The project, for which h_da has been awarded funds of around €300,000 from the Hessian state government’s “Distr@l” programme, was initiated by Beate Egner. She is a vet and head of the Veterinary Academy of Higher Learning (VAHL) as well as managing director of the VBS VetVerlag, a publishing company in Babenhausen near Darmstadt. Together with Barbara Esteve Ratsch, a Spanish-Norwegian vet specialised in physical medicine, rehabilitation and sports medicine, she is contributing her practical experience as a vet.

For example, Esteve Ratsch presents the examination form so far commonly used for dogs, on which she jots down her notes. Her idea was not to formulate her findings in the usual way as text, but to plot them on a diagram of the animal’s body. This should speed up documentation. At the same time, it creates the basis for AI-supported diagnosis. “We don’t want to completely reinvent the wheel,” says Ute Trapp, professor for computer science at h_da and an expert in app development and user experience.

Digital management systems for veterinary practices already exist and are used for managing appointments, invoicing, administration of master data and text-based documentation of medical findings. The new app created by the h_da scientists can be integrated into one of these systems as an additional feature. The interdisciplinary research team wants to develop an app for tablets that helps with documentation and diagnosis in a simple yet accurate way. Moreover, the researchers are working on an interface “that can quickly retrieve the information already recorded and make it available when Fido or Kitty pay their next visit to the vet,” explains Professor Trapp. The system should also be able to display the findings in a visual format and as a chart. Elke Hergenröther, h_da expert for computer graphics and computer vision, has a vision: a click on the app could in the future be sufficient to reel off a film that shows the changes in the animal’s health and also suggests what it is suffering from. The team is convinced that “this could especially help young vets who have not yet gathered much practical experience.”

Professor Romana Piat explains that numerical and machine learning methods constitute the technical processes used to evaluate the data and findings. She specialises in numerical mathematics. “We are still at the beginning,” she says. Her team is currently developing the numerical algorithms required for the app. For this, the researchers above all need a vast amount of data. “Firstly from healthy animals so that we can detect disease-related deviations in the first place,” says Professor Piat. They are designing the app first of all for dogs.

For the app’s software, the medical findings and diagnoses of different breeds of dog must be recorded, as well as age, height, weight or even health information such as body temperature, blood counts or blood pressure. The AI-based system must be able to make clear distinctions because different pathologies can exhibit similar symptoms. “That is why we are starting with unambiguous cases,” say Elke Hergenröther. In animal orthopaedics, that means dislocated kneecaps, for example, “which apparently occur quite frequently in young dogs,” she explains. Together with the two vets, the scientists want to proceed step by step and work on detection and diagnosis – and in this way refine the system.

The app developed at the university is designed to be a learning system. Vets will be able to upload their experiences with it to a cloud service, where these will in turn be evaluated by means of an AI system and the feedback incorporated. According to Romana Piat, the intention is to further develop the app on a continuous basis.

Thanks to the funding from the State of Hesse’s “Distr@l” programme, the Faculty of Computer Science and the Faculty of Mathematics and Natural Sciences at h_da were able to create three half-time jobs. Three experts in numerics, artificial intelligence and app development are currently conducting research on other aspects that also have to be considered during the development work, such as security and data privacy issues. “The app,” says app expert Trapp, “should only need a few tweaks here and there for it to work on all the systems used in veterinary practices.”

However, the three researchers want to make one thing quite clear: it will not be an app that pet owners can download from the internet to make their own diagnosis of their four-legged friend’s health condition. “A thorough medical examination and professional interpretation of the results are indispensable,” emphasises Hergenröther. At the end of the day, evaluating the findings and making corresponding decisions are in the hands of the medical experts. And they, too, should first of all be instructed in how to operate and use the software properly via professional training and seminars. According to the scientists, the goal is to hand the app over in two years’ time to the partner companies participating in DDiT – and then roll it out swiftly.