Stanford’s AIMI library is an incredible and powerful resource for data used in the training of AI models.
An AI System Is Only as Good as Its Data
Artificial intelligence is breaking new ground in healthcare and creating a world of possibilities in which everything will be automated and integrated. In fact, AI, IoT, and robotics are allying to automate almost everything in our world. See our article on how the combination of IoT and robotics are improving many aspects of the world’s healthcare system for further information about the phenomenon.
As Lushun Jiang and his research team identified, the biggest challenge affecting the implementation of AI in almost all industries is access to good training data.
Any given AI system is only as good as its training data, which is why AI and big data are an inseparable duo. Stanford’s Center for Artificial Intelligence in Medicine & Imaging (AIMI) is working to address this obstacle for AI in healthcare.
Artificial intelligence’s early were plagued by several pitfalls brought about by AI bias. Among other issues, the algorithmic bias created unfair outcomes that prioritized one arbitrary group of users over others. This inadequacy saw many AI algorithms labeled as sexist and/or racist. Today, with programs like Stanford’s AIMI provisioning several data sets for machine learning, the age-long challenge is receiving the proper attention.
About Stanford’s AIMI
Stanford University and Microsoft created the AIMI as a one-stop destination for annotated datasets that can be used as training data for healthcare models. It has already acquired over 1 million images and is constantly adding more, making it the world’s largest database for annotated de-identified medical images. These images are provided to researchers at no cost.
By teaming up with Microsoft’s AI for Health program, the new platform will be highly annotated, accessible, and visible. It hosts and organizes scores of additional images from healthcare institutions worldwide to create an open global repository. The platform will serve as a hub for sharing research, refining models, and identifying common differences in populations.
An Example of a banner with a button
Get the book for free
What Does This Mean for AI and Healthcare?
As AI continues to expand its role as a solution for numerous areas in our society, the need for more high-quality training data grows. Learning algorithms and models are limited only by the data they are fed; this is where Stanford University’s AIMI comes into the big picture.
Big AI, IoT, AIoT, robotics, and 5G-enabled smart systems need as much information as possible in order to create real-time solutions. AIMI’s library is expected to hit the 2 million image mark in 2022. It is the most robust resource of curated, patient-deidentified, and AI-ready data, and its ramifications for AI in healthcare are unimaginable.
AIMI’s goldmine of digital riches will be open-source, which is what the AI world has been waiting for in order to break even more new ground. With an open-source database, researchers from anywhere in the world can access data for their next AI projects.
All AI medical research will be positively affected, not just medical imaging. It will allow people to explore important clinical uses for AI beyond pixel data alone, including companion multimodal data.
Since most of the data will be available to researchers at no cost, they will have opportunities to explore several niche areas, such as medical problems particular to communities that large corporations may have otherwise overlooked.
The fact that Stanford’s AI-ready medical datasets are de-identified makes them even better and more useful. The platform also aims to create a collection of standardized machine learning tools and pre-trained models that will leverage the open-source data and its common architectures.
The Key 3 Takeaways
1) More open-source data means better training models, which means better AI-based models.
2) Stanford’s AIMI addresses AI’s age-long challenge by giving access to the most extensive library of training data in the world. And it’s growing – it will boast over 2 million images by the end of 2022.
3) All of this valuable data will be provided at no cost, which means its impact potential is limitless.
