Original advertisement on iThome

In response to the rapid population growth in Taiwan, the adoption of low-dose CT screening will help achieve disease prevention. Taipei Medical University has developed the Deep Lung Screening Assistance System, which provides screening services for lung, heart, and bone diseases, significantly reducing the workload in cargo operations. In the future, after obtaining certification from Taiwan’s Food and Drug Administration and the U.S. Food and Drug Administration, the system is expected to be implemented in hospitals.

According to a research report by the National Development Council, Taiwan has officially entered an aged society and is expected to become a super-aged society by 2025. To address the rapid increase in the elderly population and enhance disease prevention, the government will gradually and conditionally allow assistance for low-dose CT screening. In the future, hospitals will inevitably need to allocate more personnel to health examination centers and other departments, significantly increasing the workload for radiologic technologists and diagnostic physicians.

To tackle this challenge, the Big Data Center team at Taipei Medical University has developed the Deep Lung Screening Assistance System for low-dose CT scans. This system focuses primarily on chest CT imaging and provides screening services for lung, heart, and bone diseases.

Professor Chung-Hsi Lee of the Graduate Institute of Medical and Biotech Law at Taipei Medical University stated that in the development process of the AI-assisted lung nodule CT imaging diagnostic system, 6,000 comprehensive lung CT images from Taipei Medical University physicians were utilized. Additionally, heterogeneous data from the National Health Insurance Administration was incorporated to optimize the model, integrating deep learning with radiographic atlas algorithms.

Currently, the system can identify lung nodule lesions within 20 seconds, assist in determining the nature and width of the nodules, and automatically generate standardized international clinical reports. With an accuracy rate of over 95%, physicians only need approximately five minutes for final verification to complete the diagnosis.

Dr. Yen-Ting Chen, an attending physician in the Department of Medical Imaging at Shuang Ho Hospital, stated that for the detection of cerebral small vessel disease, an AI algorithm was developed using 2,000 brain CT images. The system can automatically segment white matter regions and identify small vessel disease lesions, achieving accuracy rates of 90% and 98%, respectively. Future improvements will focus on refining computational methods by calculating lesion volume or incorporating additional indicators to enhance the precision of both quantitative and qualitative analyses of small cerebrovascular lesions.

Regarding cardiac imaging for lesion identification and coronary artery monitoring, the Taipei Medical University team has also developed an AI-assisted diagnostic system for automatic coronary artery and vascular calcification detection. The algorithm has achieved accuracy rates of 91.3% for coronary artery detection and 87.2% for vascular calcification analysis.

Establishing an Automated De-Identification Mechanism: “Rainbow” Protects Patient Privacy

The research team at Taipei Medical University has been dedicated to the field of smart healthcare for many years and fully understands the importance of protecting patient privacy. All medical imaging data used in this project follows strict protocols, ensuring that patients are informed about how their data will be utilized and given the option to withdraw from the study.

The data collected from the three major affiliated hospitals—Taipei Medical University Hospital, Shuang-Ho Hospital, and Wan Fang Hospital—undergoes a thorough de-identification process before being used, fully complying with personal data protection laws.

Wan Fang Hospital follows this mechanism by completing the de-identification and image format conversion process before uploading the data to the National Center for High-Performance Computing’s (NCHC) medical imaging database. As for Shuang Ho Hospital, medical images are first uploaded to Taipei Medical University’s Big Data Imaging Center, where de-identification and format conversion are carried out before the data is uploaded to the NCHC’s medical imaging platform.

“The Ministry of Science and Technology not only funds the Taipei Medical University team’s research but also provides support through a dedicated task force with expertise in technical tools, legal frameworks, and data protection regulations, helping us navigate challenges related to personal data laws,” explained Professor  Chung-Hsi Lee. “With this support, our team can focus entirely on developing AI models without concerns, greatly benefiting the progress of our research.”

First Medical Imaging Data-Sharing Mechanism: Aiming to Establish a Feedback System

In the past, when Taipei Medical University Hospital conducted AI research and development, medical imaging and electronic medical records were kept within the hospital. However, in alignment with the Ministry of Science and Technology’s policy to promote the academic research and sharing of medical imaging data, Taipei Medical University Hospital is, for the first time, collaborating with other institutions to share valuable medical imaging data through the National Center for High-Performance Computing (NCHC) platform.The Taipei Medical University team recognizes that high-quality, annotated medical images play a crucial role in advancing smart healthcare. As a result, they remain optimistic about the mechanism for medical imaging data sharing.Dr. Yen-Ting Chen stated that in addition to sharing medical imaging data internationally, they also hope that other research teams will provide feedback on any challenges encountered while using the data. This input will serve as a valuable reference for the future development of the database.Professor Chung-Hsi Lee pointed out that the AI models developed by the Taipei Medical University team for lung, brain, and heart-related applications have demonstrated promising detection rates. The next step for the research team is to obtain FDA and TFDA certifications, enabling these research to be implemented in hospitals and benefit more patients.

Notably, Taipei Medical University Hospital has been dedicated to precision medicine for some time, aiming to provide the most appropriate medical services for patients with gastrointestinal diseases. Looking ahead, the team plans to evaluate the possibility of integrating medical imaging with electronic medical records to develop a precision medicine system tailored to Taiwan’s needs.