Stanford is looking to democratize research on artificial intelligence and medicine by releasing the world’s largest free repository of AI-ready annotated medical imaging datasets. This will allow people from all over the world to access specific data that they need for their respective projects, which could lead to potentially life-saving breakthroughs in these fields.
The use of artificial intelligence in medicine is becoming increasingly pervasive. From analyzing tumors to detecting a person’s pumping heart, AI looks like it will have an important role for the near future.
The AI-powered devices, which can rival the accuracy of human doctors in diagnosing diseases and illnesses, have been making strides as well. These systems not only spot a likely tumor or bone fracture but also predict the course of an illness with some reliability for recommendations on what to do next. However, these systems require expensive datasets that are created by humans who annotate images meticulously before handing them over to compute power, so they’re rather costly either way you look at it given their price tags–millions even if your data is purchased from others or millions more if one has created their own dataset painstakingly through careful annotation of images such as CT scans and x-rays along with MRI’s etcetera depending upon how advanced each system needs be.
In two short years, AIMI (Center for Artificial Intelligence in Medicine and Imaging) has amassed an extensive collection of annotated images from the Stanford University Medical Center and other sources. This wealth of data is available now to anyone who wants it for free – all you have to do is download it and train your AI model with little effort on their end!
In a recent partnership with Microsoft’s AI for Health program, AIMI has developed an automated platform capable of hosting and organizing scores of additional images from institutions worldwide. Part of this idea is to create an open and global repository where research can be shared on modeling programs to refine different models easier by identifying differences between population groups.
With many different ways of gathering data, the idea is to create a whole ecosystem for AI medical research. This includes more than just analyzing images! With this new system, people will be able to explore other uses outside pixel-based analysis, including clinical cases and companion multimodal datasets.
With the release of these datasets, it will be easier for researchers to spot hidden biases in their data or algorithms. Studies have shown that some AI models are more accurate than others, and with this diverse dataset from many different communities, they can detect those issues.
AI Platform: https://stanfordaimi.azurewebsites.net/