On March 21st, 2024, we held a press conference on medical research achievements, inviting Director Fang Wen-Hui from the Department of Family Medicine to discuss the research team's development of a new artificial intelligence opportunistic screening model for "Application of Artificial Intelligence Transforms Osteoporosis Screening," as well as the severity of osteoporosis globally.

Osteoporosis is the second most prevalent disease globally, second only to cardiovascular diseases. Its severity affects elderly populations worldwide. According to statistics, among Taiwanese people aged 65 and above, the incidence of vertebral compression fractures is 19.8% in females and 12.5% in males. The prevalence of osteoporosis in menopausal women is about 30%. This indicates that osteoporosis is not only an asymptomatic disease but also the leading cause of fragility fractures.

Traditional systematic screening is a comprehensive screening based on known risk factors (the so-called FRAX score), including age, gender, weight, height, fracture history, parental fracture history, and history of rheumatoid arthritis/ secondary osteoporosis, use of steroid drugs, smoking habits, drinking habits, and the risk of hip fractures and osteoporotic fractures in the next 10 years can be calculated. However, the utilization rate of traditional systematic screening is very low. In the United States, even if the state provides full subsidies, only 30% of eligible older women and 4% of older men receive osteoporosis screening.

Osteoporosis is a symptomless condition and a primary cause of fragility fractures. The mortality rate of hospitalized patients with fragility fractures is approximately 20%, making osteoporosis screening a critical task. Timely intervention and treatment can effectively reduce the risk of fragility fractures by nearly 90%. Director Fang Wen-Hui stated that the research team at our hospital has developed a new opportunistic screening model using artificial intelligence, which differs from traditional systematic screening. It can more accurately target high-risk populations for screening and related research articles have been published in the Journal of Medical Systems [2022 Impact Factor: 5.3, 13/105 (11.9%) in HEALTH CARE SCIENCES & SERVICES]. The research team utilized large-scale data within the healthcare system to train artificial intelligence algorithms for assisting diagnosis. These algorithms can identify disease characteristics from patient medical data and provide feedback to the attending physician, enabling them to screen asymptomatic high-risk populations effectively. This ensures that limited medical resources are used for populations truly in need of screening, thereby implementing risk stratification in preventive medicine to achieve the goal of early detection and treatment. Chest X-ray is one of the most commonly performed examinations among patients seeking medical attention. Therefore, the research team at Tri-Service General Hospital used this opportunistic screening model to predict the bone mineral density of the lumbar spine and hip joints based on past chest X-rays of patients. Patients were then notified to return for standard bone mineral density examinations.

Through this new model, 272 out of 315 high-risk patients (86.3%) identified within Tri-Service General Hospital were confirmed to have osteoporosis upon follow-up visits, while 43 patients (13.7%) were found to have osteopenia. Ultimately, 92 patients (33.8%) underwent active drug therapy. Among them, there was a 60-year-old male patient with no known FRAX risk factors who would have been overlooked in traditional systematic screening. However, through prediction of his bone mineral density based on chest X-rays, he was identified as a high-risk population for osteoporosis, and subsequent DXA examination confirmed the severity of his osteoporosis. Similar situations occurred with a 44-year-old male patient who was successfully identified as a high-risk population for osteoporosis through chest X-ray screening and fortunately received early detection and treatment.

In conclusion, Tri-Service General Hospital has developed a novel opportunistic screening model using artificial intelligence algorithms, which can overcome the low screening rates of osteoporosis and allocate limited screening resources to high-risk populations truly in need. This implements risk stratification in preventive medicine, aiming for early detection and treatment.