A consortium of Top US Cancer Centers has developed an AI platform that trains models from multiple institutions while protecting patients’ privacy, from several months to several months to cut the timeline for new discoveries in cancer research.
Officials Cancer AI Alliance (CAIA)Under the Fred Hutchinson Cancer Center of Seattle, this method says that this method will help researchers to detect patterns throughout the population of a single organization than their own access.
The alliance, which was announced a year ago, includes the University of Dana-Farmer Cancer Institute, Memorial Sloan Catering Cancer Center and Johns Hopkins, including Amazon, Microsoft, Google, Dellate, Slalam, Nvidia and Allen Institute for AII (AI) of Siat.
“Literally, 10 minutes ago, we were able to get a result that nobody had ever seen before … because no one was able to run this analysis across the data of four cancer centers,” said Bodhisattva Prasad MajumdarAn AI2 research scientist, in an interview on Monday afternoon.
Alliance members spent most of the time last year to create a system that allowed AI to learn from all data in the organization without recording patients in the central database.
The system is a federated learning platform: models are trained locally on the D-Native data inside the firewall of each organization and only the shortcomings of that teaching are divided. These briefs then come together to improve the strength and accuracy of the models.
“It’s a matter of moments that we got together to launch this platform in just one year and we did it as a united alliance with a shared mission to eradicate cancer,” said Brian M Bot In a press release, Fred Hutch, director of the CAIA strategic coordination center.
The AI2 has adapted its new Asta Datavoizer system for the alliance, enables the equipment to analyze data at cancer centers without removing or releasing the patient’s records. The Seatol-based Institute is officially launching Datavoizer this week as part of its broad Asta platform for scientific research.
The AI2 equipment acts as an AI agent, scientists allow to ask questions about data in simple language and take clear answers supportable codes and visualization. It gives clinicians and researchers who are not coder to explore the data and create their own insight.
According to the Cancer AI alliance, researchers from participating centers have launched eight projects using the platform – predicting treatment reactions, identifying biometers and a rare trend of cancer in the larger patient population.
The coalition says they have planned to scale the system next year, added more cancer centers and enable dozens of additional research models outside the initial projects.
More information about the Cancer AI alliance is expected to come today Madrona IA Summit Seattle, where the alliance was originally announced a year ago.