Title: Prediction of structural epilepsy in dogs

Project duration: 2021 – 2025

Research Area: Life Science and Medicine, Veterinary Medicine

Identifying the type of epilepsy in dogs is a complex task, even for experienced veterinarians. Diagnoses are often based on clinical intuition rather than objective criteria. Although numerous parameters (~75 features) are collected during examination, it remains unclear which of these are truly relevant for diagnosis.

This project aims to use machine learning to identify the most informative features and accurately classify different epilepsy types. By providing probabilistic predictions, the system is intended to support clinicians with objective, data-driven decision-making.

Aims

• Identify relevant clinical features in dogs with seizures
• Classify epilepsy types based on these features
• Provide reproducible, objective predictions
• Support clinical decision-making
• Enable integration into routine veterinary practice

The overarching goal of this study is to apply machine learning to predict the risk of structural epilepsy in dogs presenting with seizures.

Problem Statement

Clinical reasoning in veterinary medicine often relies on the clinician’s personal experience and generalizations from published studies on patient cohorts. However, scientific approaches that enable individualized, data-driven decision-making are still underutilized.

This gap is particularly evident in the prediction of the underlying cause of seizures in individual dogs. There is a clear need for tools that allow more deterministic and reproducible diagnoses.

Technology

• Data Preprocessing
• Feature Selection
• Random Forest Classification
• Bayesian Networks

Outlook

To ensure generalizability and clinical relevance, further validation of the model using data from external veterinary clinics is planned. Future research may expand to additional types of neurological disorders and explore integration with imaging data (e.g., MRI or CT scans) for improved diagnostic accuracy.

• A Living Lab Demonstrator in the form of a web application is being developed to showcase the approach in a practical, interactive format.

Publications

• Flegel T, Neumann A, Holst A-L, Kretzschmann O, Loderstedt S, Tästensen C, Gutmann S, Dietzel J, Becker LF, Kalliwoda T, Weiß V, Kowarik M, Böttcher IC and Martin C (2024) Machine learning algorithms predict canine structural epilepsy with high accuracy. Front. Vet. Sci. 11:1406107.
  • doi: 10.3389/fvets.2024.1406107

Team

Lead

• Dr. Christian Martin (Database Group Leipzig – ScaDS.AI Dresden/Leipzig)

Team Members

• Anja Neumann (Applied AI and Big Data – ScaDS.AI Dresden/Leipzig)
• Prof. Thomas Flegel (Clinic for small animals – Leipzig University)

Funding

• Federal Ministry of Research, Technology and Space; Free State of Saxony