Application of multi-modal data fusion based on deep learning in diagnosis of depression

Depression is a frequent mental condition requiring precise diagnosis in its early onset. Traditional methods are less than accurate and occur late. Following these deficits, this investigates the multi-modal data fusion and Deep Learning (DL) with the purpose of enhancing accuracy for diagnosis. A new DL model, Dynamic Dolphin Echolocation-tuned Effective Temporal Convolutional Networks (DDE-ETCN), is utilized for depression diagnosis. Different sources of data, such as physiological signals (EEG, heart rate), behavioral indicators (facial expressions), and biometric data (activity levels), are fused. Data preprocessing includes wavelet transformation and normalization of biometric and physiological data, and median filtering of behavioral data to provide smooth inputs. Feature extraction is performed through Fast Fourier Transform (FFT) to obtain frequency-domain features of depression indicators. Feature-level fusion is a good fusion of all data sources, which improves the model's performance. The DDE tuning mechanism improves temporal convolution layers to improve the model's ability in detecting sequential changes. The proposed DDE-ETCN model highly improves depression diagnosis when it is developed in Python. The model attains an RMSE of 3.59 and an MAE of 3.09. It has 98.72% accuracy, 98.13% precision, 97.65% F1-score, and 97.81% recall, outperforming conventional diagnostic models and other deep learning-based diagnostic models. The outcomes show the efficiency of the model, rendering a more objective and accurate depression diagnosis. Its higher performance justifies its potential for practical use, providing enhanced accuracy and reliability compared to traditional approaches. This innovation emphasizes the necessity of incorporating deep learning for enhanced mental health evaluations.