AWS HealthImaging
A HIPAA-compliant fully managed storage service for storing, searching, and retrieving DICOM medical images at scale
Overview
AWS HealthImaging is a dedicated storage service for storing and managing medical image data (DICOM format) such as CT, MRI, and X-ray at scale in the cloud. Built on a HIPAA-compliant security foundation, it maintains petabyte-scale image data at low cost while meeting the compliance requirements demanded by healthcare organizations. Its proprietary compression technology (HTJ2K) reduces storage costs by up to 40% compared to traditional DICOM storage while retrieving images with sub-second latency. The DICOM metadata search API enables rapid identification of image sets by conditions such as patient ID, exam date, and modality, also streamlining data supply to AI/ML pipelines.
Data Store Structure and DICOM Import
HealthImaging's data structure consists of three hierarchical levels: Data Store, Image Set, and Image Frame. A Data Store is a logical container that can be separated and managed by department or project. DICOM file import is performed via import jobs from S3 buckets. During import, DICOM metadata is automatically indexed, and the search conditions of the SearchImageSets API include Patient ID, Study Date and Time, Study Instance UID, Series Instance UID, and Accession Number. According to the developer guide, most lossless image frames are transcoded to HTJ2K (High-Throughput JPEG 2000) during import (JPEG 2000 Lossless can also be chosen when the data store is created), while some files retain their original transfer syntax. For migration from existing PACS (Picture Archiving and Communication System), online transfer using the DICOMweb standard STOW-RS endpoint is also available.
High-Speed Image Retrieval and AI/ML Integration
HealthImaging's image retrieval API is optimized for frame-level random access. Partial retrieval operations such as fetching only specific slices from hundreds of CT scan slices or extracting only specific MRI sequences are performed at high speed. This eliminates the need to download all images for AI model training data preparation, enabling supply of only needed frames to SageMaker training jobs. One example of an event-driven architecture with Lambda triggers an inference pipeline on new image imports and attaches anomaly detection results as metadata.
Security, Compliance, and Cost Design
Since medical image data contains personal health information (PHI), HealthImaging is designed as a HIPAA-eligible service. The developer guide states that data at rest is encrypted by default with a service-owned KMS key (a KMS customer-managed key can also be used), and data in transit is encrypted with TLS 1.2 (as of October 2026). IAM policies provide access control at the Data Store level, and CloudTrail records management events by default, while data events such as GetImageFrame are recorded when enabled. The cost structure consists of storage fees (GB/month), API request fees, and data transfer fees. For data lifecycle management, the developer guide describes Intelligent Tiering, which automatically moves image sets to a lower-cost access tier as access patterns change.
References (Official AWS Resources)
The primary sources for this page are the official AWS website and documentation. Check the official pages below for the latest specifications and pricing.
- Amazon S3 official page
- Amazon SageMaker official page
- AWS Lambda official page
- AWS Documentation (official)
If this page and the official documentation disagree, treat the official documentation as authoritative.