Facebook's artificial intelligence (AI) laboratory, working with the New York University (NYU) School of Medicine, generated reliable magnetic resonance imaging (MRI) scans with less data.
Using a dataset of 10,000 MRI exams from NYU, Facebook trained an algorithm via deep learning to identify the composition of the human body, feeding this knowledge into an MRI scanner's software so the AI can generate a portion of the image, saving time.
NYU School of Medicine researchers say as fastMRI moves forward, they will open-source the image dataset and Facebook will share AI models, baselines, and evaluation metrics.
The researchers' ultimate goal is to boost the speed of scans 10-fold via machine learning, and eventually to expand to other medical imaging areas.
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA