Common disease osteoporosis: AI from Munich recognises changes in bone density fully automatically

Osteoporosis is a gradual disease. Bones lose density and stability over the years - often without any recognisable symptoms. The disease is only noticed when a fracture occurs, but by then it is often too late: lengthy hospital stays, restricted mobility and significant health complications are the result. The cost of osteoporotic fractures in the EU totals 56 billion euros per year. 

Utilising existing information

Osteoporosis screenings have not yet been established across the board. Although Dual Energy X-ray Absorptiometry (DXA) is a proven method for measuring bone density, it is rarely used. ‘At the same time, thousands of cost-intensive CT scans are carried out in routine clinical practice in Germany. They already contain valuable information about bone health, but this information remains completely unused,’ says clinical radiologist Dr Sebastian Rühling.  

Sebastian Rühling, Bonescreen © TU München

This is exactly where Bonescreen comes in. The start-up was founded in 2022 by an interdisciplinary research group at Klinikum rechts der Isar. The team of doctors and AI experts had a common goal: to make better use of existing medical image data in order to detect osteoporosis earlier. ‘Bonescreen was founded by healthcare professionals, including doctors, who experience demographic change and labour shortages on a daily basis and see that healthcare systems are under massive strain,’ explains Rühling, who is co-founder and Chief Medical Officer of Bonescreen.

Sustainable improvement in care

Thanks to two grants from the European Research Council, the team was able to show that its AI-based algorithms enable more precise and automated bone density measurements. Rühling: ‘We have demonstrated with proof-of-concept studies that our cost-effective method has the potential to sustainably improve current standard care in Germany.’

This not only convinced the scientific community, but also the first clinics and radiology practices. In order to bring the system into practice, Bonescreen was spun off as a company. Other financial backers such as Bayern Innovativ and EIT Health joined in. ‘We receive non-remunerated support primarily from our mentors and key opinion leaders in our radiology network, but also from people we have never met before, who recognise the clinical added value and help us to test the product in various clinics,’ says Rühling.

Heatmap makes minimal differences visible

At the heart of Bonescreen is SpineQ - software that analyses CT scans fully automatically, without the need for additional examinations or interventions. Whereas radiologists previously had to laboriously search for fractures or density changes manually, SpineQ now takes over this task. It translates bone density values that are invisible to the human eye into a colour-coded heat map. This makes minimal differences visible. The system also recognises unnoticed fractures and provides an objective assessment of bone health.

For an AI model to make reliable predictions, it must be trained with large amounts of high-quality data. The Bonescreen team uses clinical data sets from various hospitals. All training images were annotated by medical experts so that the model not only learns, but also continuously improves. A decisive advantage: SpineQ works independently of the CT scanner or protocol used. The software adapts without the radiologists having to change their usual procedures.

Fully automated instead of semi-manual procedure

According to its developers, Bonescreen's AI solution for fracture detection differs from others in several unique selling points: The software feeds the results directly into the hospital's Picture Archiving and Communication System (PACS), meaning that doctors do not have to operate additional programmes. SpineQ also offers fully automated analysis, whereas other providers work with semi-manual processes that require radiologists to manually adjust measured values. ‘With many AI solutions, they add another layer of complexity. These are often things that tend to hinder the workflow. As we offer fully automated analysis for seamless radiological reporting, we avoid this,’ explains Rühling.

Another unique selling point is the ability to reliably analyse CT scans with contrast agents - an area in which other methods have so far reached their limits. Bonescreen also wants to set itself apart from the competition when it comes to data protection: All data remains within the clinic or practice; cloud storage is not required, which means that the highest data protection standards are met.

Market authorisation is imminent

The path from research to practice is often a rocky one - as the Bonescreen team knows. However, the company is well on the way to launching its technology on the market across the board. ‘We are starting with a phase in which radiologists can test us. They are not yet allowed to tell patients what has come out. This will be activated as soon as CE certification is obtained,’ reports Rühling. He expects market authorisation in spring 2025.

The company's business model envisages offering SpineQ both directly to clinics and via sales partners. Cooperation with large radiological practice chains, which can implement the system in several locations at once, is particularly promising. International partnerships, such as with the Erasmus Medical Centre in Rotterdam, are intended to facilitate market entry outside Germany.

But the company is already thinking ahead. The technology is to be extended to other clinical pictures, including the AI-supported analysis of MRI images.

Aiming for reimbursement by statutory health insurance funds

In the long term, the start-up wants statutory health insurance funds to recognise AI-supported bone density measurement as a reimbursable service. At present, the examination can only be offered as an individual healthcare service (IGeL), but the aim is for it to be reimbursed by the statutory insurances. Rühling: ‘Our focus is on continuously increasing the benefits for users, patients and society and we are working together with statutory health insurances and patient associations to establish Bonescreen as part of preventive healthcare.’

Bonescreen shows how artificial intelligence can help to combat one of the most underestimated widespread diseases more effectively.  The start-up shows that the future of osteoporosis diagnostics does not lie in new, expensive tests, but in the intelligent use of existing data.

Sebastian Rühling once again: ‘In the long term, we want to expand our product in preventive diagnostics. I believe there are still many ideas for identifying other diseases with risks from existing image data at an early stage, in addition to bone health. And when I look at our five-year plan, I see us as one of the providers of evidence-based AI software solutions in the radiology sector that really makes an active contribution to prevention and early detection.’

Group photo of the founders (from left): Giles Tetteh, Jan Kirschke, Anjany Sekuboyina, Sebastian Rühling, Malek El Husseini, Dominik Maurer

Customised formats for the medical technology sector

MedtecLIVE is the leading trade fair in Europe for the development and manufacture of medical technology. It is part of a Brand Family that offers the industry more than just a trade fair and supports it in advancing medical technology with its wide range of products and services as well as its strong network and partners. This creates several platforms at Stuttgart, Nuremberg and Munich, where the most diverse players in the industry can meet their target group in a customised way.