Depression is a widespread mental health disorder that may remain undiagnosed by conventional clinical methods. The rapidly growing world of social media sites such as Twitter, Reddit, Facebook, Instagram, and Weibo has provided new avenues for depression detection using Machine Learning (ML) as well as Deep Learning (DL), which analyze user behavior patterns and linguistic cues for more accurate detection of depression. Many techniques have been developed for this aim over the years. Identifying relevant publications on this topic using current academic search systems is challenging due to the rapid growth of research publications, unclear or limited search terms, and the complexity of citation networks. Several review papers have been published to ease this task by summarizing the methodologies, key findings, and recommendations for future research. However, most current reviews often do not provide a clear overview of the evolution, latest techniques, and challenges. This paper aims to address that gap by providing a comprehensive review of ML and DL methodologies for detecting depression on social media. We propose a generic architecture for these systems and present a detailed analysis of methodologies and datasets used for evaluation in this field. In addition, we highlight key open research areas, providing a useful starting point for further research and development. By narrowing our focus to social media, this review contributes to advancing the understanding and application of cutting-edge methods for depression detection. While this review highlights advancements in social media-based depression detection, it excludes alternative approaches like graph-based systems and reinforcement learning, and its focus on social media may limit its applicability to other domains.

Depression is a major global health issue, and social media offers a rich source of data for early detection. However, existing multimodal approaches often fail to effectively capture the contextual relationships between text and images and tend to ignore temporal information, thereby limiting their predictive accuracy and reliability. To address these limitations, we propose ContextVecNet, a multimodal deep learning framework. It is built on a CLIP-based architecture enhanced by learnable context vectors, which are integrated into the text and image encoding branches. This design allows the model to dynamically learn and adapt to the specific visual and linguistic markers of depression present in social media data. The architecture further incorporates a cross-modal transformer with time-aware embeddings, which jointly models temporal dynamics and cross-modal interactions across user posts. Evaluated on a publicly available multimodal Twitter dataset, ContextVecNet achieves state-of-the-art performance, with an AUC of 0.9922 and an F1-score of 0.9619, significantly outperforming existing methods. An ablation study confirms that freezing the context vectors results in notable performance degradation, emphasizing their importance in depression detection.

Background: Chest X-ray (CXR) imaging is crucial for diagnosing thoracic abnormalities; however, the rising demand burdens radiologists, particularly in resource-limited settings. Method: We used large-scale, diverse public CXR datasets with noisy labels to train general-purpose deep learning models (ResNet, DenseNet, EfficientNet, and DLAD-10) for multi-label classification of thoracic conditions. Uncertainty quantification was incorporated to assess model reliability. Performance was evaluated on both internal and external validation sets, with analyses of data scale, diversity, and fine-tuning effects. Result: EfficientNet achieved the highest overall area under the receiver operating characteristic curve (0.8944) with improved sensitivity and F1-score. Moreover, as training data volume increased—particularly using multi-source datasets—both diagnostic performance and generalizability were enhanced. Although larger datasets reduced predictive uncertainty, conditions such as tuberculosis remained challenging due to limited high-quality samples. Conclusions: General-purpose deep learning models can achieve robust CXR diagnostic performance when trained on large-scale, diverse public datasets despite noisy labels. However, further targeted strategies are needed for underrepresented conditions.