Thesis subject

Digital Diagnostic Tool For Foot and Mouth Disease

Level:

Research area/discipline: Software Engineering

Prerequisites: Software Engineering (INF-32306), Programming in Python (INF-22306)

Short description:

Foot and Mouth Disease (FMD) is a contagious animal disease which affect cloven-hoofed animals, including domestic cattle and infections with this disease can result in large economic and animal welfare consequences. Therefore FMD is a notifiable disease and to prevent major epidemics, detection of infection in a farm should be as early as possible, ideally when the first animal is infected. For detection to happen, farmers should notify any suspicion of infection to the competent authorities. In practice, suspicions of infection are raced by the farmers observing animals with clinical signs similar to those caused by FMD. Often, because of the large number of animals in a farm, farmers would detect these clinical signs only when a group of animals have already been infected, resulting therefore in delayed detection and increased risk of large epidemics.

Machine vision offers the potential to assist farmers in the fast and effective detection of diseases, hence automatic detection of FMD symptoms by means of e.g. face recognition, is a possibility which could be investigated. A digital diagnostic tool for FMD should be able to discriminate between faces of cows which are truly infected (test sensitivity) with FMD from those which are not (test specificity). Furthermore, such a digital tool ideally should not only discriminate healthy cows but also cows which are not infected with FMD but still appear to show some clinical signs similar to those caused by FMD (analytical specificity). Such signs can be caused by diseases such as infectious bovine rhinotracheitis (IBR) and bovine respiratory syncytial virus (BRSV).

Images of FMD infected and healthy cow faces are available on the internet and a small collection of images is also available at WBVR. With these face images a deep learning network could be trained to identify both FMD infected cows with a defined sensitivity and non-FMD infected cows with defined specificity. Such trained network will be the core of a digital diagnostic test, which once shown to be accurate would lead to the ultimate goal of implementing such a digital diagnostic test in farm settings to assist farmers in the timely and accurate detection of diseases.

To start the development and validation of image diagnostic algorithms, first images of faces of cows with FMD, IBR, BRSV, parainfluenza 3, other respiratory diseases and healthy cows will be collected from the internet and the WBVR repository and stored in a database. This database should be shared with scientist of WBVR and WU. These images will then be used to develop and validate the diagnostic algorithm.

For more information:

cagatay.catal@wur.nl

bedir.tekinerdogan@wur.nl