Mr Raniere Gaia Costa Da Silva1, Dr Ambika Prasad Mishra, Dr Christopher Riggs, Dr Michael Doube
1City University Of Hong Kong, ,
Thoroughbred breeding industry in Australia is responsible for generating $121.2 million in exports and makes little use of deep learning to detects abnormalities from radiographs and estimates risk to horses. This study look at the viability of deep learning to classify equine radiographs and improve racehorses welfare.
To assess the capability of deep convolutional neural networks to classify anatomical location and projection from a series of 48 standard views of racehorse limbs, we used 9504 equine pre-import radiographs to train, validate, and test six deep learning architectures available as part of the open source machine learning framework PyTorch.
The best architecture was ResNet-34 that achieved a top-1 accuracy of 0.8408 and the majority (88%) of misclassification was because of wrong laterality. Class activation maps indicated that joint morphology drove the model decision.
Deep learning has a large potential to be incorporated into tools that veterinary surgeons can use to formulate risk categories that will help prevent racehorses injury, increase racehorses welfare, and improve the public perception of racing.
Biography:
Raniere Silva is a PhD candidate at City University of Hong Kong working on Deep Learning applied to medical imaging. Before, he worked as Community Officer for the Software Sustainability in the UK. Raniere has a bachelor in Applied Mathematics from the University of Campinas, Brazil.