The field of cardiovascular disease diagnosis is moving towards more accurate and robust methods, leveraging advances in deep learning and multi-modal approaches. Researchers are exploring innovative ways to integrate multi-granularity patching, hierarchical residual learning, and self-attention mechanisms to improve the analysis of electrocardiogram (ECG) signals. Another trend is the use of multi-task learning frameworks to simultaneously model the relationships between respiratory sounds, disease manifestations, and patient metadata attributes, leading to improved diagnosis performance. Additionally, there is a growing interest in developing non-invasive methods for assessing pulmonary hypertension progression using multi-view, multi-modal echocardiography. Noteworthy papers include: Cardioformer, which achieves state-of-the-art performance on ECG classification tasks with its novel hybrid model. AI-Enabled Accurate Non-Invasive Assessment of Pulmonary Hypertension Progression via Multi-Modal Echocardiography, which proposes a multi-view, multi-modal vision-language model to accurately assess pulmonary hypertension progression. Tri-MTL, which investigates the integration of multitask learning with cutting-edge deep learning architectures to enhance respiratory sound classification and disease diagnosis.