recent projects‎ > ‎


Executive summary:

We collaborate with the VetSuisse faculty of the University of Bern since 2015. The collaboration has been supported by two grants from the Federal Food Safety and Veterinary Office (FSVO). The overall goal is the application of text mining technologies to pathology reports in order to detect monitoring signals that can be useful from an epidemiological perspective.

Project 1: 2015-2016
Funding: 213'408 CHF (total). Funding for our group: 88'000 CHF.
Principal investigator: Dr. Flavie Vial

Collaborators: Dr. Fabio Rinaldi,
Dr. John Berezowski

Project 2: 2017-2019
Funding: 269‘328 CHF (total). Funding for our group: 155'111 CHF.
Principal investigators: Dr. Corinne Gurtner

Collaborators: Dr. Fabio Rinaldi,
Dr. John Berezowski

The figures that follow illustrate schematically our approach.

1. Identify in the reports domain specific terms and organize them using lexical principles

2. Connect the identified terms to a reference domain ontology (UMLS)

3. Classify the reports (on the basis of the terminology primarily) into syndromic categories
4. Use the information from the reports to identify spatial and temporal patterns

Short Description Project 1 (2015-2016)

As recognized in the Swiss Animal Health Strategy 2010+, methods for early detection, based on the increasing abundance of data on animal health stored in national databases, can contribute to valuable and highly efficient surveillance activities. Post-mortem data, available from pathology services, are often under-exploited although they provide valuable information on the causes of death and additional health indicators for various animal species. In addition to their value for veterinarians and farmers (with regard to treatment and prevention options for the affected herd), systematic evaluation of post-mortem data could be of great value for nation-wide and international animal health and zoonotic disease early warning systems.

The academic pathology institutes of the Vetsuisse Faculty currently provide most of the veterinary pathology service offered to the food production sector in Switzerland. The overall goal of this project is to make post-mortem data, recorded at a veterinary pathology institute readily available for epidemiological surveillance and the early detection of emerging animal diseases. In particular, this project will focus on the development and evaluation of an automated text-mining and syndrome-classifying tool to 1) extract relevant information from pathology reports (written in free text) with minimal expert intervention; 2) classify pathology findings into syndromic groups to enhance the efficiency of health event detection.


Lenz Furrer, Susanne Küker, John Berezowski, Horst Posthaus, Flavie Vial, Fabio Rinaldi. Constructing a Syndromic Terminology Resource for Veterinary Text Mining. In: Proceedings of the 11th International Conference on Terminology and Artificial Intelligence, Granada, 4 November 2015 - 6 November 2015, 61-70.

Short Description Project 2 (2017-2019)

Early detection of animal disease epidemics is critical for animal and public health. To be effective, early detection surveillance should monitor data from many sources. Apart from clinical data, veterinary pathology records are an important source, but information is often in text documents and unsuitable for analyses. In a previous project funded by the FSVO a text mining tool was successfully developed to extract data from post-mortem reports from the University of Bern for surveillance. Clinical data from health services for livestock and breeder organizations are an additional source of data to be used. Due to the differences in nomenclature, these data sources cannot be combined. The purpose of this project is to: 1) Expand the text mining tool to analyze and combine both pathology and clinical data, 2) Develop tools to extract information from both sources and use spatial-temporal event detection algorithms to identify spatial clusters 3) develop an interface which enriches pathological reports with terminological information.