DYS4VET Project: Diagnosing Canine Hip Dysplasia Through Artificial Intelligence
Pets are increasingly part of the lives of families, and in that sense, they are wanted to be healthy. Canine hip dysplasia (HP) is a disease that affects the bone/cartilaginous structures and brings suffering and discomfort to dogs. The Dys4Vet project falls within the scope of veterinary medicine, with the main objective of supporting the diagnosis of canine HP through Artificial Intelligence (AI) applied to radiographic images, and consequently improving the lives and health of dogs.
The mentioned dysplasia represents a considerable economic impact, as it is an orthopedic disease that is common in both dogs and humans. The proposed solution, based on AI, will determine quantitative parameters of the hip joint, by evaluating orthopedic x-ray images, in a simple, fast way and with high levels of precision. It is intended to streamline and automate the process of analyzing radiographs and create a mechanism to support clinical decisions. Carrying out the respective classification and certification of the disease.
In this context, Neadvance and partners devised a solution called Dys4Vet, based on intelligent image analysis in order to optimize the diagnosis of HP. This project, with the acronym Dys4Vet (in allusion to the disease), consists of an artificial intelligence platform (with the same name), developed in .NET Core and Javascript. Over the last 12 months, a web platform was developed for Veterinarians to annotate radiographs, in DICOM format, in a more intuitive, quick, and easy way. This platform will also serve to obtain the best practices and know-how from experts with the creation of datasets to be used in learning deep learning models.
The project is carried out in conjunction with the University of Trás-os-Montes and Alto Douro (UTAD) and the Hospital Veterinário Montenegro (HVM), having a multidisciplinary team, in particular a specialist from Denmark, who is also a professor at a prestigious Danish University. This specialist, who is a reference with extensive experience in this subject, is currently an active member of the project and has already made available to Dys4Vet a very rich database with almost 6,000 X-ray images of dogs. Having served as the basis for the start of developments.
After the first phase of the project, which concentrated efforts on creating a synergy between partners, both in the planning/execution of a working model and in the development of the annotation tool. To date, it can be concluded that one of the main challenges is to expand the use of the web application to increase the number of annotations. This will allow the creation of an increasingly reliable dataset, containing maximum variability, in order to make the AI models robust and leverage platform expansion.
One of the main objectives of the platform is to offer Veterinarians a differentiated experience, through an innovative tool to support the diagnosis of canine hip dysplasia. Thus, it is able to load images and recognize bone structures to perform medical-veterinary index calculations automatically with a high degree of reliability and objectivity.
It is noteworthy that in recent months the project has advanced as planned, having as some of the practical results the preparation of scientific articles for participation in international conferences in the animal/canine area. Recently, the platform was invited to the HVM International Congress, in Santa Maria da Feira, Portugal, in the 2nd fortnight of October, where Neadvance will perform a demonstration of a version of Dys4Vet already has an automatic calculation of metrics that reinforce the diagnosis and prognosis/early detection of canine dysplasia, which is, therefore, a promise to improve life and animal welfare.
Article by Neadvance’s Research, Development & Innovation Team