A robust depression detection framework using Goat Paper Publishing Optimization based Mobile Maxout Forward Taylor Network from multi-modal data

Depression is a widespread and multifaceted mental health condition that profoundly affects the emotional and physical health of adults across the globe. Conventional diagnostic approaches typically depend on self-assessment questionnaires and clinical interviews, which can be subjective, time-intensive, and susceptible to underreporting due to factors like social stigma or limited self-awareness. These models frequently suffer from generalization across various populations, are sensitive to noise and missing data, and lack the contextual understanding necessary for accurate detection. Thus, this research introduces a novel framework called Goat Paper Publishing Optimization enabled Mobile Maxout Forward Taylor Network (GPPO_MMFTN) for detecting depression in adults using multimodal data. Initially, the Magnetic Resonance Imaging (MRI) image is preprocessed using the Kuwahara filter. A comprehensive set of features is extracted in the feature extraction step. In parallel, the speech signal is preprocessed employing a Gaussian filter and then, feature extraction is done to obtain the most significant and discriminative features from the preprocessed signal. Moreover, Depression detection is performed by Mobile Maxout Forward Taylor Network (MMFTN), and it is trained with Goat Paper Publishing Optimization (GPPO). MMFTN is the combination of MobileNet and Deep Maxout Network (DMN) with Taylor concept. The GPPO is the hybridization of the Goat Optimization Algorithm (GOA) with Paper Publishing–based Optimization (PPBO). Finally, the outputs from both modalities are fused using the chord distance. The evaluation results indicate that GPPO_MMFTN has acquired an accuracy of 96.323%, sensitivity of 95.399%, specificity of 97.202%, F1-score of 95.213% and precision of 95.148%.