CEPEF-HARAT

The CEPEF-HARAT project is an ongoing doctoral research initiative conducted at the University of Zurich in collaboration with the Swiss Data Science Center (SDSC) and the international CEPEF consortium. The goal is to develop the first machine learning–based tool to predict mortality risk in equine anaesthesia, helping veterinarians identify high-risk patients before surgery and reduce anaesthesia-related fatalities.

The project builds on the CEPEF4 dataset, comprising more than 48 000 anaesthesia cases collected from 94 centres worldwide — the most comprehensive dataset of its kind. My current work focuses on data cleaning, preprocessing, and exploratory data analysis, preparing this large and heterogeneous dataset for machine learning applications.

Together with the SDSC, I’m working on the integration of Natural Language Processing (NLP) and Large Language Models (LLMs) to structure free-text fields such as “breed,” “reason for anaesthesia,” and “comments,” allowing important clinical details to be incorporated into predictive modelling.

Future steps include feature engineering, model development, and validation using explainable AI techniques like SHAP and LIME, with the long-term goal of creating a transparent, data-driven risk assessment tool (HARAT) that can support clinical decision-making and improve patient safety in equine anaesthesia.

This project is a collaboration with Prof. Dr. Regula Bettschart-Wolfensberger, Head of the Department of Anaesthesiology of the University of Zurich Veterinary Hospital (Tierspital Zürich) and Guillaume Obozinski from the Swiss Data Science Center.