The American College of Radiology (ACR) and the Society for Imaging Informatics in Medicine (SIIM) have released the results of their first joint machine-learning challenge to develop artificial intelligence (AI) algorithms for the detection of pneumothorax on x-ray.
In all, 1,475 teams participated in the SIIM-ACR pneumothorax detection and localization challenge. The competition involved developing algorithms capable of prioritizing patient chest x-rays based on their likelihood of having pneumothorax.
The radiograph dataset, made publicly available by the U.S. National Institutes of Health, was annotated by radiologists from SIIM and the Society of Thoracic Radiology using a web-based tool from imaging AI developer MD.ai. The groups used the modeling and analytics platform Kaggle to host the competition.
The top 10 teams with the most effective algorithms took home a total of $30,000 in prize money. The winning teams and their open-source algorithms are posted online at the SIIM website.
