Published Paper Details:

Depression is a globally recognized mental health disorder that affects millions of individuals, often going un- diagnosed due to limitations in traditional screening methods. The growing prevalence of social media platforms has created an alternative space where users express their emotions and thoughts, providing valuable behavioral data for mental health analysis. Most existing AI-based depression detection systems focus on single modalities--primarily text--while overlooking non-verbal cues such as facial expressions that are equally critical in identifying depressive states. This paper presents a comprehensive review of recent research in multimodal AI systems that integrate text, image data to improve the accuracy and robustness of depression detection. We explore the use of Large Language Models (LLMs) like BERT and GPT-4 for textual analysis, Convolutional Neural Networks (CNNs) for facial emotion recognitions. A hybrid framework is proposed for fusing these modalities, allowing the system to assess depression risk with high reliability. The review also identifies existing research gaps, such as the need for diverse datasets, real-time implementation challenges, and ethical concerns related to user privacy. Ultimately, this study highlights the significant potential of multimodal AI in supporting early detection and intervention in mental health care, while emphasizing the importance of responsible AI deployment that complements professional psychological assessment.