Medical Report Generation

Medical Report Generation (MRG) is a task focused on training AI to automatically generate professional reports based on input medical imaging data. The aim of this task is to assist clinical doctors in making faster and more accurate diagnostic decisions, which can be time-consuming and prone to errors, even for experienced physicians. Although deep neural networks and Transformer-based architectures are currently the most popular methods, pre-trained models often exhibit performance degradation in the medical domain, primarily due to significant differences between specific domain language datasets and the vast datasets available on the internet, as well as the imbalance in medical datasets during the fine-tuning phase. Recently, multimodal learning and contrastive learning have shown potential in this field, but they still face numerous challenges that require further research.