Deep learning is a sub-discipline of artificial intelligence that uses artificial neural networks, a machine learning technique, to extract patterns and make predictions from large datasets. In recent years, it has achieved rapid development and is widely used in numerous disciplines with fruitful results. Learning valuable information from complex, high-dimensional, and heterogeneous biomedical data is a key challenge in transforming healthcare. In this review, we provide an overview of emerging deep-learning techniques, COVID-19 research involving deep learning, and concrete examples of deep-learning methods in COVID-19 diagnosis, prognosis, and treatment management. Deep learning can process medical imaging data, laboratory test results, and other relevant data to diagnose diseases and judge disease progression and prognosis, and even recommend treatment plans and drug-use strategies to accelerate drug development and improve drug quality. Furthermore, it can help governments develop proper prevention and control measures. We also assess the current limitations and challenges of deep learning in therapy precision for COVID-19, including the lack of phenotypically abundant data and the need for more interpretable deep-learning models. Finally, we discuss how current barriers can be overcome to enable future clinical applications of deep learning.

The application of Large Language Models (LLMs) in the medical field signifies a revolutionary shift in medical informatics and patient care. LLMs trained on vast and specialized medical corpora bring precision and efficiency to medical diagnostics, treatment planning, and service delivery. This paper outlines the essential methodologies for preparing LLMs for deployment in healthcare as well as pre-training and fine-tuning processes. Innovative pre-training methods are introduced, such as the multimodal and Contrastive Language-Image Pre-training (CLIP)-based approaches that enhance the model's understanding across different data types. Besides, fine-tuning techniques such as Supervised Fine-Tuning (SFT), Instruction Fine-Tuning (IFT), and Parameter-Efficient Tuning (PET) are discussed, which can tailor the LLMs to specific medical tasks with relatively low resource requirements. Moreover, our paper delves into advanced prompting strategies, including zero/few-shot, Chain-of-Thought (CoT), and self-consistency prompting, which refine the model's capacity to handle complex medical queries, as well as Retrieval-Augmented Generation (RAG) and interactive web search technologies in improving the accuracy and reliability of model responses, which can enhance the applicability of LLMs in real-world medical practice. At last, our article depicts the applications of multimodal models in medicine and LLMs in medical texts and images. Our paper not only provides the common methods and applications of LLMs in medicine but also the future directions, ethics, and privacy in this rapidly advancing field.