• Latest information
  • Timeline
  • Conditions of entry
  • Competition Awards
  • Contents of the Track
  • Contest Rules
  • 2018FHC
Latest information
MedAI Challenge获奖名单
  • MedAI “创新挑战奖”: 迪小骄(迪英加)
  • MedAI “最佳人气奖”: 心维度.家链医
  • MedAI “最具潜力奖”: 体素科技,云病理万达信息联队
MedAI Challenge决赛入围名单(排名不分先后)
  • 病理赛道: 云病理万达信息联队
  • 病理赛道: HistoAI
  • 病理赛道: 迪小骄
  • 病理赛道: AION
  • 心电图赛道: 心维度.家链医
  • 心电图赛道: 心声医疗
  • 心电图赛道: AIphaECG
  • 糖眼赛道: 体素科技
  • 糖眼赛道: 依未科技
  • 糖眼赛道: 新眼光
  • 糖眼赛道: 致远慧图
  • May 20, 2018: Launch.
  • June 13, 2018: Registration.
  • July 20, 2018: Team merger deadline.
  • July 31, 2018: Submission deadline.
  • August 10, 2018: Finalists released.
  • August 25, 2018: The final at Shanghai, China.
Conditions of entry
MedAI Challenge 人工智能卓医创新挑战赛
It is open to the world and encourages medical experts and doctors from the medical field to positively collaborate with artificial intelligence experts from universities, scientific research institutions, and startup companies from the field of artificial intelligence.
Competition Awards
1、SAIL(Shanghai AI Leader)

The SAIL (Shanghai AI Leader) Awards will be awarded by the organizing committee from the winners of each sub-seminars, the winner will be awarded.

2、Cash awards
MedAI Innovation Challenge Award150,000 yuan (1)
MedAI Most Popularity Award100,000 yuan (1)
MedAI Most Promising Award50,000 yuan (2)
3、Investment fund

Artificial intelligence investment fund of 200 million yuan will be set up by famous investment agencies, and the winning team will give priority to the support of the venture capital fund.

4、Business incubation

The winners of the competition will be supported by the incubation base, entrepreneurial mentor, resource matching, and policy counseling.

Contents of the Track
  • Pathology track
  • ECG track
  • Multiple fundus diseases

Intelligent and integrated scanning/auxiliary diagnosis system for pathology


Pathology is a subject studying changes in morphology, structure, metabolism and function of diseases, so as to reveal the etiology, pathogenesis and prognosis of diseases. In clinical medicine, pathological diagnosis is by now the gold standard to diagnose cancer. In foreign countries, pathologists are reckoned as the doctor of doctors, whereas in China, for various reasons, the lack of pathologists is still significant.

The National Health Commission requires 2 pathologists for every 100 hospital beds, but at present there are only about 10,000 registered pathologists, and the application of artificial intelligence to healthcare is an important way to fill the huge gap for pathologists.

Besides, the National Health Commission requires compulsory pathological examination for surgical operations, and the average waiting time for pathology results is about 30 min. Therefore, intelligent equipment is urgently needed to simplify the examination and accelerate the identification. By now, there is still a technological gap regarding to such equipment, and if China could first break the technical difficulties, it will lead the research and development process of the world’s pathological equipment.

Last but not the least, medical resources are unevenly distributed in China. Many central and western areas lack basic conditions for clinicopathological diagnosis, so if artificial intelligent products, equipped with the wisdom and experience of top histopathologists in Beijing, Shanghai and Guangzhou, could be used to make up the technical gap between the East and West, the national strategy of “healthy China” and “accurate poverty alleviation” could hopefully be realized.


This competition is focused on the intelligent scanning /auxiliary diagnosis system in the field of pathology, including tissue sections, intraoperative frozen sections, cytopathological smears, and immunohistochemical quantitative analysis. This competition mainly considers the accuracy and speed of AI algorithm calculation, but the usability as well as the market prospect of the hardware system are also taken into account. Aiming at “clear visualization and correct identification”, the system should integrate digital slice scanning and AI algorithms, which therefore realize integrated operation of AI and digital pathological analysis.

For example, for histopathology of certain diseases such as breast cancer or thyroid cancer, the slice scanning/reading system should, based on embedded AI algorithm, rapidly identify whether the tumor is benign or malignant. For cytopatholigical smears, the system should immediately determine whether the smear is positive or negative. For immunohistochemical sections, the system should be able to perform accurate quantitative analysis.

For this competition, some pathological photography / auxiliary diagnosis devices have already been put into the market. For example, related products from PKI has been introduced to the market 2 or 3 years ago. Also, some products are being used in hospitals or institutions, such as Peking Union Medical College Hospital, or the Clinicopathological Center of Fudan University. However, the overall market is still in the exploratory stage, which is far from large-scale clinical application. Therefore, as a new invention, AI-assisted diagnosis is of broad prospect for development.

This competition will test the accuracy and speed of AI from three aspects, i.e. cytopathology, histopathology, and immunohistochemistry. For cytopathology, we will test the accuracy and speed of algorithm; for histopathology, we will test the accuracy in determining benign and malignant tumor; and in immunohistochemistry, we will test the accuracy of quantitative analysis. Also, we will consider whether the candidates could provide an overall solution for pathological analysis.

4、Match system

This is an application contest, which will focus on mature products combining hardware devices and artificial intelligence algorithms. The areas targeted include all directions of pathological analysis, including but not limited to histopathology, cytopathology and immunohistochemistry. The participating teams use their own AI (Artificial Intelligence) diagnosis and management model to obtain optimization results in terms of the digitalized data of cytopathology, histopathology, and immunohistochemistry provided by the organizers, thereby to accurately determine the cytopathology, histopathology, and immunohistochemistry with AI.


The models will be scored by the results of algorithm model and the effects of diagnosis and treatment embedding. The results of algorithm model will account for 45% of the total score, the effects of diagnosis and treatment embedding will account for 30% of the total score, and presentation will be account for 25% of the total score. The scoring rubric is listed as below:

Review Scoring rubric Dimensions percentage
Results of algorithm model (technical review) The accuracy in determining cytopathology, histopathology, and immunohistochemistry Determining benign or malignant histopathology 15%
Determining benign or malignant cytopathology 15%
Determining benign or malignant immunohistochemistry 15%
Effects of diagnosis and treatment embedding (expert review) Time cost for calculation 15% 30%
Whether the team provides integrated soft- and hard-ware solutions 15%
Presentation (expert review) Comprehensive demonstration of the technology path, application scenario, and business mode, and so on 25%
Total 100%
6、Sample data

In the final stage of the contest, the Concord Hospital and the National Institute of Health (NIH) of the United States will provide a number of histopathologic sections that are labeled benign and malignant, as well as cytopathological data and immunohistochemical digital sections that are labeled positive/negative for testing of AI algorithms. The following figure is a sample of 5 labeled images for reference for participating teams.

downloadComputer terminal open

Name Units
Prof.Zhu Hongguang
Department of Pathology, Shanghai Medical College of Fudan University
Professor, doctoral tutor
Pathological Specialized Committee, Shanghai Branch of Chinese Medical Association
唐峰 教授
郑建明 教授
殷于磊 教授
平波 教授
王漱阳 教授
崔全才 教授

ECG diagnosis


According to the report of WHO in 2017, cardiovascular disease (CDV) has become the leading cause of death worldwide, and there are about 17.7 million people in the world die from CVD every year. The Chinese Cardiovascular Disease Report (2016) estimated that, in China, there are about 290 million people with CVD, and it is accountable for more than 40% of residents dying from disease; CVD is also the leading cause of death in China, and the mortality is higher than other diseases such as cancer or lung diseases. Annually, the number of people experiencing sudden cardiac death is as high as 500 thousand, and the main reasons include myocardial ischemia and malignant arrhythmia. Dynamic and long-term ECG monitoring is an important method used to identify cardiac ischemia, malignant arrhythmia, or other cardiovascular risk events, and it is also a key to increasing the rate of cardiovascular rescue and reducing the mortality rate of the disease.

With the aging of population, the incidence and mortality of CVD in China is still rising, especially in rural areas. This is largely owing to the huge medical differences provided in different areas. In rural areas, diagnosis and monitoring of CVD is still underdeveloped; and in big cities, inexperienced doctors and grass-roots medical institutions could hardly provide immediate and accurate ECG diagnosis, and even experienced doctors might give misdiagnosis to those special cases. Once a patient is delayed or misdiagnosed, s/he would face an increased risk of sudden cardiac death, and would suffer significant harm caused by treatment errors. For example, if “supraventricular tachycardia” on ECG is recognized as “ventricular tachycardia”, this will not only lead to misdiagnosis, but also falsely change the treatment plan from radiofrequency ablation to cardioversion defibrillator implantation. By now, artificial intelligence (AI) has been widely used in cardiovascular diseases and other medical fields, and it, in combination with the internet of things, might hopefully elevate the overall level of cardiovascular disease monitoring (myocardial ischemia and malignant arrhythmia-ECG) by improving ECG diagnosis in remote areas, and therefore realize early identification of cardiovascular risk events, and reduce the death rate of patients.

The National Health Committee proposed to develop grassroots medical care as well as to improve management of chronic diseases, and CVD is at the center of chronic diseases. Focusing on monitoring of cardiovascular risk event, this competition takes intelligent ECG diagnosis, particularly rapid recognition and diagnosis of cardiovascular risk events as the topic, and aims to explore the application in AI in ECG diagnosis, and open up a new milestone for AI in cardiovascular scientific research.

3、Contest contents

The participants will be required to use their own ECG AI diagnostic model to accurately calculate the basic ECG parameters (heart rate, PR interval, QRS duration, QT interval, QTc), so as to achieve accurate diagnosis and rapid identification of hazardous events of cardiovascular diseases, such as cardiac ischemic and malignant arrhythmias.

4、Match system

This application contest will focus on mobile ECG monitor + ECG AI diagnosis system. The mobile ECG monitor will be integrated with the ECG AI algorithms to achieve individualized long-range, dynamic ECG monitoring and ECG AI diagnosis, thereby to achieve early detection of myocardial ischemia and malignant arrhythmias and other hazardous events of cardiovascular diseases.

The contest will be divided into two stages. In the first stage, participants will be required to design their own software to read the ECG voltage based on the format of “digital data of the original ECG voltage”, and use some original ECG voltage data provided by the organizer or sample data of corresponding ECG atlas, so as to optimize the existing ECG AI diagnostic models.

The second stage: The optimized AI diagnostic model will be used for ECG diagnosis to accurately calculate basic electrocardiographic parameters (heart rate, P-R interval, QRS duration, Q-T interval, QTc), which will be comprehensively analyzed to achieve accurate diagnosis. This stage will also provide comprehensive exhibition in terms of the overall design, commercial pattern and technical route of application scenarios of mobile ECG monitor + ECG AI diagnostic system.

5、Rules for review

In this match, the screened projects will be reviewed in terms of ECG diagnostic techniques and comprehensive exhibition. For the technical review, a certain number of specific ECGs selected by the expert group will be used for on-site evaluation of the AI algorithm provided by the project, of which the accuracy and speed of calculation of the basic ECG parameters account for 20% of the total score, and the ECG diagnostic accuracy accounts for 20% of the total score. For the expert review, the comprehensive exhibition of the overall design, commercial pattern and technical route of the application scenarios of mobile ECG monitor + ECG AI diagnostic system accounts for 60% of the total score. The details of the rules for review are listed in the following table.

Mode of review Review content Percentage
Technique review Accuracy and speed of the calculation of the basic ECG parameters (heart rate, P-R interval, QRS duration, Q-T interval, QTc) 20%
ECG diagnostic accuracy 20%
Expert review Expert group evaluate and score the comprehensive exhibition of the overall design, commercial pattern and technical route of the application scenarios of mobile ECG monitor + ECG AI diagnostic system. 60%
Total 100%
6、Sample data

In the contest, we will refer to the data format of the International Electrocardiogram Database to provide a number of sample data for AI calculation, as well as data or image recognition, as shown below. The sample contains the digital format of the original ECG voltage values marked by an ECG expert, or a corresponding ECG map, as well as a corresponding expert diagnosis report.

downloadComputer terminal open

Name Units
Chen Yihan
Vice Chancellor of Tongji University
Deputy director of the Eastern Hospital Affiliated to Tongji University
Director of the Key Laboratory of the Ministry of Education of Arrhythmia
Prof.Su Yangang (review expert)
Professor of the Department of Cardiology, Zhongshan Hospital Affiliated to Fudan University
Vice chairman of the Arrhythmia Committee, Chinese Medical Doctor Association
Standing committee of the electrophysiology and pacemaker branch of Chinese Medical Association
Successor chairman of the cardiac pacemaker physiology branch of Shanghai Institute of Biomedical Engineering
Prof.Bu Jun (review expert)
Director of the Department of Cardiology, Renji Hospital Affiliated to Shanghai Jiaotong University
Changjiang Scholar of the Ministry of Education
Outstanding Youth of the State
Vice chairman of the Youth Committee of Shanghai Academy of Cardiovascular diseases
Prof.Yuan Jinqing (review expert)
Deputy director of the Department of Coronary Heart Diseases, Fuwai Hospital, Chinese Academy of Medical Science; Director of ward of the Department of Internal Medicine
Chairman of cardiovascular branch of Chinese Medical Doctor Association and deputy chairman of Specialized Committee of Thrombosis Prevention and Treatment
Standing Committee and secretary of cardiovascular branch of China Geriatrics Society
Vice Chairman of Specialized Committee of Accurate Cardiovascular Disease, China Association for the Advancement of International Medical and Health Communications
卢永昕 (评审专家)
姜萌 (评审专家)

Assisted-diagnosis and management of diabetic fundus complications


According to the International Diabetes Federation, about 425 million adults have diabetes worldwide, and 1.60 million people died of diabetes in 2016. In China, there are 114 million diabetic patients in 2017, and diabetes has become a chronic disease with the highest incidence as well as mortality in China. As a lifelong metabolic disease, diabetes has more than 100 complications, involving the eyes, heart, brain, blood vessels, and kidney, and it is believed to be the diseases with the most complications. More than that, related clinical studies also suggest diabetes to be the risk factor for glaucoma, age-related macular degeneration, and retinal vein occlusion.

For diabetes and fundus complications, early diagnosis and treatment, in combination with long-term monitoring, is an important way to prevent serious consequences such as blindness. However, since the awareness and treatment rates of diabetes are very low, fundus complications can hardly be diagnosed, and more than 50% diabetic patients could not screen for such complications in time. The main reasons include:

1.the endocrinologists:

1) Uneven distribution of medical resources made patients in rural and remote areas lack of effective diagnosis and treatment;

2) The increase of doctors is lagging behind the increase of patients. This makes it difficult to manage large numbers of patients with chronic diseases;

3) Diabetes can lead to various complications involving other departments, so endocrinologists might not diagnose independently. Taking fundus lesions for example, many unexperienced endocrinologists may fail to identify the lesion and result in misdiagnosis or missed diagnosis.

2.the diabetic patients:

1) Diabetes is a life-time disease with long treatment periods and multiple complications, and it could hardly be reversed by drug therapy. According to the American Diabetes Association, 46% of diabetic patients would develop at least one complications within 3 years after onset, and the incidence would soar to 98% within 10 years after onset, which indicates that both doctors and patients should focus on long-term monitoring in order to prevent complications;

2) The treatment of diabetes lacks effective follow-up. Doctors could not determine when to perform examinations, or to change the treatment plan for prevention of complications.

Therefore, advanced technology is needed to realize a one-step management system for diabetes and fundus complications. By introducing artificial intelligence (AI) to endocrinologists, the ability of long-term patient management could be improved, and the diagnosis rate of fundus complication could be elevated. Also, the system might realize automatic follow-up of patients at different stages, and the above problems might be solved by monitoring risk factors and changing treatment plans.

3、Contest contents

This competition concerns not only the grading and identification of simple diabetic retinal lesions, but also the assisted-diagnosis and management of all kinds of diabetic fundus complications, which is rarely seen in previous domestic or oversea competitions. This competition hopes to find an AI product that truly matches the diagnosis and treatment scenarios, and meets the essential needs of doctors. During long-term follow-up of diabetic patients, endocrinologists need to timely monitor the development of diabetic retinopathy, but they may fail to identify fundus lesions in time, so AI is needed to help diagnosis of such complications.

4、Match system

This application contest will focus on the auxiliary diagnosis and management of ocular complications of diabetes, which intends to make the AI ​​products truly adapt to realistic medical scenarios, and allow the follow-up management of chronic disease represented by diabetes to become a possible task in primary hospitals. The participating teams will use their own AI-aided diagnosis and management model to participate in the contest. In the first stage, the participating teams will be asked to use a number of desensitization data samples that are labeled by medical experts and provided by the organizers to optimize their algorithm models. In the second stage of the finals, the organizer will provide the team with the desensitization case data marked by medical experts to test their models, so as to evaluate the models in the medical scenarios. Based on this, the organizer will further evaluate the overall design, commercial pattern and technical route of the application scenarios. The data used in this contest will be collected by the Metabolic Management Center (MMC).


The models will be scored by both results of model algorithm and effects of diagnosis and treatment scene embedding. The results of model algorithm accounts for 40% of the total points; the effects of diagnosis and treatment scene embedding accounts for 30% of the total points, and presentation accounts for 30% of the total points.

Detailed scoring rubric is listed in below:

Review Scoring rubric Dimension Dimension
Results of model algorithm Grading and assisted- diagnosis of fundus complications Diabetic retinopathy grading and lesion identification 15% 40%
Identification of age-related macular degeneration 10%
Identification of glaucoma 5%
Identification of retinal vein occlusion 5%
Identification of other complications 5%
Effects of diagnosis and treatment scene embedding Time cost of algorithm 10% 30%
Rationality of process embedding 20%
Presentation Comprehensive demonstration of the technology path, application scenario, and business mode, and so on 30%
Total 100%

The percentage for diabetic retinopathy grading and lesion identification (15%) is further divided as:

Diabetic retinopathy grading and lesion identification 15% Dimension Output requirement percentage Note
Grading 5 grades 5%
Microaneurysm Present / not present 1%
Bleeding point No / < 20 / >20 2%
Hard exudate No / outside 2 disk diameters from the macular center / between 1-2 disc diameters from the macular center / within 1 disc diameter from the macular center 3% The distance of hard exudate from the macular center is critical for determining the degree of macular edema
Cotton-wool patches Present / not present 1%
Neovascularization or proliferating membranes Present / not present 1% Since neovascularization is often hidden behind proliferating membranes, the two are combined as one lesion.
Preretinal or intravitreous hemorrhage Present / not present 1%
Laser spot Present / not present 1% Laser spot is the scar left after fundus treatment. It is very important for patient management.
6、Sample data

The desensitization case data labeled by medical experts used in the contest will be collected from the MMC. The data samples are illustrated as follows.

downloadComputer terminal open

Name Units
Contest Rules
1、Entry Rules

1) The team should be composed of at least 1 person, and a captain must be appointed. One person can only participate in 1 team at the same track.

2) Each person can participate in multiple track races at the same time.

3) Ensure that the registration information of the participating team is true and accurate, otherwise, the qualification and winning rights will be disqualified.

4) No limitation for nationality and age.

2、Data Agreement

1) Contestants must hold and use data only for competition purposes, including but not limited to reading and learning from data, and analyzation, unless otherwise specified in the terms of the competition website. Contestants must agree to take appropriate measures to prevent those who do not formally agree to these contest rules from gaining access to the data and agree not to transmit, reproduce, publish, redistribute or otherwise provide data to any party not participating in the contest.

3、Code sharing

Contestants are not allowed to privately share data-related or data-based development source code or executable code. Any such sharing violates these competition rules and may result in disqualification.

4、Submission of preliminary contest Report

downloadComputer terminal open

2 0 1 8 未 来 医 疗 科 技 大 会
2018 Science and Technology Conference on Future Healthcare
Combined with the MedAI Challenge, this conference contains three parallel meetings, which are named 2018 Frontier Forum on Artificial Intelligence, 2018 Frontier Symposium on Brain Science and Brain-inspired Intelligence and 2018 Symposium on Digital Pathology and Clinical Application, respectively.
  • 2018 Frontier Forum on Artificial Intelligence

    -Market Analysis of Artificial Intelligence in China

    -Artificial Intelligence

    -Medical Application of Artificial Intelligence

  • 2018 Frontier Symposium on Brain Science and Brain-inspired Intelligence

    -Basic Research of Brain Cognitive Function

    -Brain-Inspired Artificial Intelligence

    -Diagnosis and Intervention of Brain Diseases

  • 2018 Symposium on Digital Pathology and Clinical Application

    -Digital Pathology

    -Clinical Application of Digital Pathology

    -Developments of Digital Pathology and Trends of Diagnosis in the Future

  • 2018 MedAI Challenge

    -Working with multiple partners, this challenge aims to find excellent projects and teams, to provide free support on resource integration and innovation.

Medical Assessors
AI Assessors
  • Chen Yihan
    Member of Chinese Academy of Sciences,Vice president of Tongji university
  • Ning Guang
    Member of Chinese Academy of Engineering,Tenured professor of Ruijin hospital, Shanghai Jiao Tong University School of Medicine
  • Mao Junfa
    Member of the Chinese Academy of Sciences,President of the Artificial Intelligence Institute,Shanghai Jiao Tong University
  • Prof.Su Yangang
    Professor of Cardiology department of Zhongshan Hospital, Fudan University
  • Prof.Bu Jun
    Dean of Cardiology department, Renji Hospital, Shanghai Jiao Tong University School of Medicine
  • Prof.Yuan Jinqing
    Deputy director of National Center for Cardiovascular Center and Director of the first Department of Internal Medicine, Fu Wai Hospital, Chinese Academy of Medical Sciences
  • Prof.Zhu Hongguang
    Pathology Department of Shanghai Medical College of Fudan University
  • 唐峰
  • 郑建明 教授
  • 殷于磊
  • 平波 教授
  • 王漱阳 教授
  • 崔全才 教授
  • 卢永昕 教授
  • 姜萌
  • 高鑫
  • 林浩添
  • 赵明威
  • 袁源智
  • Prof.Yang Xiaokang
    Executive vice-president of Artificial Intelligence Institute, Shanghai Jiao Tong University
  • Prof.Shen Dinggang
    Joint CEO of United Imaging
  • Wei Qing
    Microsoft CTO
  • 张堃
  • 金熙
  • 张静