This year marks the 25th anniversary of the National Health Insurance (NHI) program. The program has accumulated up to 64.9 billion claims records and 2.2 billion medical images since 2018, which has become an asset for Taiwan to develop big data analysis and medical AI. In support of the SMART Gov Action Plan, the National Health Insurance Administration (NHIA) opens up data in order to maximize value-added applications. While complying with the Personal Data Protection Act, from June 2019, applicants are entitled to apply to the NHIA for accessing identified medical images under the purpose of industry-academia cooperation. Today (27th), the administration held the 25th Anniversary of the NHI Program, Conference on Utilizing NHI Data for AI Applications. Various guests were invited to speak at the conference, including Vice President Lai Ching-Te, Deputy Minister of the National Development Council Shien-Quey Kao Deputy Minister of Ministry of Science and Technology Dar-Bin Shieh and Chairman of Quanta Computer Barry Lam. The conference and exhibitions showcased successful AI applications adopted by many hospitals, which gave conference participants from the industry, government, and academic sectors an eye-opening experience and built confidence in Taiwan's capability to build a global presence through its AI development.
A total of 12 teams showcased their AI research results at today's conference. Each research project has its own exceptional features, and they are summarized as follows:
⊙ Taipei Veterans General Hospital—a comprehensive, AI-assisted diagnostic system for brain metastases to assist outpatient care: DeepMetsR-Plus is the upgraded version of an AI-assisted diagnostic system for brain metastases. It can precisely identify metastatic lesions and calculate crucial figures such as quantity, maximum diameter, and sizes of the tumors. It has been used in neuroimaging and chest medicine at Taipei Veterans General Hospital to help detect lesions of lung cancer brain metastases and improve the efficiency and quality of medical workflow.
⊙ National Taiwan University—BodyPart AI recognition for medical imaging of body parts and cross-national federated learning for COVID-O2 model: The BodyPart Smart Retrieval System was developed to create identifiable images of corresponding body parts without annotations. The system can also accurately and immediately identify body parts such as head, neck, chest, abdomen, and organs. The system was part of the research of NVIDIA, a large international company. Its AI model can predict the demand for oxygen supply for patients with COVID-19 symptoms.
⊙ National Taiwan University Hospital—PANCREASaver, an AI-assisted system diagnostic for the pancreas: PANCREASaver is equipped with the world's first AI predictive model that can automatically identify pancreatic cancer using CT images and conduct a large-scale verification through a national database, with an accuracy of up to 91.1%. The system is incorporated into the PACS (Picture Archiving and Communication System), which efficiently diagnoses pancreatic cancer.
⊙ National Taiwan University Hospital—a smart, automated contouring of metastatic brain tumors and performance evaluation and analysis system: The system built by the team is a smart consultant that can accurately develop an entire plan for stereotactic radiosurgery, automatically annotate metastatic brain tumors and endangered surrounding organs, and provide evaluation results of tumor annotation. The system provides metastatic recurrence rates after surgery based on the MRI images of the brain and also offers parameters of the distance between the tumor and its endangered surrounding organs. Information provided by the system can serve as a reference for doctors when interpreting surgical and post-surgical conditions.
⊙ National Taiwan University—building a predictive model for risks of cardiovascular disease among Taiwanese citizens by integrating NHI CT images and big data: HeaortaNetR is a fully automatic AI model developed to analyze cardio-aortic calcification and fat. It is the only AI model in the world that can categorize and calculate thoracic calcification/quantify fat, and it can achieve an accuracy of heart segmentation of up to 94.2%. This model can significantly maximize the added value of chest CT scans and achieve precision medicine.
⊙ Taipei Medical University Hospital—an AI assistive diagnostic system to verify diagnosis of the lungs using NHIA big data:
Deep-Lung: LungRads is an assistive system built to diagnose pulmonary nodules using CT images. The system is incorporated with deep learning and radiological atlas algorithms, which can screen for the locations of pulmonary nodules within 20 seconds, classify benign and malignant pulmonary nodules, and automatically produce clinical management reports that are in compliance with international standards.
⊙ Far Eastern Memorial Hospital—developing an AI model for automatic staging (T stage) for nasopharyngeal carcinoma (NPC): NPC is most common in Southeast Asia. However, the complex anatomical structure of the head and neck makes tumor interpretation difficult. The AI model is built with NMR images of the head and neck, linked to the clinical process, and integrated into the RIS (radiological information system) system. It can automatically identify tumor location, perform tumor staging, and provide real-time tumor staging results.
The NHIA Director-General Po-Chang Lee pointed out that the NHIA has a valuable database with large amounts of medical information. Today's conference showcased the preliminary results of AI applications through industrial cooperation and synergies created by industrial, governmental, and academic circles after access was opened to the NHI database. The NHIA will keep an open mind and encourage willing sectors to make the most of database applications under the precondition of complying with relevant laws and regulations. This approach facilitates national development, enhances national health and welfare, and benefits all citizens