Philips Accelerates Insights at Scale for Healthcare Organizations through Innovative Data Annotation Solution Built on AWS

Philips is a health technology company focused on improving people’s lives through meaningful innovation across the health continuum, from healthy living and prevention to diagnosis, treatment, and home care. Since 2014, the company has been offering customers its Philips HealthSuite Platform, which orchestrates dozens of AWS services that healthcare and life sciences companies use to improve patient care. The Philips HealthSuite Platform is a healthcare-ready and regulatory-approved platform that focuses on providing healthcare infrastructure and data services in the cloud by connecting devices, unlocking data, and fostering collaboration. As part of the HealthSuite Platform, Philips is working on a new end-to-end artificial intelligence (AI) development and deployment platform using Amazon SageMaker, which helps data scientists and developers to build, train, and deploy ML models for virtually any use case. Philips had created a suite of annotation tools that it was running using on-premises infrastructure. As part of the new AI development and deployment platform on HealthSuite, it wanted to expand the use of existing annotation tools beyond on-premises capabilities and centralize its data annotation process in the cloud, saving time and money while protecting the security and privacy of sensitive medical information, which is a priority as part of the Philips Data Principles. Previously, someone had to drive to the annotators’ locations—a hospital, for example—to hand deliver data to annotators and collect completed annotations.

Philips’ annotation solution in the new AI development and deployment platform centers around Amazon SageMaker Ground Truth, a data labeling service that makes it simple to label data while giving organizations the flexibility to use human annotators through third-party vendors or their own private workforce. Philips can continue to incorporate its existing annotation applications—some that run on premises on Windows or Linux software and others that are web based, occasionally running on isolated instances. “This project is a way for us to centralize data and use our existing AI annotation tools in a cloud environment so that our annotators can access data from a browser that is connected to cloud-based storage,” says René van Erp, software architecture lead at Philips.

Built natively on AWS, Philips HealthSuite Platform stores the data that it collects for annotation as well as ML model training in a data lake built on Amazon Simple Storage Service (Amazon S3), an object storage service built to retrieve any amount of data from anywhere. To read data and push it to the web-based annotation tools, Philips uses AWS Lambda, a serverless, event-driven compute service that lets developers run code for virtually any type of application or backend service without provisioning or managing servers. After this is complete, annotations get pushed back into the HealthSuite data lake using an additional Lambda function. To route web requests to Lambda, Philips uses Amazon API Gateway, a fully managed service that makes it simple for developers to create, publish, maintain, monitor, and secure APIs at virtually any scale. “It’s very important that we can use our existing annotation tools and don’t have to rebuild them from scratch to make them web based,” says Frank Wartena, program manager for data and artificial intelligence at Philips. The solution built on AWS saves developer time and effort by removing the need to rewrite annotation tools, effort that Philips can now spend on innovative solutions to improve healthcare outcomes.

Annotation tools installed as desktop applications run on Amazon AppStream 2.0—a fully managed nonpersistent desktop and application service for remotely accessing work—and render using SageMaker Ground Truth. “Bringing nonweb applications into Amazon SageMaker Ground Truth is something that wasn’t done before,” Wartena says. “It’s not just relevant in the healthcare industry, but it’s also valuable for others.” The workflow on AWS also simplifies Philips’ ability to keep a full record of anyone who comes in contact with sensitive data—an important regulatory requirement.

As an additional benefit, annotators don’t have to alter the methods that they currently use in their clinics. Physicians working with CT scans, for example, can continue to use the same methods for labeling images in their day-to-day practices that they use to perform annotations for Philips, thereby preserving most of the existing user experience and workflow. And by shifting offline annotation processes to the cloud, Philips speeds up the process, decreasing the total lead time of ML projects in development and reducing time to market. “Through our use of Amazon SageMaker Ground Truth, the clear benefit we see is that we move much faster with existing tooling to a cloud-based annotation workflow,” van Erp says.

Using SageMaker Ground Truth, Philips can reach a much larger qualified global workforce and continually add new datasets at scale, resulting in a wider distribution and volume of data annotations. ML models trained on these datasets can provide more accurate predictions, helping healthcare professionals better diagnose diseases, such as cancer or infections. It also speeds up the annotation process, which reduces the throughput time of the ML projects and brings insights at scale faster to customers.

Philips is collaborating closely with AWS on the development of its AI development and deployment platform, which will include additional Amazon SageMaker capabilities. In addition to annotation, these include model training, validation, and model deployment and monitoring. As a result, Philips will be able to help doctors save time and make better-informed decisions. “Using Amazon SageMaker Ground Truth, we have a much more efficient process, especially within our full AI development and deployment platform,” van Erp says. “It’s an efficiency and productivity boost, which has a tremendous impact on lead time and overall cost.”

To learn more, visit https://aws.amazon.com/sagemaker/data-labeling/.