Medical training in Asia could also be headed to a future the place establishments are usually not solely coaching college students but additionally coaching AI fashions.
Within the HIMSS24 APAC session, “Implementing AI in Healthcare,” Dr Hyung-Chul Lee, deputy CIO of Seoul Nationwide College Hospital (SNUH), shared how their hospital develops, validates, and implements AI applied sciences.
“For greater than 100 years, our hospital has been educating medical college students and residents. However now we even have to coach LLMs (giant language fashions).”
When LLMs ultimately made their option to healthcare not too long ago, researchers at SNUH shortly validated their potential utility. Whereas LLMs scored excessive in one of many world’s hardest medical examinations and outperformed present predictive AI fashions in such use instances as predicting in-hospital mortality and real-time admissions, their capacity stays restricted.
“Our analysis reveals that no LLMs have but scored larger than 60 [in the Korean Medical License Exam], particularly in medical legislation.”
“Additionally, these LLMs have restricted capacity [in performing] multimodal duties. For instance, if I uploaded photographs associated to ventricular tachycardia – a situation requiring emergency resuscitation – from VitalDB (a multi-parameter very important indicators database), even the [latest] ChatGPT-4o [wrongly] solutions that it is a regular ECG and no resuscitation is required.”
Acknowledging the necessity for LLMs to get fed with extra various information, Dr Lee and his group constructed a vector database and a mannequin for fine-tuning search. “We’re at present engaged on vector embedding of all information in our hospital and loading it into the VectorDB.”
This initiative led to the event of SNUH’s in-house LLM platform to securely develop AI fashions for his or her clinicians. The LLM service, Dr Lee stated, required 40 H100 GPUs and 6 petabytes of storage.
“VectorDB might be our textbook and fine-tuning it will likely be [in] our curriculum.”
“I stay up for seeing AI fashions skilled at our hospital and [how these] will change our follow and enhance affected person outcomes.”