Spinal X-ray image analysis to aid early osteoporosis detection

Specialist computer software is being developed in Manchester to help identify the early onset of osteoporosis.

With £660,000 from the Department of Health and the Wellcome Trust through the Health Innovation Challenge Fund, Manchester University will work with Optasia Medical and Central Manchester University Hospitals NHS Foundation Trust to develop a system that automatically analyses the subtle changes that indicate the presence of the disease.

According to Tim Cootes, a professor of imaging from the Institute of Population Health based Manchester University, the research will focus on vertebra in the spine and how they appear in x-rays, particularly DXA dual energy, low dose X-rays.

Vertebral fractures provide an early indication of osteoporosis, a disease that weakens bones, leaving them prone to fracture. Osteoporotic vertebral fractures also double the risk of future hip fracture.

However, the asymptomatic nature of the fractures means they can go unnoticed by the patient.

‘Fifty per cent of these injuries don’t cause any pain at all, the patient just gets a bit shorter and doesn’t notice,’ Cootes said.

According to the university, osteoporosis affects 1 in 2 women and 1 in 5 men over 50 and treatment of hip fractures will cost £2bn in the UK by 2020. Early stage diagnosis can, however, be problematic with clinicians occasionally struggling to reach consensus.

Cootes said: ‘Essentially, the clinicians would grade [vertebral fractures] by saying its either normal, deformed but not fractured, or it’s a grade 1,2 or 3 fracture and they…argue about what exactly is meant by grade 1,2, or 3.

‘By accurately fitting our statistical shape models of the outline of the vertebra we’re able to give them [clinicians] more detailed measurements, so rather than just having a grading… we can have a set of numbers [that] is more specific.’

The project’s core software is able to analyse an image by finding the outline of a shape very accurately and then quantifying subtle changes. Cootes believes the system – which will be integrated with radiography equipment used in hospitals – has the potential to analyse images held on centralised NHS hospital databases.  

‘People might have chest x-rays that include the spine but usually the radiologist isn’t looking at that, they’re looking at the lungs and they don’t necessarily pay as much attention to the spine,’ said Cootes. ‘So, potentially, there’s some evidence that things that are visible on the images are being missed because…the image was taken for another reason.’    

Cootes added that the algorithms being used are the result of more than two decades of incremental research and that they’re some of the most accurate in the world.

He said: ‘We’re applying similar algorithms to looking at x-rays of the knee to look for signs of osteoarthritis or looking at x-rays of the hip to look for changes in shape which are associated with osteoarthritis.’