Diverse Datasets — How to Find, Curate & Leverage them to Fuel AI Research

R&D departments are gravitating toward cloud-scalable, data-driven strategies to take advantage of the artificial intelligence (AI) boom. But AI is incredibly data-hungry, and many life sciences organizations are looking for better ways to feed the beast, by accessing and leveraging both internal and external datasets. To make internal data most useful, enterprises need to establish basic data management practices for centralization, curation and computing. This has been especially challenging for imaging data, given its complex nature. MR and CT scans, X-rays and other imaging data hold tremendously valuable information for AI but extracting and cataloging this data can seem prohibitive at an enterprise level. Similar problems apply when seeking external data — with additional challenges in costs, compliance and ensuring quality and consistency. Some researchers are beginning to leverage federated learning & analytics — sending algorithms to data where it resides — to sidestep challenges with compliance and data privacy. This approach holds great promise, but the basic data management issues remain, and researchers need mechanisms to understand the quality, scope and other attributes of a dataset before using it in a federated learning effort. Join Flywheel’s experts in imaging research and data governance for a discussion of how these issues are playing out in life sciences organizations. How can research teams better take advantage of datasets both inside and outside their walls to train AI models and accelerate R&D? And how are organizations planning for the future while also dealing with research challenges in the here and now?