Complete Recording of Voice Data in Emergency Situations… Anticipated to Improve Patient Safety and Healthcare Quality

Director Young-Hak Kim of the Digital Information Innovation Headquarters “Close Collaboration with Clinical Settings… Leading Innovation in Digital Health Technology”

▲ Professor Ji Wan Kim of the Department of Orthopedic Surgery at Asan Medical Center is consulting a patient using the AI medical voice recognition system.

1) A patient collapses from cardiac arrest in a hospital ward. While the Medical Emergency Team (MET) promptly arrives on-site to perform CPR and administer medications, the medical voice recognition system, installed on a tablet PC mounted on the emergency cart, captures and summarizes the medical team’s urgent conversations in real time. The data, including the patient’s condition and treatment details, is automatically stored in the medical information system, enabling the attending physician to utilize it during patient care.

2) As the doctor consults the patient, asking about the affected area and pain level, the entire conversation is transcribed into text in real time. While the medical voice recognition system identifies key symptoms, classifies diseases, and summarizes the conversation, the doctor concentrates on listening to the patient instead of looking at a monitor to input medical records. After the consultation, essential information for patient care is automatically stored in the electronic medical record (EMR).

A technology has been developed to automatically capture and store medical staff’s conversations as medical records, even in urgent emergency situations where every second counts, ensuring patient safety and improving healthcare quality.

Asan Medical Center has recently announced the implementation of Korea’s first AI-powered medical voice recognition system. The system records and summarizes conversations between medical staff and patients in real time across various clinical settings, including emergency rooms, hospital wards, and consultation rooms, and automatically generates medical records.

Until now, voice AI technology in the medical field has primarily been used in the form of ‘Voice EMR,’ where AI converts spoken input from medical staff into text and stores it as part of the medical record. The newly implemented medical voice recognition system at Asan Medical Center goes beyond outpatient consultations and diagnostic procedures, capturing the voices of both medical staff and patients even in emergency situations. This advancement is expected to enhance the accuracy of medical records and contribute to the development of more precise treatment plans.

In 2023, Asan Medical Center developed the medical voice recognition system and initially conducted pilot tests in the outpatient clinics of the Departments of Orthopedic Surgery and Plastic Surgery. After undergoing a validation process to assess its efficiency and accuracy, the system has now been fully implemented for use across all clinical settings.

Powered by a large language model (LLM), the medical voice recognition system utilizes voice data from both medical staff and patients during consultations to perform real-time speech-to-text conversion, record key symptoms, classify diseases, and summarize conversations. Additionally, the system is integrated with the medical information system of Asan Medical Center (AMIS 3.0), allowing users to format data as needed and automatically store it in electronic medical records (EMR) and other relevant databases.

By training the AI model with department-specific medical terminology and tens of thousands of hours of clinical voice data, the system has significantly improved the accuracy of recognizing conversations between medical staff and patients. Dedicated microphones filter out background noise and irrelevant speech while precisely analyzing vocal amplitude, further enhancing speech recognition accuracy.

This system allows medical staff to focus on patient interaction rather than manually documenting medical records, ensuring that critical symptom details, which serve as the foundation for treatment plans, are accurately and comprehensively recorded. Particularly in critical emergencies, such as cardiopulmonary resuscitation (CPR), the system converts urgent medical conversations into real-time text and automatically stores them as part of the medical record. This capability is expected to enhance patient safety by allowing for future review and analysis.

Currently, the medical voice recognition system is in use across 16 departments, including Oncology, Otolaryngology-Head and Neck Surgery, and Psychiatry, as well as emergency rooms and orthopedic wards. Asan Medical Center plans to gradually expand its use based on ongoing monitoring.

Professor Young-Hak Kim, Director for the Digital Information Innovation Headquarters at Asan Medical Center, stated, "The AI-powered medical voice recognition system enables the effective recording and storage of vast amounts of voice data that would otherwise be lost during consultations. By incorporating not only the voices of medical staff but also those of patients, the system enhances the accuracy of symptom documentation, ultimately improving the quality of healthcare and laying the foundation for personalized treatment."

He added, “We will continue to collaborate closely with clinical teams to optimize emerging digital technologies, including AI, for practical use in medical settings.”

Asan Medical Center is driving digital innovation across various fields, including robotic process automation (RPA), digital pathology system, mobile personal health record service, and precision medicine system, to establish a future healthcare environment based on digital healthcare.

At the end of last year, Asan Medical Center received the highest Level 7 certification in INFRAM (Infrastructure Adoption Model), a global standard for evaluating the digital infrastructure of healthcare institutions, administered by the Healthcare Information and Management Systems Society (HIMSS).