Artificial intelligence learned diagnosis of heart noise in dogs with the accuracy of an experienced doctor. The team of researchers from the University of Cambridge has developed a machine learning algorithm that reveals pathologies with a sensitivity of 90%.
According to the research manager, Dr. Andrew McDonald from Cambridge, if cardiovascular diseases are a serious problem for people, then the situation is even more critical for dogs.
Statistics shows an alarming picture: every thirtieth dog, which comes to an appointment with a veterinarian, suffers from heart noise. Especially often pathology is found in small breeds and elderly animals. Cardiac noises serve as the main indicator of the mitral valve disease – the most common heart disease in adult dogs.
Professor Jose Novom Matos from the faculty of veterinary medicine Cambridge notes: determining the degree of heart flare and The appointment of treatment requires great experience. Often, owners have to contact narrow specialists-cardiologists and pay for expensive surveys.
To create an algorithm, researchers have collected the world’s largest cordial sound data database. Almost 800 animals were examined in four specialized veterinary centers of Great Britain. Each patient conducted a complete physical examination and scanning of the heart, recorded noises using an electronic stethoscope.
It is interesting that the algorithm was originally developed to diagnose diseases in people based on a database that includes data about a thousand patients. Professor Anurag Agarval, who headed the study, explains: mammal hearts are quite similar, and their pathologies develop according to similar mechanisms.
The team adapted the technology for working with dogs, teaching the algorithm not only to reveal noise, but also to determine their degree. The system is able to distinguish between mild violations and signs of serious pathologies.
Unlike human mitral failure, which can only be treated with surgically, there are effective medicines for dogs. However, the success of treatment directly depends on the timeliness of its beginning. Despite the fact that heart pathologies are more common in small breeds, dogs of all sizes and ages were included in the study.
Another important plus – the technology will be available even for small veterinary clinics. The use of a digital stethoscope in combination with the machine learning algorithm allows for high-quality diagnostics without expensive equipment.