Classification of the long bones and fractures of dogs

One of the most popular topics of recent years, deep learning algorithms find wide application in the biomedical field. Disease and fracture detection, biological data estimation, tissue and organ segmentation, missing data completion, and many more applications can be successfully performed using images obtained by various imaging methods, thanks to deep learning algorithms. In particular, fracture detection is one of the most popular subjects in this field. Nevertheless, most of these applications are made for human medicine. Therefore, veterinary medicine have been ignored drastically. The realization of the deficiency in this area has been the biggest motivation for the thesis. In the thesis, a comprehensive data set containing X-ray images of dogs gathered from Ankara Metropolitan Municipality Stray Animals Temporary Care Home is created. It is aimed to determine the type of long bone from the images, then detect the presence of the fracture and, if present, classify the fracture type. According to X-ray images, the determination of the fracture time and the maturity of the dog are other studies carried out within the scope of the thesis. Like many studies in the field of the biomedical image processing, different deep learning architectures are compared, and the results are tried to be optimized.