+49 30 838 62618
The overall objectives of this thesis were to develop a specific training protocol for training dogs to identify cows in estrus under practical conditions and to identify steps in the procedure of scent detection dog training and testing that can eventually bias the study outcome.
My first study revealed multiple differences in methodology and a high variability of the results of 14 studies on training and testing scent detection dogs recruited in a systematic literature research. For evaluation of the studies a check list detailing relevant information about the study design and the training and testing process was used. For the experimental part of this study seven dogs were trained to indicate black tea as target scent in two different testing systems, a testing platform and a scent detection board, and against two different types of negative samples, tap water and smelling distractions. We could achieve a high sensitivity (92.1%) and specificity (97.4%). There was no difference of performance in the two testing systems. Quality of negative samples had no influence on the results.
In the second study six dogs were trained for heat detection in dairy cows. A specific training protocol was developed for training dogs to identify cows in estrus from the feed alley. Four of those dogs participated in the final test after an average training time of 50 h per dog. Finally the dogs overall correctly identified positive cows as being positive in 23 out of 32 cases (i.e. sensitivity of 71.9%) and falsely classified positive cows as being negative in nine cases (28.1% type II errors). Out of 128 cases 119 were correctly classified as true negative cows (i.e. specificity of 93.0%) and in nine cases negative cows were falsely identified as positive cows (7.0% type I errors).
To process lessons we learned from the first and the second study, we wanted to identify factors in the procedure of dog training and testing that can eventually bias the study outcome. Therefore we performed a systematic literature research and systematic assessment. The objective was to evaluate publications on scent dog studies. Finally I wanted to summarize this information to recommend a best practice standard for scent dog studies. A total of 54 studies could be included for final evaluation. We were able to identify the target odour, the type of scent detection task and experimental set-up, the samples used for training and testing, the test design, the dog trainer and training procedure chosen and the used dog breeds as factors potentially influencing the outcome of scent detection studies with trained dogs. With these findings a best practice standard was recommended.
We were able to show that trained dogs can detect cows in estrus on farm with better sensitivity and specificity than is often obtained by conventional heat detection methods. Further research is warranted to develop an optimized training plan that allows training estrus detection dogs for practical use within an appropriate time and to assure a certain sensitivity and specificity in working estrus detection dogs over time. The use of dogs as a diagnostic tool for medical and biological applications is becoming increasingly important. It is mandatory, however, to follow standardized test protocols for scent detection dogs to allow valid comparisons of results. We recommend the application of the best practice standards to reduce possible bias in training and testing scent detection dogs as far as possible.