Severance Hospital Develops AI Emergency Activity Support Platform

"Ultimate Goal: Improving Patient Survival Rates"

A research team at Severance Hospital has developed artificial intelligence (AI) to assist with the rapid treatment of patients, starting from inside the ambulance and continuing through to the emergency room.

Hyukjae Jang, Professor of Cardiology at Severance Cardiovascular Hospital, Severance Hospital

Hyukjae Jang, Professor of Cardiology at Severance Cardiovascular Hospital, announced on the 30th that his team has completed the first phase of research and development for the “Intelligent Emergency Activity Support Platform,” a project carried out as part of the National Fire Agency’s research and development (R&D) initiative, and has produced an integrated prototype.

It is crucial to maintain the golden hour for treatment while in the ambulance before arriving at the emergency room. However, in addition to emergency procedures, it is easy to miss this critical window due to processes such as checking various vital signs and confirming available hospitals. There have also been challenges because records that need to be conveyed to emergency room staff have relied on the memory of paramedics.

In this first phase of research, the team integrated AI models to support on-site record-keeping by paramedics and the transfer of information to hospitals, focusing on facilitating fast communication between ambulances and emergency rooms.

The AI-integrated model developed by Professor Jang’s team incorporates core functions necessary for emergency transport into a single platform, including automatic creation of emergency activity logs, support for optimal transport decision-making, and transmission of on-site photos and assessment reports.

The four categories created by integrating a total of 10 types of artificial intelligence include: ▲ Emergency information conversion AI, which uses a speech recognition model specialized for emergency conversations, and ▲ Emergency situation prediction AI, which integrates models that predict patient deterioration at the emergency scene.

Paramedics who actually used the models during the first phase of research and development gave high marks for overall user convenience, improved work efficiency and response speed, and reliability. The overall satisfaction score reached 86 points, far exceeding the first-phase evaluation benchmark of 80 points. In particular, the optimal hospital recommendation feature was praised as a useful reference indicator in the field.

In the second phase, the research team plans to quantitatively verify the effects of improved response speed, reduced documentation burden, enhanced accuracy of communication between field and hospital, and system stability through real-world operation. Based on this, they will also pursue further enhancement of features by incorporating field feedback.

Professor Jang stated, “In the first phase, we integrated core functions needed for collaboration between field and hospital, and established a foundation at the level of completed development by advancing 10 types of AI models to support on-site record-keeping, judgment, and information transfer. Most importantly, our ultimate goal is to improve the efficiency of emergency activities in ambulances and ensure that records about the patient’s condition are quickly delivered to the appropriate emergency room physician, thereby increasing patient survival rates.”