The Most Joint Research Center for AI Technology and All Vista Healthcare (AINTU) was established under guidance of the Ministry of Science and Technology (MOST) in Taiwan in 2018. AINTU mainly focuses on two research fields, AI Technology and Biomedicine, and is determined to form an international AI innovation research center to establish academic benchmarks, serve as a bridge between domestic and international professionals, promote industry-academia collaboration, cultivate AI talents, and provide solutions to social problems. Not long after AINTU is established, a Memorandum of Understanding (MOU) has been signed between the center and Stanford University, and several international forums were held in collaboration with the Artificial Intelligence Research Center of National Institute of Advanced Industrial Science and Technology (AIST AIRC) of Japan. And in 2019, with the referral from the MOST and the Ministry of Foreign Affairs (MOFA), AINTU has also served and collaborated with many important international guests to promote the development of AI technology in Taiwan. The guests including: AI delegation of UK, AI delegation of Israel, Finland Trade Center in Taiwan, Senate Committee on Health and Social Policy of Czech, Deputy Minister of Digital Economy Promotion Agency (DEPA) of Thailand, delegation of American Institute in Taiwan (AIT), delegation of Ho Chi Minh City University of Technology, President of University of Waterloo, delegation of Federal Ministry of Education and Research of Germany, delegation of Directorate General of Internal Market, Industry, Entrepreneurship and SMEs of European Commission (DG Grow), delegation of French National Research Agency (ANR), President of the committee for ethical research in digital science in France (CERNA), and Dean of the Faculty of Applied Science & Engineering of University of Toronto.
Meanwhile, our center actively visited global organizations leading in AI research, especially AI in medical and healthcare, such as Harvard Medical School, MIT, MGH/BWH, Stanford University, and Google in US, and Amii. Vector, and Mila in Canada. During 2020, several physical activities with international institutions and cross-strait interschool collaborations have been planned; however, due to the COVID-19 pandemic, most of the events were suspended.
In 2020, there were totally 31 MOST AI innovation programs executing under the AINTU. And the members including 174 professors (as principle investigators or co- principle investigators), 105 PhD students, 581 master’s students, and 11 fulltime researchers.
With the signed MOU and collaboration project agreements, AINTU has been actively bridging the industry and academia, to promote possible collaborations. And the center has also been continuously organizing educational events for students, researchers or AI practitioners, to help them learn deeper about AI technologies. In 2018 and 2019, AINTU has organized more than 10 seminars, 17 AI Meetup events, 11 workshops and refresher courses, 4 summer schools and hands-on courses. More than 3000 participants have joined the events.
With the understanding to the society’s concern on the issues of ethical, law, and regulation generated in the research process of AI technologies, in the past 3 years, AINTU has hosted 6 forums on the topics of legal issues faced during the research of technologies, and has also established the team of AI-ethical, legal, and social impact (AI-ELSI) at the beginning of foundation to provide possible solutions for the issues.
▎學術卓越、樹立標竿
Establishing Academic Benchmarks
臺大AI中心轄下計畫團隊均具有擁有雄厚研究實力,研究成果豐碩亮眼,於2020年1月至12月期間,共發表191篇國際期刊論文、13篇國內期刊論文、367篇國際研討會論文、並取得6件國外專利,1件國內專利、申請7件國內專利、1項國內商標註冊獲准及3項美國臨時專利申請,取得TFDA通過執行先導臨床試驗1件。舉辦了9場國際級競賽,包括亞洲區腦腫瘤切割競賽及急重症智慧醫療數據松比賽。同時,中心也建置Taitk Server 網站 (taitk.org),將科技部AI創新專案中的計畫衍生的資料集、工具集,以及AI模型統整釋出,三年來已累積59項,協助擴大專案成果效益。
The top-notch domestic research teams gathered within AINTU have published numerous excellent research results. From January to December in 2020, the teams have published 191 papers on international journals, 13 papers on domestic journals, and 367 papers on international conferences; obtained 6 foreign patents, a domestic patent, a domestic trademark, and a permission from TFDA to execute pilot clinical trial; applied 7 domestic patents, and 3 provisional applications of USPTO.
And the teams have also hosted 9 international competitions, including the Asia Cup Brain Tumor Segmentation Challenge and Taiwan Smart Emergency and Critical Care Data Science Conference, Workshop, and Datathon.
AINTU has built a Taitk Server (taitk.org) to aggregate 59 items of resources, including datasets, toolkits, and AI models, of the projects among MOST sponsored AI innovation program.
In 2020, AINTU has announced several significant achievements: on April 2021, Prof. Hong-Yuan Liao has published Yolov4, which is the fastest Object Detection technology on earth, and was listed the first and second place on international competitions. The team has made the source code public, which has been cited more than 300 times in the first half year, and is definitely a great contribution to the research community; Prof. Winston Hsu has also announced the explainable AI module – xCos in May, 2020. The technology not only preserves high accuracy of face recognition, but also provides explanations on the inferenced results. The technology can let users understand about how AI makes decisions, and helps to increase human’s trust in AI technologies. On the side of biomedicine, the project team of Prof. Fei-Pei Lai, co-PI of the center, has been invited by the MOST to present the “AECOPD System for Precision Medicine” on a press conference. The system can offer COPD patients precise healthcare services and remind them for early treatment by a smart wearable device, including to control their body state immediately by a long-term continuous evaluation and to predict possible seizures in future 7 days. It is now implemented in National Taiwan University Hospital (NTUH) to support clinical decisions.
▎迎向國際、接軌世界
Bridging to International Professionals
中心自成立以來,積極尋求國際研究夥伴,以提升臺灣AI技術及生技醫療領域在國際間之能見度。為達成此一目標,中心已與美國、加拿大、歐洲、亞洲多國知名學術機構建立深化之合作關係。在美、加方面的合作,中心與美國史丹佛大學於2018年簽署合作備忘錄,並派駐研究人員於史丹佛大學進行移地研究,研究課題包含運用電子病歷進行心律不整、急重症等疾病預測;於2019年與加拿大多倫多大學簽署合作備忘錄,同年10月也與美國哈佛大學醫學院關係研究機構 (Hinda and Arthur Marcus Institute for Aging Research at Hebrew Senior Life Inc.) 簽署合作備忘錄,並於2020年初,雙方就人工智慧在基因體學、生技及醫療應用、臨床研究、老年與老年病學研究等議題申請國際合作研究計畫。
▲ 中心陳信希主任擔任 FinNLP 2019 最佳論文獎頒獎人
Prof. Hsin-Hsi Chen, director of AINTU, served as presenter of Best Paper Award of FinNLP 2019
Since establishment, in order to raise the prominence of domestic AI technology and biomedicine fields on the world stage, AINTU has been actively searching for international research partners. For instance, AINTU has built partnerships with several prestigious foreign institutes in the US, Canada, Europe, and Asia.
In the US and Canada, since the fulfilment of MOU in 2018, AINTU has stationed a visiting research member in Stanford University, and has been collaborating on research projects ever since. The topics of research include using digital Electronic Medical Records for Arrhythmia and critical patient prediction. In 2019, AINTU has built partnerships with institutes including University of Toronto and the Hinda and Arthur Marcus Institute for Aging Research at Hebrew Senior Life Inc. And the partnership with Marcus Institute for Aging Research has led to application for international collaboration projects on biogenetics, biotechnology, clinical research, and gerontology in 2020.
In Asia, AINTU has built partnerships with the prestigious institutions in Japan, including the AIST AIRC, RIKEN Center for Advanced Intelligence Project (RIKEN AIP), the Research Institute of Electrical Communication of Tohoku University (RIEC), and Kyoto University. AINTU and AIST AIRC have co-organized a “FinNLP” competition in a top-tier conference, the IJCAI-PRICAI 2020. There are 17 papers accepted and published. Furthermore, AINTU and AIRC have co-organized a “FinNum2” competition in NTCIR-15 in 2020, which has enrolled by 13 teams from 7 countries and 15 institutes. AINTU have also co-organized a “SHINRA2020-ML” competition in NTCIR-15 with RIKEN AIP. “When AI Meets Human Science”, a symposium focused on interdisciplinary AI and human studies, was co-hosted with the RIEC in order to promote collaboration and communication between domestic and international practitioners in the fields of AI technology and humanities and social sciences. An online and onsite hybrid symposium called “AI & Smart Medicine for Digital Health” has also been co-hosted with Kyoto University in December 2020 to promote the research achievements and applications of the domestic medical AI field. With the dedication in the past years, AINTU has also been granted the right to host ACM SIGIR 2023, which is the first time for Taiwan to host the top-tier conference in the field of information retrieval.
▲ 中心於 2020 年 12 月偕同台大醫院及轄下 8 個生技醫療領域團隊於「醫療科技展」中展出
AINTU has participated in the “Healthcare+ Expo” in December 2020 with the teams and NTUH.
In order to achieve commercialization and promote industry-academia collaboration, AINTU has continuously been working on building partnerships with domestic and international corporations. Until 2020, AINTU has signed MOU and Non-disclosure agreement (NDA) with numerous corporations and institutes, and provide resources for project team spin-off. Here is a good example, from July 2019 to June 2020, the project team of Prof. Yu-Chuan Li was granted to execute Taiwan Reputed University Startups to Taiwan Unicorns of MOST Program (MOST TRUST-U Programs) which titled “AI – enhanced Safety of Prescription and clinical application.” The team has established a start-up named “AESOP Technology, Inc.” and released “MedGuard”, a system powered by AI engines to detect potential errors in prescriptions and provide optimal recommendations. Besides, it is notable that the license fee from the AINTU project teams reached over 60 million NTD in 2020, showing the strong value creation through technology transfer from academia to industry.
▲ 中心與中國信託簽訂產學合作計畫
AINTU has signed a collaboration project with CTBC Bank Co., Ltd.
To further promote the project teams and bridging them with the industry, AINTU has participated in the “Healthcare+ Expo” in December 2020 with the teams and NTUH. During the 4 days exhibition, more than 1,100 potential collaboration discussions took places, and this surely can help project teams’ business break-through. As one of the exhibitors, Apollo Medical Optics, Ltd., the company co-founded by Prof. Sheng-Lung Huang and develops OCT-based devices for dermatology, thought that the exhibition has brought more potential partners (representatives of hospitals or clinics, and doctors) than expected, which is very helpful for them to promote their products in the future.
▲中心偕同王偉仲教授與健保署,與 NVIDIA 合作參與全球 COVID-19 聯邦學習計畫。
AINTU has joined the EXAM initiative of NVIDIA with the National Health Insurance Administration of Taiwan and the project team of Prof. Wei-
Chung Wang.
AINTU has signed a collaboration project with CTBC Bank Co., Ltd. in 2020. The project was mainly focused on using natural language process technology to analyze dialogues and to understand customers’ intention. The technology has got CTBC’s approval and will be applied in real bank services in next year.
On the aspect of collaboration with global industry, the center has joined the EXAM (EMR CXR AI Model) initiative of NVIDIA with the project team of Prof. Wei-Chung Wang. With the help of federated learning (FL), medical practitioners can acquire prediction for the level of oxygen required by incoming SARS-CoV-2 patients. The FL model has been released on NVIDIA GPU Cloud and the result has also been submitted to an international journal. The partnership has led to a collaboration between the center, National Health Insurance Administration of Ministry of Health and Welfare in Taiwan, and NVIDIA, which is certainly a great example of an international tripartite collaboration.
▲Capstone 計畫團隊於「2020 醫療科技展」中展出
Teams of the Capstone project has exhibited in the “2020 Healthcare+ Expo”
We all know that the application field and high-quality data are critical for developing useful AI technologies. On December 2019, AINTU hosted a seminar on AI technologies in NTUH, and introduced the hospital information system, the process and pitfalls of applying AI technologies, medical image analysis, and suggestions on the steps of building the domain and how to integrate external interdisciplinary resources. AINTU has also worked closely with the Center of Intelligent Healthcare of NTUH in 2020. The centers have collaborated to co-host 2 internal Smart-Healthcare seminars, develop a procedure for NTUH’s Software as Medical Device trial, and promote field applications of AI technologies in NTUH.
On March 2020, AINTU submitted a Capstone project proposal, titled “Intelligent Emergency Treatment: The Comprehensive Strategy to Resolve Overcrowding Patient Flow in Emergency Department with Artificial Intelligence Technology,” which was approved by MOST and started to execute by the center since then. This three-year project has targeted the Department of Emergency Medicine of NTUH as the experimental field, and planned to introduce AI models to each stage of patient flow.
The innovative emergency room (ER) system powered by AI technologies can assist medical practitioners on precision ER triage, analyses of medical record, immediate identification of critical condition, clinical decision support for ED discharge, and prognosis prediction for in-hospital cardiac arrest. An intelligent edge device is also equipped in the domain to consolidate the medical systems and assist medical practitioners to take immediate action on emergent situations, to provide a better and more comfortable ER environment.
除此之外,中心也透過辦理學術活動的方式培養人才,並以不同面向推廣AI知識。首先,針對初接觸人工智慧的大學生與研究生,於每年八月舉辦「AI Summer School」人工智慧技術夏季學習營,邀請多位知名教授,進行為期3~4天的課程與實作練習。2020年為第三屆,參加人數從第一年的86人成長到225人。為使人才培育能更向下扎根,中心於2020年7月與王偉仲教授合作,針對高中生規劃一系列Medical AI School之暑期營,共計216名高中生參與本次暑期營,透過實體課程搭配線上課程,並設計兩個實際醫療影像專案,編列專屬於高中生之Medical AI教材,對高中生相關Medical AI基礎教育貢獻心力。
針對研究人員與學者,中心著力於鏈結國際AI研究單位促進交流與合作。中心與Google連續三年共同舉辦『Google AI Bootcamp創新研究營』,參與人數從2018年的150人成長到2020年的550人。每年皆邀請十餘位Google Brain及Google Research學者及本中心的AI學者就最新AI研究成果進行分享交流。全幅健康照護子中心也透過跨中心/學術單位合作(包括成大AI中心、臺大生醫電資所、臺大應用數學所、臺大醫院、中研院等),合辦多場專題演講、內訓課程、Medical AI Meetup等,在2018~2020年間已有逾2000人次參與。透過定期舉辦訓練課程,培育更多優秀的研究人員,讓臺灣在AI應用於生醫照護的成果與能量能持續發展與成長。
▲ 中心主辦科技部 AI 創新研究專案 – 跨域觀摩交流暨成果發表會
AINTU has hosted the “Exhibition for AI Innovation Research Center Program of the MOST”.
The foundation of developing scientific researches is to cultivate research talents. In 2020, the project teams within AINTU have gathered 105 domestic and foreign PhD students, 581 master’s students, 82 undergraduate students, 2 international students for the project. And to promote the communication between domestic AI research teams, the four AI centers under guidance of the MOST have been taking turns to host interdisciplinary research exhibitions in each quarter. The events were composed of poster sessions and live demo by the project teams within each center, and keynotes provided by industrial professionals. On November 2020, AINTU has hosted a research exhibition including all 72 project teams within the four AI centers. The event not only promoted the connection between domestic researchers but also let the society understand more about researches of the centers.
Besides cultivating research talents within the project teams, AINTU is also dedicating in promoting knowledge of AI and sharing new perspectives by hosting educational events. First of all, targeting the beginners, undergraduate and graduate students, AINTU has organized “AI Summer School” in August in the past 3 years, and invited prestigious AI professionals to conduct the lectures. 2020 is the third year of the summer school, and the number of participants has increased from 86 (the 1st year) to 225 (the 3rd year). Besides, in order to nurture talents and build the base, the center has collaborated with the team of Prof. Wei-Chung Wang to conduct a Medical AI School for high school students, which was composed of live and online lectures. The materials used in lectures were specially designed for high school students. And the team has gathered and compiled the data used in the event and submitted a paper to an international journal in hope of making contribution to the basic education of medical AI.
▲ 人才培育活動:Medical AI Meetup Educational Events : Medical AI Meetup
For AI researchers and faculties, AINTU has focused on creating opportunities for our teams to communicate and to explore potential collaborations with top-tier international institutes. AINTU and Google have been co-organizing the “Google AI Bootcamp” for successive 3 years. The number of participants increased from 150 (2018) to 550 (2020). The camp not only provided lectures by professionals from Google Brain, Google Research, and the AINTU, but also served as a platform for researchers to connect and collaborate. The All Vista Healthcare sub-center of AINTU (MAHC) has also co-hosted lectures, forums and Medical AI Meetups with several research centers and institutes including: MOST Artificial Intelligence Biomedical Research Center, Graduate Institute of Biomedical Electronics and Bioinformatics of NTU, The Institute of Applied Mathematical Sciences of NTU, NTUH, and Academia Sinica. It is notable that the events have attracted more than 2,000 participants in the past 3 years. By conducting the educational events, the center hopes to cultivate more and more research talents and promote the research energy and results of AI application on medical and healthcare in Taiwan.
We understand that strong AI talents are the key to competitiveness. AINTU was dedicated to attract talents worldwide and to make Taiwan an AI talent hub of the Asia-Pacific region. With the goal in mind, AINTU obtained the right to host Machine Learning Summer School (MLSS 2021 Taipei) in 2021. More than 20 renowned international scholars and professionals are invited to give lectures to thousands of students from all over the world.
而自前(2019)年年底起,COVID-19的散播持續對全球社會的各個面向造成重大的影響,而臺灣不論是政府、民間或是產學研醫各界都在各自的崗位上齊心協力抗疫,為了盡一份心力,也讓世界看見臺灣的努力,中心於2020年舉辦 “TAIWAN is Helping:全方位AI x防疫論壇” 三場系列論壇,結合國內外產官學研醫專家包含前副總統陳建仁等超過30位意見領袖參與擔任講者與談。系列影片在會後也受到熱烈迴響,顯見社會大眾對於此議題的關注,也展現本次系列活動的重要性。
Another important dedication of the AINTU is to provide solutions to social problems, and to execute research and guide communications on the issues of ethics, law and regulations of the society. Since establishment, the center has hosted forums on the topics of legal issues faced during the research of technologies. On October 2020, AINTU has co-hosted a forum with the project team of Prof. Weijing Wang, and invited 8 prestigious professionals to share their points from the aspects of the framework, law, ethics, psychiatry, and science. The speakers have also shared examples of difficulties faced in practice, and provided solutions from other countries as possible guidance.
The MAHC has also established the team of AI-ethical, legal, and social impact (AI-ELSI) at the beginning of foundation. The main mission of the team is to support the possible legal and ethical issues generated during execution of the projects within the center. The team has not only conducted several forums and lectures to improve the communication between professionals in separated fields, but has also been continuously working on related studies to provide resource for supporting the establishment of domestic regulations and management systems.
In hope of making positive impacts on the society, and let the world see Taiwan’s efforts made to combat the SARS-CoV-2 virus, the center has conducted a series of “TAIWAN is Helping: AI x Epidemic Prevention Online Forum”. The 3 forums were composed of talks and panels given by more than 30 key opinion leaders from the industry, government and academia, including Dr. Chien-Jen Chen, the former Vice President of Taiwan. The videos of the series have received many attentions and positive responses after the events, which shows the society’s concern on the issue, and the importance of conducting the event.
▲ “TAIWAN is Helping:全方位 AI x 防疫論壇 ”
“TAIWAN is Helping: AI x Epidemic Prevention Online Forum”