PICTURES is investigating data science, engineering and cyber security to enhance the SMI. Supporting bigger datasets, smarter searches, more ambitious high-tech projects.
The Project
The PICTURES project is about increasing the capabilities of Safe Havens to support emerging technologies such as artificial intelligence (AI) and new medical image data types like, MRI, CT, Xray etc in health data research. PICTURES is investigating data science, engineering and cyber security to enhance our capabilities in supporting bigger datasets, smarter searches, and more ambitious high-tech projects.
It is a 5-year MRC funded programme of work (ending 30th June 2024) led by the University of Dundee, in partnership with the University of Edinburgh, Abertay University, NHS Scotland, and industry collaborators, with additional support from EPSRC and as part of HDR UK.
Safe Havens, or Trusted Research Environments (TRE), are a place where pseudonymised linked health data can be made available for approved research in a safe and secure manner. In Scotland they enforce the 5 Safes and associated measures to maintain public trust that your personal health data is used for the benefit of patients and / or the population as a whole.
Information about how the project unfolded is summarised in this Recording:
• PICTURES Overview (11 mins)
The Solution
The PICTURES programme of work has delivered a fully functioning Scottish Medical Imaging (SMI) platform able to manage and provision population level, research-ready medical images and associated Radiology Reports, routinely collected since 2010, linked to other health-care data. Anonymised research-ready data is made available in a secure, Trusted Research Environment or ‘Safe Haven’ with the tools and compute power needed for large scale analysis. The software developed for SMI and PICTURES are open-source wherever possible, and can be found on GitHub at https://github.com/SMI/SmiServices. Contributions are welcome and appreciated. Please let us know via Contact Us if you are using any of our software for your own project.
Routinely collected or ‘real-world’ data are extremely useful for healthcare research. However, using these datasets is challenging because they are large and unwieldy, and require specialist tools and skills to work with. Platforms like SMI provide the capabilities and privacy controls needed for impactful healthcare research.
The three research questions we had going into this project were:
- How to build research relevant cohorts from messy, unstructured, identifiable data
- How to handle big data in a scalable manner
- And how to protect patient confidentiality
Our team of researchers, developers and clinical experts delivered innovative solutions to address challenges and risks, summarised in the following Recording and Posters:
Video
• SMI Overview (11 mins)
Posters
• Text-based Medical Image Classification by Body Part.
• Natural Language Processing of Radiology Reports.
• Identifying MRI sequence type from pixel data.
• Scottish Medical Imaging (SMI).
Publications
◦ The Scottish Medical Imaging Archive: 57.3 million Radiology Studies Linked to their Medical Records
◦ Next-Generation Capabilities in Trusted Research Environments: Interview Study.
The Context
Safe Havens were introduced in Scotland around 2008 as places where pseudonymised linked health data could be made available for approved research in a safe and secure manner.
As diagnostic technologies improved, vast amounts of high-quality clinical image data have become available. Every x-ray, CT scan, ultrasound and MRI contains a wealth of information – not just about disease but also about what healthy bodies look like. As compute power has increased, the tools and knowledge for tapping into data have become more common.
Data science now provides the means to support medical practitioners through machine learning and Artificial Intelligence (AI). AI in health care can help speed up diagnoses by computing vast amounts of information in seconds, easing the reporting burden for Radiologists by pre-populating routine fields of information for them and spotting disease early through pattern recognition. For diseases such as cancer, early detection from imaging greatly improves the chances of survival.
Machine learning works best when it is trained on very large datasets containing all the patterns we want it to recognise. Those patterns can be found in the images and other health data collected through routine visits to healthcare settings. AI can detect subtle changes invisible to the human eye and will give consistent results 24×7 which even the best radiologists cannot deliver 100% of the time.
The aim of PICTURES continues to be identifying best practice for both the technical environment and governance of safe havens to allow such research to happen safely and securely. But most importantly, maintaining the trust of the public that enabling healthcare research with routinely collected medical image data in this way delivers genuine public benefit.
