Mayo Clinic recently unveiled the development of RadOnc-GPT, an assistant designed to analyze cancer data from Mayo Clinic hospitals. This cutting-edge technology fine-tunes the Llama 2 model from Meta, utilizing data from the hospital’s network. RadOnc-GPT is trained to perform three key functions: developing radiation treatment plans, recommending radiation modalities, and explaining diagnoses with corresponding ICD-10 codes.
The main challenge faced by the development team is sorting through patient data, as the existing database only separates ICD-10 codes clearly. Other notes such as diagnoses, patient statuses, and treatment plans are not neatly categorized, necessitating manual data extraction. The team curated treatment data from 3,282 cases, adding details like the type of radiation therapy used, and implementing training with the LoRA technique.
RadOnc-GPT is initially designed to assist in answering patient queries during follow-up visits, with plans to create a chatbot for patient service in the future.
Source: Mayo Clinic
TLDR: Mayo Clinic introduces RadOnc-GPT, an AI assistant trained on cancer treatment data to provide detailed analysis and support for patient care.
Leave a Comment