We are excited to announce the general availability of AWS HealthImaging, a purpose-built service that helps builders develop cloud-native applications that store, analyze, and share medical imaging data at petabyte-scale. HealthImaging ingests data in the DICOM P10 format. It provides APIs for low-latency retrieval, and purpose-built storage.
Our healthcare customers tell us they want their care teams to have the best medical imaging applications, and they want to reduce the complexity of managing infrastructure. Our research focused customers want to analyze imaging data at scale, and to accelerate collaboration and discovery across their organizations. Both of these customer groups express a desire to have all their organization’s medical imaging applications work from the same store of data. The cloud can help address these customer needs. With HealthImaging, builders, like AWS Partners who provide medical imaging applications and research solutions, can focus more on tackling these customer’s challenges, instead of worrying about infrastructure.
Expansion of medical imaging in care delivery and research
For over 100 years, healthcare providers have used medical imaging, such as X-ray, MRI, and ultrasound, to noninvasively look inside patients. The Nobel Prize laureate Marie Curie was one early pioneer of medical imaging. She developed vehicles with x-ray equipment, called “petite Curies”, to help battlefield surgeons provide better care during the first World War. Today, medical imaging is used to diagnose and monitor a wide range of health conditions, spanning oncology, trauma, stroke, and more. Globally, more than 3.6 billion medical imaging procedures are performed each year, collectively generating exabytes of medical imaging data.
The healthcare system is struggling to meet the growing demand for medical imaging procedures. Since 2008, the average number of imaging procedures assigned to radiologists in the US has increased from 58 per day to 100 per day. In that same period, the typical imaging study size doubled to nearly 150 MB. As a result, radiologists need new technologies to increase productivity, and AI is increasingly being used to streamline cognitively taxing workflows and minimize errors.
Healthcare provider IT groups are often responsible for the infrastructure that hosts new and archived medical imaging studies. These organizations are managing rapidly growing imaging archives, typically on-premises. They see infrastructure consuming considerable square footage, IT staff, and operational budgets. They are also managing a growing array of enterprise applications that require access to the same medical images, each with different latency and resolution needs. This results in their storing several cached copies of each image for various applications, plus additional copies for long-term storage. Consequently, they see higher storage costs due to data duplication, and uncertainty around which version of an image is authoritative.
Collaborations across care teams and research groups can lead to still more copies of data. Care teams typically require fully identified data, whereas teams building AI models may prefer to use de-identified data. With traditional on-premises architectures, customers may need to make additional copies of data for each use case, resulting in higher storage costs and greater operational complexity. The expansion of medical imaging has created vast data sets that can be used to develop new language and computer vision AI models. Yet legacy data silos impede innovation by limiting researcher’s access to data.
Introducing AWS HealthImaging
HealthImaging provides a purpose-built medical imaging data store that simplifies provisioning infrastructure, giving customers more time to care for patients and perform research. With HealthImaging, all the applications across an organization can access a single authoritative copy of data without duplication, and users can securely access the data from anywhere. With just a few clicks in the HealthImaging console, you can provision a data store capable of hosting petabytes of medical imaging data, maintaining every image ready for low-latency retrieval. Further, HealthImaging reduces the amount of infrastructure required to operate enterprise imaging solutions, and that helps you save cost and reduce operational complexity.
Customers can use applications built on HealthImaging to take advantage of low storage costs for image archives, without worrying about hardware refresh cycles or capacity planning. As new data is generated by imaging modalities, it can be imported to HealthImaging and be immediately available for retrieval by diagnostic applications like PACS (picture archiving and communication systems). AWS DataSync, AWS Direct Connect, and purpose-built gateways from AWS partners can simplify moving data from the edge to the cloud.
Figure 1. How AWS HealthImaging works, from importing data generated by modalities (like CT, X-ray, etc.) to low-latency retrieval by clinical applications (like PACS) and research workflows.
AWS Partners are innovating on behalf of customers
AWS Partners are already using HealthImaging to reimagine the medical imaging solutions used by radiologists, care teams, and researchers.
Wake Forest Baptist Health is making clinical content more accessible to radiology students, with solutions from Apollo Enterprise Imaging and HealthImaging.
“We need the ability to scale, share, and view medical images for research and education across our enterprise and with collaborators around the world. Working in collaboration with Apollo EI, using AWS HealthImaging and their cutting edge enterprise imaging repository technologies, makes that possible.” — Josh Tan, Systems Manager at Wake Forest Baptist Health
Philips, a global leader in medical imaging, intends to use HealthImaging as a foundational element of their next-generation medical imaging suite.
“Our vision is to help clinicians and staff manage growing workloads and optimize workflows to speed time to diagnosis and treatment of patients. AWS purpose-built services, like AWS HealthImaging, can help Philips innovate faster and serve our customers and their patients. Our cloud-enabled HealthSuite Imaging PACS intends to use AWS HealthImaging to improve experiences and accessibility for clinicians all over the world.” — Shez Partovi, Chief Innovation & Strategy Officer and Chief Business Leader of Enterprise Informatics, Philips
Moving your data from on-premises silos to the cloud opens new opportunities for innovation. HealthImaging is integrated with Amazon SageMaker for machine learning, giving you access to GPU accelerated computing. NVIDIA is investing in hardware accelerated tools and open-source frameworks that work seamlessly with HealthImaging to advance algorithm development and AI adoption across medical imaging.
“MONAI, co-founded and accelerated by NVIDIA, is a domain-specific, medical imaging AI framework that speeds the translation of research breakthroughs and AI applications to clinical impact. With the integration of MONAI and AWS HealthImaging, medical images can be viewed, processed, and segmented in near real time — optimizing physician workflows, enhancing patient experiences, and helping hospitals improve efficiencies.” Prerna Dogra, Global Lead for Healthcare AI Products, NVIDIA
Partners are also making it easier to realize the benefits of HealthImaging. Dicomatics is a medical informatics company offering a range of solutions, including ones that assist enterprise-class data migrations from legacy environments to modern cloud-based environments using HealthImaging.
“Dicomatics is a leader in seamless and scalable medical imaging data migration. From on-prem to the cloud, we excel in handling petabyte-scale and complex migrations. With the power of AWS HealthImaging, our customers now have a specialized cloud service dedicated to storing their valuable data, fueling clinical workloads, and groundbreaking research.” – Aviram Biton, Strategic Partnership, Dicomatics.
Cost-effective storage for medical images
HealthImaging provides cost-effective storage that can reduce the total cost of ownership for storing new data and image archives of any size. HealthImaging offers a Frequent Access storage tier for new and frequently accessed data, and an Archive Instant Access tier that is cost-effective for infrequently accessed data. Data stored for more than 30 days is automatically moved to the archive tier. The behavior is similar to the Amazon Simple Storage Service (Amazon S3) Intelligent-Tiering storage class, passing on cost savings to customers.
Both storage tiers of HealthImaging support data retrieval in milliseconds. Every image frame stored in HealthImaging can be accessed and rendered with sub-second latency, reducing the need for customers to stage data in expensive block storage volumes.
Scalable data ingestion and processing
You import DICOM P10 files to HealthImaging by launching asynchronous import jobs. You can scale up by running multiple import jobs concurrently. Individual DICOM P10 files are imported as image frames, and automatically organized in image sets with consistent metadata at the patient, study, and series levels. Copy and update APIs make it easy to manage image sets, as your specific workflow requires.
The pixel data for each DICOM P10 file is encoded as High-Throughput JPEG 2000 (HTJ2K), a state of the art image compression codec offering efficient lossless compression and resolution scalability. Customers with large archives may see their storage volumes reduced with HTJ2K, which can help reduce costs. HealthImaging validates that all pixel data is transcoded successfully by providing checksums for each image frame imported. These checksums are added to the image set metadata, so you can independently verify the lossless image processing upon retrieving an image frame.
The metadata in DICOM P10 files, like patient identifiers and procedure details, is automatically normalized at the patient, study, and series levels. This helps eliminate inconsistencies and improve data quality. All metadata is preserved, with normalization performed based on the Registry of DICOM Data Elements. Further, the normalized elements can be accessed with developer friendly keys, like
PatientID, rather than hexadecimal DICOM tags.
There is no charge for importing data to HealthImaging. The pixel data encoding and metadata normalization are performed automatically. This means customers can reduce costs when migrating to HealthImaging from self-managed infrastructure for DICOM ingestion.
Optimized APIs for real-time applications
Existing image transfer protocols (such as DIMSE and DICOMweb) can provide poor latency performance when streaming from the cloud. Yet, radiologists require low-latency experiences for interactive workflows and diagnostic applications. Therefore, HealthImaging offers optimized APIs for low-latency retrieval of pixel data and metadata.
HealthImaging is purpose-built to deliver industry leading performance in data retrieval and image loading, thanks to its efficient metadata encoding, lossless compression, and support for progressive resolution image data access. Applications and AI algorithms can efficiently access study metadata via APIs, without having to load image data. Similarly, applications can directly load image data via APIs, without any compromise in image quality thanks to state of the art image compression.
The HTJ2K codec is an order of magnitude faster than JPEG2000, and at least twice as fast as all other DICOM transfer syntaxes. With HealthImaging, applications can leverage HTJ2K with single instruction multiple data (SIMD) instructions, for exceptional image decoding performance. Further, modern browsers can utilize Web Assembly SIMD (WASM-SIMD) to bring industry leading performance to zero-footprint web viewers. Thus, HealthImagining enables applications to retrieve and load data with latencies meeting the most demanding interactive use cases.
Our customers and partners are tirelessly innovating on behalf of patients. HealthImaging is helping them provide high-quality care to more patients in need. HealthImaging makes it easier to manage medical imaging data at petabyte-scale, with industry leading performance, and freedom from worrying about infrastructure.