LETI: LETI Researchers Taught AI to Detect the Hyperactivity Degree in Dogs

LETI researchers, together with scientists from the University of Haifa (Israel) and Newcastle University (UK), have developed a machine-learning model that helps determine the degree of ADHD in dogs by analyzing their behavior recorded on video.

ADHD is a neurodevelopmental disorder that can appear in humans and animals from an early age. It is characterized by inattention, hyperactivity, and impulsivity. Although today ADHD is treatable and in many countries of the world is not considered a disease, people and animals affected by the syndrome can be dangerous to themselves and others.

Previously, LETI researchers, together with researchers from Israel, the UK, and France, trained a neural network model to determine whether or not dogs have ADHD using video recordings. The new algorithm allows for a more accurate assessment of an animal condition. The findings are published in the Animals journal.

“We have come up with a new method for classifying hyperactivity conditions in dogs based on video analysis. Now our machine-learning-based model can not only detect ADHD but also determine the degree of its manifestation with high accuracy. This is important for more effective therapy,” says Dmitry Kaplun, the project manager, Associate Professor of the Automation and Control Processes Department at LETI

The researchers used videos from veterinary clinics recording the activity of dogs with ADHD to train the neural network model. The experiments took place in a room specially marked with a coordinate system. This automatically built up the trajectories of the dogs’ movements in various situations. Then, based on the statistics collected, the artificial intelligence detected patterns and drew conclusions about the degree of hyperactivity of the animal.

“Our solution can be used in telemedicine, which is actively developing, including veterinary medicine. And here, the main task of the developed system is to help the doctor diagnose the degree of ADHD faster and more accurately. The accuracy of determining the severity of the syndrome with our model is about 81%.”

Dmitry Kaplun, Associate Professor of the Automation and Control Processes Department at LETI
The researchers are now working to improve the accuracy of diagnosing ADHD conditions and the ability to detect its symptoms at an early stage. One of their ultimate goals is to use the experience to treat people.