AI image challenge

RSNA organizes data challenges to spur the creation of artificial intelligence (AI) tools for radiology. Data challenges engage the radiology community to develop datasets useful for training AI systems to perform clinically relevant tasks. Researchers then compete to create applications that perform defined tasks according to specified performance measures. The goal of each challenge is to explore and demonstrate the ways AI can benefit radiology and improve patient care.

These AI data challenges are organized by the RSNA Radiology Informatics Committee. Please direct questions about the AI data challenge program to informatics@rsna.org.

RSNA Pneumonia Detection Challenge (2018)

In 2018 RSNA organized a challenge to detect pneumonia, one of the leading causes of mortality worldwide.

Download datasets

Please note: These are very large files. We recommend you have sufficient internet bandwidth and storage available before downloading the datasets.

About the challenge

We worked with colleagues at the Society for Thoracic Radiology and MD.ai to label pneumonia cases found in the database of chest x-rays made public by the National Institutes of Health (NIH).

In the challenge, we invited teams of data scientists and radiologists to develop algorithms to identify and localize pneumonia. Kaggle, a subsidiary of Google, provided a data-sharing platform for the challenge. Kaggle also identified the challenge as socially beneficial and contributed $30,000 in prize money.

Response to the Pneumonia Detection Challenge was overwhelming, with over 1,400 teams participating in the training phase. The 10 top entries in the test phase were recognized at an event in the AI Showcase at RSNA’s 2018 annual meeting.

Further reading

RSNA Pediatric Bone Age Challenge (2017)

RSNA conducted a challenge to assess bone age from pediatric hand radiographs, a routine task that determines an important developmental indicator.

The challenge used a dataset developed by Stanford University and the University of Colorado, which was annotated by multiple expert observers.

Over 250 participants, comprising radiologists, technology companies, computer scientists, engineers and other medical specialists, competed in the challenge. Participants worked in 37 teams to submit the outcomes of their algorithms. Teams with the most accurate predictions were recognized at RSNA 2017.

Further reading