The field of radiology and medical AI is witnessing significant advancements, particularly in the integration of multimodal data and the application of large language models (LLMs). Recent developments emphasize the importance of temporal information in medical image analysis, with models now capable of synthesizing data across different time points to enhance diagnostic accuracy. This shift is exemplified by the incorporation of temporal-aware multimodal large language models (MLLMs) that leverage both visual and textual data to generate more precise radiology reports. Additionally, there is a growing focus on optimizing LLMs for specific radiological tasks, with strategies emerging for effective prompting and fine-tuning to better align these models with clinical needs. The integration of expert knowledge into the optimization process of LLMs is also gaining traction, offering a more efficient and effective approach to leveraging these powerful models in healthcare settings. Notably, the fusion of early modality-specific features in multimodal architectures is proving to be a superior method for disease classification, outperforming previous models in accuracy.