Artificial intelligence has been quite a controversial topic lately. Not the AI that powers the movements of NPCs in games, but the elements behind LLM that power chatbots, create generative images, and create websites. While concerns remain, a team of researchers at the University of Cambridge have shown they can diagnose heart murmurs in dogs, and I think it’s a pretty good idea.
As shared in Nvidia Developer Blogthe study used models that “were trained using PyTorch and Nvidia CUDA on Nvidia GeForce 10-series GPUs, which enabled efficient data processing.” A machine learning algorithm “listened” to digital heartbeat data and diagnosed heart murmurs and heart disease based on symptoms.
At any murmur grade (i.e., any intensity), the algorithm was able to detect murmurs at a rate of 87.9%, which corresponds exactly to the cardiologist’s assessment in 57% of the recordings.
as test states, “the model is a promising tool for enabling accurate and affordable screening in primary care.”
This is a great potential apply case for AI because for the pet owner, scanning the pet will not only be cheaper, but may be more proficient and accessible. Instead of booking a vet appointment, you can have someone trained to apply the device who will simply scan your dog and let you move on. This can free up time for skilled veterinarians to work on various procedures.
There is, however, one concern that is worth examining here and that researchers are likely aware of when entering this study. Just a few months ago, a study found that up to 20% of local doctors in the UK could be using generative AI tools.
Using artificial intelligence to create calendars and even draft letters is not a problem in itself, but it is vital to remember that artificial intelligence cannot reason like humans. It can approximate reasoning by synthesizing information, but it requires a human to overcome the final hurdle of understanding.
He doesn’t actually see or hear patients’ concerns or feelings and doesn’t diagnose problems legally. He can compare potential symptoms, connect them to potential causes, and that’s it.
Artificial intelligence, such as the one that can be used to diagnose patients, must collect a lot of information, both for scientific purposes and to understand the language that patients speak. However, with so much data, LLMs are susceptible to “hallucinations” where bad or false information is treated as if it were true.
Although they may be more right, it takes a human being to see this hallucination in the first place. Just last month, a counterfeit artificial intelligence website tricked thousands of Halloween attendees into attending the counterfeit event.
For this reason, artificial intelligence is not used to diagnose patients, but can only act as a tool to support the diagnosis procedure. In this case, when it comes to auscultating potential heart murmurs in dogs, it appears to have a high degree of accuracy, which may be partly attributable to the fact that the tool is built to make a very specific diagnosis, but requires a human to check and suggest treatment .
If used responsibly, it could be a great tool for both vets and dogs, and one of the best uses of artificial intelligence I’ve seen to date.
