Applying artificial intelligence and machine learning (AI/ML) within the realm of medical imaging has proven to be remarkably effective. The FDA has approved close to 700 AI/ML-enabled medical devices, with 77% specifically tailored for radiological purposes. Medical imaging plays a vital role in diagnostic strategies, evidenced by approximately 500 million imaging exams conducted on adults across the United States in 2023 alone—averaging nearly 2 exams per adult. Given that medical data retention laws require archiving for a minimum of 5 years in most states, the need for efficient imaging solutions is pressing. Medical imaging processes are often time-consuming and labor-intensive, requiring extensive training for radiology technicians and professionals. While a basic X-ray may take less than a minute, advanced imaging techniques like Magnetic Resonance Imaging can extend to over an hour.
Healthcare organizations encounter significant challenges when trying to leverage medical imaging data for machine learning. The first hurdle is the fragmented nature of data storage, as imaging data often resides in various disconnected repositories within healthcare systems. Additionally, strict compliance regulations necessitate meticulous data pre-processing, including de-identification, standardization, and harmonization, before integrating it into ML models. The sheer volume of medical imaging data exacerbates these issues; hospitals typically generate around 50 petabytes of data annually, with approximately 80% made up of imaging files, thus requiring solutions that can efficiently and securely handle such vast datasets.
Flywheel Solution Overview
Flywheel operates on AWS, providing a software as a service (SaaS) solution that facilitates comprehensive processing of medical imaging data at scale. It integrates seamlessly with Amazon Simple Storage Service (Amazon S3), ensuring secure, scalable storage while optimizing costs. Through this partnership, healthcare organizations can manage, analyze, and share imaging data effectively, cutting down on costs and time spent on data preparation significantly. Reports indicate that AWS customers have achieved up to a 94% reduction in medical imaging workflow costs, transforming months-long manual operations into mere days.
Amazon S3 effortlessly scales to accommodate growing data demands while maintaining performance. Given the sensitivity of health data, security remains a top priority; Amazon S3 is HIPAA eligible and offers enhanced encryption, access controls, and audit logging to uphold data integrity and compliance. By utilizing Flywheel’s Interfaces tab, users can easily mount multiple Amazon S3 buckets for data import, export, or both with just a click.
To ensure secure access to your imaging data, connect your Amazon S3 bucket to Flywheel by configuring AWS credentials. Specify your S3 bucket (for instance, s3://my-flywheel-bucket) within the Flywheel interface and provide the AWS Access Key and Secret Access Key linked to an AWS Identity and Access Management (IAM) role with suitable permissions. This setup, as illustrated in Figure 2, establishes a secure connection between Flywheel and your S3 storage.
Once mounted, you can import data into a Flywheel Project using the Web Import tool. For example, we’ve loaded one subject from the NSCLC-Radiomics-Interobserver1 project, which is part of the Imaging Data Commons and freely accessible from the Registry of Open Data on AWS.
Within a Flywheel project, data is categorized by Subject, Session, Acquisition, and File. With Flywheel, there’s no need to search multiple locations for varying data types; everything is consolidated under one Subject. For instance, a patient may receive an X-Ray during one visit and an MRI during another, along with surveys stored as PDFs or CSVs—all filed under the same Subject.
The zero-footprint radiology viewer in Flywheel allows users to visualize, annotate, and segment medical images efficiently. Flywheel also optimizes storage costs by transitioning data according to usage patterns through Amazon S3 Intelligent Tiering. The Reference-in-Place feature allows Flywheel to read source files without duplicating them, reducing storage costs while treating the source location as primary storage.
Preparing Data for Analysis with Flywheel on AWS
Once imported into Flywheel, data can be processed and analyzed using Flywheel Gears, which leverage configurable Amazon Elastic Compute Cloud (Amazon EC2) instances for time and cost efficiency. If your Project has a Flywheel Gear Rule activated, data pre-processing can be automated. For example, our Example Project employs a Gear Rule that extracts Digital Imaging and Communications in Medicine (DICOM) header metadata and classifies incoming DICOM files. The results of these operations are displayed in the Session dashboard.
Moreover, training machine learning models can be executed directly within the Flywheel platform using Flywheel Workspaces. Cohort building is straightforward with Flywheel’s Search feature, allowing searches based on medical image attributes or relevant structured data.
Sharing data on Flywheel is uncomplicated; you can collaborate with others within the platform using granular roles and permissions or utilize the Flywheel SmartCopy feature. Alternatively, you can export a Flywheel project to an Amazon S3 bucket for retrieval.
Integrating Flywheel into your AWS solutions reduces the time and costs associated with data aggregation, processing, and curation, allowing you to focus more on generating valuable insights.
Empowering Healthcare with Flywheel and Amazon S3
The combination of Flywheel and Amazon S3 presents a robust solution for enhancing medical imaging workflows. This integration enables healthcare organizations to manage, analyze, and share extensive imaging data efficiently while significantly lowering costs and preparation time. As medical imaging data volumes continue to surge, Flywheel on AWS establishes a solid, scalable foundation for advancing AI-driven healthcare innovation.
To discover how Flywheel can maximize the potential of your medical imaging and health data, visit Flywheel.
For further insights, you can check out this authoritative resource on the topic. Additionally, this is an excellent resource for those interested in learning about Amazon’s leadership development interviews.
Leave a Reply