Imagine a world where medical students can practice their conversational skills with virtual patients, or where chatbots can provide accurate and personalized medical advice. This is the potential of medical dialogue systems (MDS). But building these systems requires massive amounts of real-world doctor-patient conversation data, which is difficult to obtain due to privacy concerns. Researchers are tackling this challenge by creating synthetic dialogues from clinical notes. A new approach called SynDial uses a single large language model (LLM) and a clever feedback loop to generate realistic conversations between doctors and patients. Starting with clinical notes as a base, SynDial iteratively refines the generated dialogue, checking for similarity to real conversations and ensuring the dialogue accurately reflects the information in the notes. This iterative process is key to producing high-quality synthetic data. Testing SynDial on real clinical notes showed promising results. The generated dialogues excelled in accurately capturing the medical facts and extracting key information from the notes. While other methods might create more diverse conversations, SynDial prioritizes factuality and faithfulness to the original medical records. This research opens doors to new possibilities in healthcare. Synthetic dialogues can train better medical chatbots, provide realistic scenarios for medical training, and even help analyze patient data for insights. While more research is needed to scale and refine the system, SynDial offers a promising path toward more accessible and effective healthcare through AI.