ACR/SIIM/STR sponsor machine learning challenge for pneumothorax detection

The American College of Radiology (ACR), the Society for Imaging Informatics in Medicine (SIIM), the Society of Thoracic Radiology (STR), and MD.ai are jointly sponsoring a machine learning challenge on pneumothorax detection and localization. The Pneumothorax Detection Challenge will be launched at SIIM’s 2019 Annual Meeting being held June 26-28 in Aurora, CO.

Participants in the challenge will develop high quality pneumothorax detection algorithms to prioritize patients for expedited review and treatment and to promote the development of clinically relevant use cases for artificial intelligence (AI). They will use augmented annotations on the public chest radiograph dataset from the National Institutes of Health (NIH) created by radiologists from SIIM and STR. These annotations were developed using a commercial web-based tool from MD.ai, and follow the ACR Data Science Institute’s structured artificial intelligence (AI) use case for pneumothorax detection. Standards-based healthcare API's (FHIR and DICOMweb) will be used to reduce the interoperability barriers to clinical implementation post-competition.

“This Kaggle competition will result in open source algorithms to help solve a serious healthcare problem that can lead to death if not identified and treated quickly,” said Bibb Allen Jr., MD, chief medical officer of the ACR Data Science Institute in a joint press release. “By co-hosting this challenge to engage data scientists in solving real clinical problems defined in a structured AI use case, we are bringing together the radiology and technical communities to generate new healthcare solutions and improve patient care.”

Winning teams will be announced at the 2019 SIIM Conference on Machine Intelligence in Medical Imaging being held September 22-23 in Austin, TX.

A similar Kaggle challenge, the Pneumonia Detection Challenge, is sponsored by the Radiological Society of North America (RSNA). The 2018 challenge winners, announced at the 2018 RSNA annual meeting, were Alexandre Cadrin-Chênevert, MD, a practicing radiologist in Lanaudière, Quebec, Canada, and Ian Pan, a third-year medical student at Brown University in Providence, RI. A total of 346 teams participated in this challenge designed to foster the development of AI algorithms.