Pediatric Cancer LLM for Symptoms and Side-Effects Prediction with Explanation

Pediatric oncology, a medical field that deals particularly with the research of childhood cancer, which is significantly different from adult cancer. Due to its rarity and the age group it affects, there are multiple challenges that need to be addressed. The lack of communication with child patients about their disease, informing the guardians of the patients about the treatment plan in layman's language, and providing affordable consultation and therapy are just a few of them. In India, these problems are amplified due to the lack of medical awareness and the existence of taboos related to cancer among the public. This research paper proposes a novel solution to tackle these challenges by building an application, powered by LLMs and Agentic AI, combining OpenAI's Swarm framework and Mistral 7B, fine-tuned on a custom-made pediatric cancer dataset. The solution takes the domain-specific user queries, generates an accurate and concise response, and modifies the tone and language automatically to make the response child-friendly and appropriate. It helps the child patients and their guardians understand their condition, predict potential side effects based on their treatment data, and provide emotional support when needed. This application is specifically for patient care to help reduce fear and confusion among the vulnerable patients and their guardians. By integrating AI with pediatric healthcare, this study explores how technology can be used not only to inform but also to comfort and empower children during one of the most difficult journeys of their lives.