Skip to main content

Research (scientific computing)

Research development – ‘perfusion2diagnosis’

perfusion2diagnosis (p2d) is clinical software developed by scientific computing, in collaboration with nuclear medicine physics, which analyses specific scans performed by nuclear medicine, and produces clinical reports that can aid differential diagnosis of dementia.

Recently, Sofia Michopoulou, in nuclear medicine physics, has developed ML models which provide supportive information for clinical diagnosis funded by an NIHR clinical lectureship called BRAIN AI (Biomarker Research Assessing Inflammation in Neurodegeneration using AI).

Data from specific regions of the brain scans are input into three distinct models, which generate predictions as to whether the current dataset is considered normal or abnormal, shows neurodegeneration, or shows Alzheimer’s disease including mixed dementias. Multiple presentations and publications have been delivered outlining potential benefits and limitations of these models.

Working with Sofia, next steps include integrating the models to become part of the existing clinical software and contributing additional data to the reports it generates, subject to satisfying the needs of clinical users and requirements of the MDSG (medical device software group) QMS.

Glucose management – technologies development

Our work on glucose management in insulin dependent diabetes is centred around three key themes:

  • Dynamic risk management (intelligently trading safety, flexibility and patient burden)
  • 'Access to success' - focusing on sustained effective use of technologies, including through education and interface design
  • Trustworthiness of artificial intelligence powered insulin delivery.

Together with colleagues and students at Bristol and Southampton universities, we have secured funding through a variety of grant funding bodies and charities (including UKRI, Diabetes UK, The Health Foundation) and are contributing to healthcare AI projects linking more than a dozen university, healthcare and commercial partners across Europe. Multiple presentations and publications have been delivered to both scientific and healthcare audiences, highlighting both the opportunities and risks presented by these technologies, and reinforcing the importance of sustainably connecting people (patients, carers, and healthcare professionals) to these technologies.

Research imaging data management service

Our service is tailored to accommodate the dynamic requirements of researchers and the evolving landscape of technology. We have implemented and continuously refine automated, in-house developed systems centred on XNAT (Extensible Neuroimaging Archive Toolkit), ensuring thousands of DICOM research images are pseudonymised and accessible to researchers.

We utilise XNAT, which offers a solution for managing imaging research data, both within and beyond the UHS network. XNAT is widely used across imaging research institutions both here in the UK and globally, and, despite its name, is not limited to neuroimaging. XNAT is an open-source platform designed for storing, sharing, and managing medical imaging and related data. Those features mean it can be an effective tool for your projects requiring storage and sharing of imaging data securely with collaborators both within and outside of UHS.

Key components of our data management system include:

  • Study request forms: These forms capture essential details such as data and project summaries, lead investigators, and specific project needs like pseudonymisation protocols and data retention policies. Completing a study request form is the initial step to engaging with our service, allowing us to evaluate your project's viability and provide information on potential costs. To request a form, please contact scicom@uhs.nhs.uk.
  • Data acquisition and pseudonymisation: Our system facilitates the direct transfer of studies from scanners or retrieval from UHS's PACS via bespoke automated software. We ensure the privacy of imaging data through rigorous in-house pseudonymisation processes, which meticulously cleans identifiable DICOM tags, covering 250 DICOM tags with flexible protocols.
  • Data storage and sharing: Our infrastructure supports sharing imaging data both within UHS and externally. This arrangement enables data transfers to trusted research environments (TREs), secure data environments (SDEs), and other necessary locations.

Through these measures, we aim to provide a robust framework that supports the research community's imaging data management needs, ensuring security, flexibility, and efficiency.