There are few areas in medicine that require higher precision or data than radiation oncology. RadOnc-GPT is a fine-tuned LLM built using Meta Llama 2 that has the potential to significantly improve radiotherapy decision-making.
There are few areas in medicine that require higher precision or more data than radiation oncology. A patient’s life depends on receiving the right treatment in this professional field.
Mayo Clinic’s groundbreaking RadOnc-GPT, a large language model (LLM) that leverages Meta Llama 2, has the potential to significantly improve the speed, accuracy and quality of radiotherapy decision-making, benefiting doctors and patients. The patients they serve. It is fine-tuned based on a large dataset of radiation oncology patient records at the Mayo Clinic in Arizona. Because the model is trained locally using Llama 2 running a local GPU server, patient data will not be shared outside the secure network. All studies were approved by the Institutional Review Board.
“Open source LLMs, with appropriate fine-tuning, have huge potential to revolutionize radiation oncology and other highly specialized areas of health care,” said Dr. Wei Liu, Mayo Professor of Radiation Oncology and Research Director in the Department of Medical Physics, at the Arizona clinic.
The direct clinical use case for RadOnc-GPT is patient follow-up. Liu’s team plans to develop a chat robot to answer routine questions from patients after radiotherapy, reducing the workload of nurses and clinicians and allowing them to focus on higher-priority tasks.
Future developments may include expansion to other clinical tasks, such as building models to predict patient outcomes in radiation oncology. In addition, Liu said the team is considering leveraging the recently released more advanced Llama 3 model to enhance its performance.
AI-driven tools can automate daily tasks, quickly analyze complex data sets, and identify patterns that may escape human attention, saving healthcare providers valuable time. This acceleration and efficiency allow clinicians to focus on top priorities, such as direct patient care and decision-making in complex cases.
Liu said the Mayo Clinic team has been working with a health care natural language processing team at the University of Georgia for many years to actively follow the latest developments in the industry.
They chose Llama 2 as the basic model for deriving RadOnc-GPT, which adjusted instructions for three key tasks: generating radiation treatment plans, determining the best radiation method, and providing diagnostic descriptions/diagnostic details about patients based on the International Statistical Classification of Diseases (ICD-10) code.
Compared to general LLMs, Llama 2 ‘s RadOnc-GPT has improved specificity and clinical relevance.
“This process has traditionally been very time-consuming, relies on manual analysis of large amounts of unstructured clinical data, and is susceptible to changes in human interpretation,” Liu said. “Effective tools for speech processing can significantly enhance every stage of radiotherapy and have the potential to improve treatment outcomes.”
The team performed extensive manual processing to overcome the challenges of organizing and preparing radiation oncology datasets, which involved extracting, isolating and labeling relevant information from patient records.
Impact of open source
Liu said open source of advanced artificial intelligence models allows the Mayo Clinic to use cutting-edge models directly in its research and accelerate the development process. The clinical impact has also been amplified, improving patient care. The goal of the plan is not even just to improve the accuracy of therapeutic interventions.
“It is also about cultivating an ecosystem where data security is crucial,” Liu said. “In oncology, patient data is highly sensitive, and it is especially important to ensure patient confidentiality is never compromised.”
For smaller companies and institutions, open source artificial intelligence systems are critical to democratizing innovation and promoting collective progress in medical science. Liu said that open source methods could allow small organizations with limited resources to develop their own customized models.
If you want to learn more, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
More details: https://go.fb.me/ma1vim
Oil tubing: