NTUH unveils newly developed AI-powered telecare system

Taipei, Jan. 9 (CNA) National Taiwan University Hospital (NTUH) on Thursday presented its newly developed telecare system that uses artificial intelligence (AI) to predict the occurrence of patient-related emergency events by analyzing text conversations between patients and their live-in caregivers.

The "Intelligent Voice Emergency Prediction System of the Fourth-Generation Real-Time Synchronous Telehealth Care Platform" developed by NTUH and Chunghwa Telecom won a "21st National Innovation Award - Clinical Innovation Award" in November 2024.

At a news conference, Ho Yi-lwun (何奕倫), director of the Department of Internal Medicine and Telecare Center at NTUH, said the system converts conversations between patients and caregivers into text and can analyze keywords in conversations, such as blood pressure and other physiological data, for earlier predictions patients require emergency treatment, thereby reducing unnecessary complications in patient care.

Taking patients with acute myocardial infarction as an example, Ho said if intervention occurs within the golden six hours for treatment, there is a chance occluded blood vessels can be recanalized.

The intelligent voice emergency prediction system can accurately identify patients with emergency needs, distinguish between high risk and low risk patients, and allow those with greatest need to receive necessary treatment more quickly, he further explained.

This system will not only help reduce emergency room visit rate, but also the length of hospital stays and the risk of death, due to immediate medical responses, Ho added.

Hung Chi-sheng (洪啟盛), a doctor from NTUH's Telehealth Center, said that since its establishment in August 2009, the Telecare Center has handled 540,000 phone calls and cared for more than 7,000 patients.

This time, it cooperated with Chunghwa Telecom to introduce the telecom company's multi-language, medical staff speech recognition technology and conduct AI speech recognition model training, which can recognize speech in Mandarin Chinese, English and Taiwanese languages, Mandarin accents of migrant caregivers, medical terms, Chinese and English names of drugs and convert the speech into text.

Hung explained that through language processing and analysis of integrated nursing records, the AI system can monitor and predict the probability of patients experiencing emergency events within two weeks of analysis. The system's predictive accuracy rate is approximately 90 percent, he said.

The center has published several papers demonstrating that telecare can improve patients' blood pressure, reduce mortality, reduce emergency room visits and the number and length of hospital stays, Hung added.

Home-based telecare has become an important tool to address healthcare needs in Taiwan's aging society, particularly during the COVID-19 pandemic.