Researchers at Harvard Medical School and Massachusetts General Hospital released a platform for annotating medical images for artificial intelligence. Recent developments in medical machine learning created a need for large, multi-institutional radiological datasets. A team led by Dr. Synho Do developed a collaborative system to speed up the workflow of AI researchers and practitioners.
MarkIt allows users to upload DICOM and non-DICOM images, organize cases into projects and annotate for classification and object detection tasks. Built-in AI tools allow annotators to speed up the labeling process by showing suggestions and pre-annotating. Researchers from around the world can be invited to the project, making the annotating collaborative.
Harvard’s new AI solution also implements blockchain to decentralize the annotation process. This allows annotators to prove their contributions, and model developers to verify the consistency of the dataset. In the future, this technology might further secure machine learning models and distribute them for researchers across the globe.
MarkIt is now available for beta testing at https://markit.mgh.harvard.edu.
