Asan Medical Center has developed a voice-based clinical documentation system powered by a large language model.
HOW IT WORKS
The system automatically records and analyses conversations – including identifying symptoms and classifying diseases – between and among staff and patients in near real-time. It also creates patients’ medical records on the spot which are automatically saved to AMIS 3.0, AMC’s EMR system.
It works in many medical settings, including emergency rooms, wards, and treatment rooms. Installed on the emergency cart’s tablet computer, it can also be used during emergency responses.
The LLM was trained with “tens of thousands of hours” of clinical voice data covering various specialty medical terminologies. To improve the scribe’s voice recognition capability, a dedicated microphone is used to filter background noise and calculate speakers’ vocal range precisely.
First developed in 2023 and piloted in orthopaedic and plastic surgery outpatient clinics, the AI scribe runs live across 16 AMC departments, such as oncology, otolaryngology, and psychiatry. The hospital plans to expand the scribe’s scope of use gradually.
WHY IT MATTERS
Touted to be a first among hospitals in South Korea, the clinical scribe was built to assist in improving medical record accuracy and support making precise treatment plans. It also helps ensure patient safety by not missing out details or information, particularly during urgent medical situations.
THE LARGER CONTEXT
The development of the generative AI-based voice scribe is part of many projects enabling digital healthcare at one of South Korea’s biggest hospitals. The foundational IT infrastructure running these projects, which include digital pathology and robot process automation, was validated at Stage 7 of the globally recognised HIMSS Infrastructure Adoption Model late last year.
AMC also applied AI in its recently launched web portal for international patients, automatically translating eight major world languages and supporting data management and verification.
ON THE RECORD
“With the AI-based voice recognition system for medical treatment, we can effectively record and store a large amount of voice-based information that often gets lost during the treatment process,” said Dr Young-Hak Kim, director of AMC Health Innovation Big Data Center.
“Accurate symptom information reflecting the voices of medical staff and patients can be the basis for improving the quality of medical care and providing personalised treatment,” he added.
