CT scanners use X-ray radiation and software to create 3D images of patients to help diagnose a variety of conditions. Every year, medical physics teams across the NHS perform legally required audits to ensure their scanners are working correctly. Part of this process involves optimising the radiation level of their scanners: ensuring the minimum amount of radiation is used to deliver an image of sufficient quality.
In 2026, we are collaborating with the Medical Physics team in NHS Lothian to support the use of real-world patient data for their optimisation process.
Optimising CT scanners
CT scans are widely used in the NHS to diagnose injuries, detect diseases, and guide medical procedures. The optimal radiation dose for an individual patient depends on several factors, such as the body part being scanned and the patient’s body mass index (BMI).
Modern CT scanners have sophisticated hardware and software features designed to assess the size and shape of the patient and adjust the radiation level accordingly. How well these scanners adjust to different patients needs to be legally checked, otherwise they might deliver fully optimised images for one patient, but high dose or low-quality images for another. Importantly, if too much radiation is used, then this can be harmful to the patient in the long-term.
Optimising NHS Lothian scanners
Traditionally, NHS Lothian’s Medical Physics team has used CT phantoms to optimise radiation levels. A phantom is a highly specialised physical object used to check scanning equipment and radiation safety. Alongside the CT phantoms, a small sample of patient data was occasionally utilised, which the team had to access and process individually. Unfortunately, it has not been possible to repeat this manual process for large numbers of tests, due to the time needed to organise samples of patient images by sex, age, and BMI.
The Medical Physics team wanted to see if using larger amounts of real patient data would give better results for scanner optimisation. In a pilot collaboration, they sought DataLoch’s involvement to automate a data pipeline to sample de-identified patient data for using in their audit process.
Addressing the storage challenge
The key issue that we needed to mitigate was the available data-storage capacity, since the images produced by scans require significant space. We have had to work closely with the Medical Physics team to understand which parts of the scans were needed for their analysis. If we filled the available storage capacity with unnecessary parts of the scans, this would have limited the data that would be genuinely helpful for optimising the scanners.
Initial impacts of the collaboration
Our pilot project is split into two initial phases.
In phase 1, DataLoch delivered around 1800 de-identified CT scans of patient heads – linked to data about the patient’s sex and age at the time of scan. Using this data, the Medical Physics team has already detected differences in the setup of two CT scanners that affect their ability to image the brain appropriately. These differences would not have been discovered from standard audits or phantom scans, and has been achieved without diverting frontline radiologists from their patient-facing tasks or taking scanners out of use for manual testing.
We are now ready for phase 2: moving on to full-body scans. For this stage, BMI data from patients is an important addition to the information used in the traditional optimisation approach. Our Medical Physics collaborators are optimistic about what we will learn in the next stage of our pilot.
Case Studies
