Building a serverless MRI pipeline for precision medicine on AWS

This article describes how the Australian Epilepsy Project built a serverless MRI processing pipeline on AWS to accelerate epilepsy diagnosis and treatment across Australia.

• Reduced MRI processing time from 25 hours to 20 hours per participant using AWS services
• Achieved over 35% cost reduction compared to previous infrastructure solution
• Uses AWS HealthImaging for DICOM storage and retrieval of medical imaging data
• Employs ECS on Fargate for scalable, serverless compute for long-running MRI workflows
• AWS Step Functions orchestrates workflow state management from ingestion through export
• Amazon S3 stores all data with lifecycle management; RDS PostgreSQL tracks metadata and audit trails
• Lambda handles lightweight postprocessing; Systems Manager enables secure data export
• Custom neuroimaging portal provides staff with study management and quality control capabilities
• Supports 1,500+ patients across seven scanning sites in five Australian states

The serverless architecture demonstrates how AWS services enable scalable, cost-effective medical imaging platforms for precision medicine research at national scale.