Philips uses AWS ML to improve healthcare interoperability

Interoperability is a critical enabler for exchanging patient information between healthcare professionals and systems/applications to support evidence-based care at scale. The core of interoperability is sharing data between healthcare IT (HIT) applications, within and across healthcare organizations. Patient data is siloed in HIT applications, including Electronic Health Records (EHRs), medical devices, lab systems, radiology PACS systems, and more.

International standards have been developed to support data sharing, including HL7/FHIR, IHE, and DICOM. However, implementation of these standards has been variable, with each HIT application having different requirements to meet customer-specific workflows and business needs. Every integration between HIT applications is unique and complex, requiring extraordinary effort by teams of engineers—over periods of weeks or months—to manually facilitate data sharing across applications.

Philips is a multinational healthcare company providing innovative solutions to drive high-quality care delivery across the health continuum. Philips diagnostic informatics solutions are used by healthcare providers around the world to streamline workflows, improve care collaboration, enable high quality diagnoses, guide treatments, and more. Interoperability is at the core of diagnostic informatics, and data sharing across the health continuum is a crucial enabler. Philips interoperability solutions are used by numerous healthcare enterprises and health systems to create connected suites of information systems. Working at scale, Philips understands the complexity of healthcare IT environments, including the need to develop novel AI solutions to improve interoperability and reduce the effort required to share data between HIT solutions.

Philips and AWS are exploring the ability to use natural language processing (NLP)  algorithms to automatically detect commonalities in data elements between HIT applications. Simultaneously, they’re exploring how NLP algorithms can be used to predict appropriate translations and allow two or more HIT applications to share data seamlessly. Philips and AWS are applying the NLP algorithms to a real-world use case to demonstrate the ability to reduce manual effort when integrating two HIT applications.

Interoperability use case

AWS has developed sophisticated services that enable advanced AI/ML solutions to be developed and deployed across the healthcare value chain. Amazon Simple Storage Service (Amazon S3) is used for data staging. Amazon SageMaker and Amazon SageMaker Studio feature a suite of ML modeling and deployment services used to prepare, build, train, and tune ML models. This is facilitated by Amazon SageMaker Studio notebook instances, which seamlessly integrate with backend ML services like TensorFlow on AWS.

Philips and AWS are building value proofs to establish the ability for AI-algorithms to facilitate data sharing between Philips PerformanceBridge and a hospital’s EHR system. Philips PerformanceBridge integrates operational data from one or more EHRs, providing analytics and real-time workflow intelligence to enhance healthcare operations. The focus of this proof-of-concept (POC) study is to establish the ability of AI to reduce the number of manual steps required to integrate Philips PerformanceBridge with an EHR.

For the POC study, multiple AI-NLP algorithms are being trained to translate EHR data into a standardized format easily consumed by Philips PerformanceBridge. The Philips and AWS teams have trained these AI algorithms in Amazon SageMaker to facilitate deployment at scale for use by the Philips data integration team. Amazon SageMaker was key for rapidly training and deploying algorithms for the POC study. Given that Philips PerformanceBridge processes more than 100 million data sets every year, Amazon SageMaker is central for running inference at scale.

Solution overview

AWS ML accelerates interoperability using advanced deep learning models. The AI algorithms can detect and translate data elements provided by the EHR system into the format required by Philips PerformanceBridge system. Using AI algorithms will reduce a multi-step integration process for Philips PerformanceBridge down to a few steps. This has profound implications for the accelerated delivery of PerformanceBridge and other diagnostic informatics solutions to Philips customers.

Conclusion

Philips and AWS ML are creating the foundation to automate data interoperability and reduce lengthy implementation times across Philips diagnostic informatics solutions. Accelerating data interoperability using ML has the potential to improve healthcare in many ways, including  enabling seamless patient-centric care, reducing the cost and effort in managing complex IT environments, monitoring data quality, and democratizing interoperability so vendors of any size and/or solution can easily share data. Philips and AWS see a future where data interoperability is seamless and cost-efficient without sacrificing data quality. AI will make interoperability an automated process that occurs behind the scenes, enabling healthcare stakeholders to focus on patient-centric, data-driven care at scale.

To learn more about AWS for Health—an offering of curated AWS offerings and AWS Partner Network solutions used by thousands of healthcare and life sciences customers across the globe, visit AWS for Health and AWS Healthcare Solutions.

To learn more about Philips Interoperability Solutions—an offering to streamline workflows and improve care collaboration by enabling smooth data exchange across healthcare players, visit Philips Interoperability Solutions.