Core Concepts
Developing a robust and scalable process to securely transfer, process, and integrate radiology images with structured and unstructured clinical data in a research environment.
Abstract
The paper describes the experiences and challenges in establishing a trusted collection of radiology images linked to the United States Department of Veterans Affairs (VA) electronic health record database. Key insights include:
Uncovering the specific procedures required for transferring images from a clinical to a research-ready environment, including image identification, batch transfers, and data integration.
Identifying roadblocks and bottlenecks in the process that may hinder future efforts at automation, such as metadata management, data quality, and trust issues.
Highlighting the need for having images linked to both structured and unstructured clinical data within the same research environment, as clinical systems are not set up for research.
Discussing the importance of addressing security, privacy, access, compute, and cost considerations when scaling the image integration process to multiple modalities and facilities.
The iterative approach allowed the team to develop an automated pipeline for transferring images and associated metadata while ensuring the security, privacy, and integrity of the image data. This process facilitated researcher capabilities with creating multimodal predictive modeling using deep learning techniques on chest X-Rays and MRI data.
Stats
The pilot received 263,000 chest x-rays and 729,000 MRI files for a sum total of 1,011,000 medical image files.
The VA corporate data warehouse also contained 24.8 TB of structured data and 13.7 TB of unstructured clinical notes.
The image transfer pipeline can process approximately 157 images per second (500k+ files per hour).
Quotes
"The sheer volume of data as well as the computational needs of algorithms capable of operating on images are extensive."
"Having a trusted paradigm for gathering and organizing imaging in a research environment allows researchers to focus primarily on research and less on planning and engineering."