Introduction
Bayer Radiology, a unit of a renowned pharmaceutical company solving pressing societal issues across healthcare and life sciences, faced an increasingly common challenge in large organizations of making their multi-modal biomedical data universally valuable for various users for AI purposes.
Steffen Vogler, a Senior Imaging Technology Scientist at Bayer Radiology, noted that AI developers were spending around 50% of their time making their data AI-ready. They required a platform that could enable internal researchers and partners to build AI medical software while maintaining high standards of scientific rigor.
The goal was to find an efficient, flexible, and scalable database for AI. The search for a suitable solution led Steffen and his team to Deeplake.
The Challenge
Bayer Radiology’s primary challenge lies in the time-consuming data preparation process for AI. Developers had to manage complex data subsets, control data versions, and constantly add new data to their projects.
The multimodality of data – image class, electronic health records, lab data, genomics, proteomics – further complicated matters.
"The elephant in the room is it takes a lot of time to reach the finish line. Especially the data curation part."
Steffen Vogler
Senior Imaging Technology Scientist, Bayer Radiology
Activeloop's Solution
To overcome these challenges, Bayer Radiology implemented Activeloop’s Deeplake solution. Deeplake unifies different data modalities, including healthcare datasets for machine learning, into a single storage solution that can be streamed for pre-processing, training, and inference. The solution was deployed on Bayer Radiology’s Google Cloud account, allowing for efficient search or inference capabilities and the ability to fine-tune own models.
Activeloop's Tensor Query Engine
Querying datasets is a critical aspect of data science workflows that enables users to filter datasets and focus their work on the most relevant data. Deep Lake offers a highly-performant query engine for filtering your data, generating SQL-like queries from natural language.
"We went from a GitHub issue to a working prototype I could use in under 3 weeks."
Steffen Vogler
Senior Imaging Technology Scientist, Bayer Radiology
It's next-level. We've enabled a new human-machine interface that is natural to use and yields high-accuracy results for end-users.
Steffen Vogler
Senior Imaging Technology Scientist, Bayer Radiology
Results
In only 2 weeks, Activeloop was fully integrated into Bayer Radiology’s enterprise cloud platform. The time spent on data preparation was reduced from 50% to a mere fraction, allowing developers to focus more on optimizing AI architectures.
Notably, the solution prioritizes security, adeptly managing sensitive data within Bayer Radiology’s secure data centers, essential for safeguarding Personally Identifiable Information (PII).
2 weeks
Time to IntegrateProcess AI Data in Days
Instead of Spending Up to 50% of Project TimeWords instead of SQL
In-App Chat with Biomedical Data
Future Plans
Bayer Radiology’s collaboration with Activeloop and the successful implementation of Deeplake has paved the way for future collaborations.
Powerful query features to curate subsets in natural language.
Chat with multi-modal data, at scale
Powered by Intel
The Intel RISE and Intel Disruptor Initiative Program further bolstered Bayer Radiology’s collaboration with Activeloop. Intel technology was used at multiple stages in the project, including feature extraction and processing large batches of data. Intel Xeon 4th Generation processors, powering Activeloop’s Tensor Query Engine, facilitated quick and efficient queries across millions of scans, significantly shortening innovation cycles.
As Bayer Radiology continues to innovate and push the boundaries of AI in healthcare, Activeloop’s Deeplake, powered by Intel technology, remains an integral part of its AI tech stack.
Overall, the collaboration between Bayer Radiology, Activeloop, and Intel showcases a successful example of combining cutting-edge technology with heavily regulated industry needs to deliver a sizeable impact. This collaboration promises to redefine how AI enables the pharmaceutical and healthcare industry, with Bayer Radiology leading the charge.
"The Intel RISE program gave us flexibility and room to experiment, setting us up for success."
Steffen Vogler
Senior Imaging Technology Scientist, Bayer Radiology
I am confident that adopting Activeloop was a value-add through improved innovation that will compound as our AI operations grow.
Steffen Vogler
Senior Imaging Technology Scientist, Bayer Radiology
The collaboration between Intel, Bayer, and Activeloop highlights the critical role of fast-paced innovation and economic scalability in advancing healthcare AI. Utilizing Intel’s latest AI Hardware and IA based software assets, Bayer has unlocked the full potential of Activeloop Deeplake. This collaboration has not only delivered significant performance enhancements but also achieved a more cost-effective approach to AI initiatives, paving the way for future advancements in the sector.
Arijit Bandyopadhyay
CTO of Enterprise Analytics & AI, Head of Strategy Cloud and Enterprise – CSV Group, Intel Corporation