger part of the population deals with anxiety-related issues, efficient and accurate detection of physiological markers of anxiety-related arousal is important to reduce the detrimental and irreversible impacts. In this paper, novel fuzzy VGG-16 neural network with custom classification layers is proposed for anxiety detection utilizing the one-channel electrocardiogram (ECG) signal from wearable sensors. The proposed method is tested on a publicly available ECG database for anxiety assessment with images of ECG signal having 50, 100, 150, 200, 250 beats per sample. Highest performance is obtained with images having ECG signal samples with 250 beats each i.e. 0.95, 0.94, 0.88, 0.92, 0.96 for accuracy, F1 Score, kappa score, sensitivity and specificity respectively for anxiety detection using proposed fuzzy VGG-16 model. The results of the proposed method also demonstrates that when fuzziness is added to the existing VGG-16 model, significantly improves the classification performance. This substantiates the robustness of the proposed method for automatic identification of anxiety and supporting clinicians in appropriate treatments.