Wavelet-Enhanced CNN for Depression Classification Based on MRI Images

Depression profoundly deteriorates the quality of life. Timely and accurate diagnosis remains challenging. While MRI shows promise in depression detection, existing methods lack the diagnostic accuracy required for clinical practice. Herein, we develop an integrated convolutional neural network combining VGG16 and EfficientNet architectures with wavelet transform to enhance depression identification from MRI scans. Prior to this, the data was also filtered using information entropy for preprocessing. By synergistically consolidating discriminative spatial and spectral features, our model achieves superior performance over state-of-the-art approaches on five diverse MRI datasets as quantified by accuracy, recall, precision, and F1 score.