Analyse and visualise medical image data more easily and securely with new technology

A lot of data is generated in medicine, for example during computer tomography. This data is important on the road to personalised medicine. Artificial intelligence methods, such as machine learning, learn from them and help to personalise diagnoses and therapies in the future. However, such techniques are still fraught with uncertainty. A team of researchers from Kaiserslautern and Leipzig is working on a system that automatically analyses and visualises medical data, including its uncertainties.

The team will be presenting its technology at the Medica medical technology trade fair in Düsseldorf from 13 to 16 November at the Rhineland-Palatinate research stand (Stand E80, Hall 3).

A stroke requires haste. With the help of computer tomography (CT) images, doctors can quickly determine where in the brain a blood clot has formed and what treatment is appropriate. Such imaging procedures play an important role in medicine. They are also used in other areas, for example before operations. For example, images from magnetic resonance imaging (MRI) help surgeons to plan an operation before surgery.

A common feature of all these technologies is that they generate a lot of data. „Analysing and visualising it automatically is an important step on the way to personalised medicine,“ says Dr Christina Gillmann, computer scientist at the University of Leipzig. „This area has gained enormously in importance in recent years.“ This is made possible by AI methods such as machine learning and neural networks. They learn on the basis of data with which they are trained or fed. For example, from CT image data that a doctor has previously processed. In this way, technical information as well as medical experience is incorporated. The more data these processes can analyse, the better the results will be.

In a few years, such technologies could be used in everyday clinical practice to enable personalised diagnoses and therapies, for example. However, they are still in their infancy. „Each medical case has to be trained individually. The data has to be processed individually in advance, which is very time-consuming," says Robin Maack from the Computer Graphics and Human Computer Interaction research group at the Rhineland-Palatinate University of Technology (RPTU) in Kaiserslautern. For each medical case, for example, doctors have to individually „label“ the data. This means, for example, that if such a system is to be trained to automatically recognise a tumour, hundreds of images with known tumours have to be drawn in by hand so that the neural network has a basis on which it can learn,

explains Gillmann.

Maack continues: „In addition, there are no standardised interfaces with which trained networks can be handled, loaded and used. But there are also no standardised guidelines on how medical professionals should deal with uncertainties in the data situation, be it in training data sets or in the models used.

Such uncertainties arise, for example, in the case of lesions. These are certain areas of the brain that are no longer supplied with sufficient oxygen or no oxygen at all due to the blockage of vessels in the brain in the event of a stroke. They are no longer able to function. The core of the lesion is often easy to recognise, but there is usually no clear demarcation at the edge and regions where even doctors cannot agree on whether they should be classified as a lesion or not. Ultimately, it takes medical experience to know how to deal with this.

This is where the work of Gillmann and Maack comes in. Their team is currently developing a standardised system for processing and evaluating medical image data and visualising its uncertainties. It goes by the name of GUARDIAN, which translates as „Hüter“. The researchers have designed their technology to be easy to use. For example, clinics can load their trained neural networks here and combine them with the processed data provided, for example when recording a stroke.

The system evaluates the data and visualises the results. „This happens automatically, without the need for IT knowledge“, continues Maack. Our method also shows the uncertainties, which means that the doctors can look at them again and, if necessary, make a joint decision on what is the best treatment in individual cases, for example.

The two computer scientists will be presenting their technology at the trade fair. The system is freely available as an open source application.

The „Visualisation and Human Computer Interaction“ working group led by Professor Dr Christoph Garth at RPTU in Kaiserslautern has long been researching how to process data from imaging procedures for medical use in such a way that it can be used easily and reliably in everyday clinical practice.

The appearance of the researchers from RPTU Kaiserslautern-Landau at Medica is being organised by Klaus Dosch from the Department of Transfer, Innovation and Sustainability. He is the contact person for companies and, among other things, arranges contacts with the scientific community.

Article from the "innovations report" from 6 November 2023