MIT’s Data to AI Lab (DAI Lab) has developed a software system, Cardea, to marshal reams of hospital data through machine learning models to extract valuable insights. The framework is open-sourced and uses generalisable techniques to encourage teamwork and drive innovation. Cardea could come in handy to deal with a range of problems, from black swan events like pandemics to no-shows.
Cardea will be able to help hospitals solve “hundreds of different types of machine learning problems,” said Kalyan Veeramanchaneni, principal investigator of the DAI Lab. He is also a principal research scientist in MIT’s Laboratory for Information and Decision Systems (LIDS).
Automated Machine Learning or AutoML is a set of methods and processes to make machine learning available for non-experts. AutoML can either optimise an existing model or scope out the best models for specific datasets. It automates and accelerates machine learning model development.
Cardea is an AutoML system. To build Cardea, the MIT researchers put together several machine learning tools in healthcare to make a powerful reference for hospital problem-solvers, said Sarah Alnegheimish, a graduate student in LIDS.
Cardea is designed to work with Fast Healthcare Interoperability Resources (FHIR), the standard-bearer in electronic healthcare records.
How does it work
Cardea takes the users through a pipeline with multiple choices and safeguards at each step. The data assembler first feeds the information. Cardea’s data auditor finds discrepancies in the data. Then, the system asks what queries the user has or the specific problem to be addressed. The users will then be presented with different ML models, and Cardea explores the dataset and models to learn patterns to make predictions.
As of now, Cardea can help with four types of resource-allocation questions. As Cardea improves, it will be able to “solve any prediction problem within the health care domain,” said Alnegheimish.
The team presented Cardea at the IEEE International Conference on Data Science and Advanced Analytics in October 2020. In the efficacy test, the system outperformed 90 percent of users on a data science platform.
The next step in Cardea’s evolution is the model audit. The system will empower the users to decide whether to accept a particular model’s results.
Cardea was released to the public as an open-source platform earlier this year. Developers around the world are welcome to improve on the model.
The team has plans to add more data visualizers and explanations to make Cardea more accessible to non-experts.
Healthcare & AI
The Healthcare sector is immensely benefitting from the advances in AI and machine learning. Below we look at a few companies that offer transformative digital health technologies leveraging AI, data analytics and machine learning.
Canada-based BlueDot’s cloud platform delivers outbreak risk awareness in near real-time. with its cloud platform. BlueDot Insights claimed to have identified the COVID-19 outbreak in Wuhan before the US Centers for Disease Control and Prevention and the World Health Organisation. The company’s geographic information system combines over 100 datasets to help users run complex risk assessments. BlueDot provides specific information and intelligence to deal with an infectious disease threat.
Google Health recently launched an AI tool to detect breast cancer after parsing datasets from mammogram scans. Google has teamed up with Deep Mind’s health team to accelerate the AI adoption to tackle blindness in diabetics, cancer detection, postpartum depression, and more.
Philadelphia based-Oncora Medical has an intuitive visualisation software to connect real-world data generated from various clinical sources. The platform leverages analytics to identify research questions, design more efficient clinical trials, establish patient cohorts, and test trial accrual.
Innovaccer has become the first healthcare Unicorn in India. Check out their journey here.