On April 7, 2021, Vingroup Big Data Institute (VinBigdata) announced the results of Chest X-ray abnormalities detection challenge with a total prize value of 50,000 USD. The competition provided the largest Vietnamese medical dataset, attracting 1,277 teams from more than 60 countries and territories worldwide.
Chest X-ray abnormalities detection challenge was organized globally by Vingroup Big Data Institute on Kaggle (a prestigious playground for the data science community) from December 31, 2020 to March 31, 2021.
After 3 months of implementation, the contest found the owners of the three leading positions and a special award for the excellent Vietnamese team out of a total of 1,277 teams.
The first prize belonged to ℳS²Ƒ team consisting of members from Argentina’s National Atomic Energy Commission, Indian medical technology company HealthGenYZ, Intel Corporation and H2O.ai (Google). They would receive 20,000 USD prize. SZI – ranked second with three members from Russia. The third was Scumed, which included five members from China and Malaysia. ℳS²Ƒ, SZI and Scumed were also the top 3 teams during process of training model on validation dataset.
The Special Prize for the best Vietnamese team was given to Chu Tan Kiet, Faculty of Computer Science, Drexel University, Philadelphia, Pennsylvania (USA). He ranked 10th and would be awarded 8,000 USD.
Chu Tan Kiet said: “The medical image processing competition organized by VinBigdata is more practical than most of the current medical imaging contests on Kaggle (*). It requires both abnormalities classification and localization for 14 groups of diseases. In addition, with special prize for the best Vietnamese team, the contest also motivates Vietnamese people to participate in learning and approaching practical medical problems.”
With the challenge of building a machine learning algorithm to detect and localize 14 common types of cardiopulmonary abnormalities, the Chest X-ray Anomaly Detection competition provided a dataset of 18,000 specific medical images of Vietnamese people. Data were collected and annotated by experienced doctors from Vinmec Hospital, 108 Military Central Hospital and Hanoi Medical University Hospital. The problem simulated the real work of radiologists at the hospitals, which combined with standardized real data sources, helped technology lovers to approach practical problems. The final results were automatically evaluated by the system based on the accuracy of the algorithm sent by participant teams, through direct comparison with the doctor’s sample results.
Dr. Nguyen Quy Ha, Head of Medical Imaging Department, VinBigdata said: “With a large-scale and quality dataset, our competition has attracted thousands of participants, more than 10% of whom are Kaggle Master or Grandmaster – and experts from leading technology companies in the world such as Google, NVIDIA, Intel, ABBYY, Oracle, Preferred Networks, etc. The top three excellent teams have taken effective approaches from which practical solutions could be developed. Especially with 103 Vietnamese teams participating, this is the global data Science competition with the highest number of Vietnamese participating teams in ever.”
At the end of the contest, a set of 18,000 X-ray images collected and labeled by reputable radiologists will continue to be shared by VinBigdata on Kaggle. The competition aims to provide domestic and foreign Data Science community opportunities to access to typical Vietnamese medical data, thus resonating intellectually to jointly develop solutions to Vietnam’s medical problems.
“Vingroup aims at creating positive changes for the Vietnamese by promoting scientific research and technological innovation. Through this contest, VinBigdata hopes to create opportunities for domestic and international technology lovers to approach real problems and real data, thereby solving medical problems. We also want to inspire other businesses to share more diverse and multi-disciplinary datasets for the community ”, Professor Vu Ha Van, Scientific Director of VinBigdata shared.
Since April 1, 2021, VinBigdata has also launched VinDr Lab, an open-source software for managing and labeling medical image data. It is to promote data sharing and the development of Data Science and AI in Viet Nam./.
Kaggle.com (developed by Google) is the world’s most famous platform in the field of Data Science and AI, with a community of more than 5 million members. Kaggle competitions are strictly assessed on the impact of the problem and the quality of data. Amazon, Facebook, Microsoft, GE, Intel… are some of the big technology companies that have held the contest on this platform. As the first organization in Vietnam to hold a contest on Kaggle, VinBigdata’s award is in the top 10 highest prize levels currently.