54 Image Based Sampling to Assess Horn Fly Abundance on Cattle

Abstract

Abstract Horn flies are a major irritant to livestock. In cattle, horn flies can cause increased heart and respiratory rates, decreased feed efficiency, decreased weaning weights, and reduced milk production, resulting in substantial economic loss. Fly abundance varies within and across breeds, and genetic analyses have shown sufficient variation to permit selection of hosts for resistance to horn flies. A major limitation for selecting against horn fly abundance is the difficulty associated with measuring on-animal fly counts under pasture conditions. Traditionally, trained evaluators examine cattle in the pasture and estimate fly numbers. Fly movement and animal movements, as well as evaluator expertise are major sources of variation. Image analysis has also been proposed as a means to evaluate horn fly abundance. However, several issues related to image acquisition and processing need to be resolved before reliable estimates can be obtained. Cows (n = 752) and heifers (n = 129) were assessed by trained evaluators and several digital images were taken for each animal twice during the summers of 2019 and 2022. For each image, a box that runs roughly from the withers to the hooks and from the chest floor to the fore of the udder was established. Using ImageJ software, all horn flies within the box were manually counted. Additionally, a density region-based method was used. It consisted of dividing the box into regions of high, moderate, and low fly density. A grid was placed over the box and 20%, 10%, and 5% of squares were sampled randomly within the high, medium, and low-density regions of the grid, respectively. The total counts were used to assess the accuracy of the density region method and the subjective assessment by trained evaluators method. The Pearson correlation between fly counts assessed based on and the methods was 0.954. A lower correlation was obtained between subjective assessment and (0.403) and (0.432). When fly abundance was discretized into 4 classes, the image-based methods showed high concordance (~0.90). However, the subjective method yielded only a 37% concordance. These results seemed to indicate that the proposed collection can be used with high accuracy to estimate the total number of flies on the animal. It significantly reduced the amount of time it takes to estimate fly counts from images. Furthermore, it has a greater accuracy than subjective evaluations. Despite its good performance, the proposed method still does not overcome the difficulties associated with capturing images under pasture settings. Alternative approaches that can cheaply and efficiently estimate fly abundance should be explored.