This week we will continue to look at imaging in veterinary medicine. We have covered ultrasound, CT scanning and MRI scans. Most readers are familiar with the older imaging technology – x-ray imaging, which has been around for about a century and is available in most veterinary practices. But now we will turn to the interpretation of the images generated by these various modalities. There are some really exciting changes afoot in this specialism. Traditionally, the images of your pet were taken via x-ray, ultrasound, CT or MRI and then reviewed by a qualified person – the vet. The veterinary surgeon spent years training at university, then years in practice to acquire the experience to interpret these images. Then they would combine that interpretation with other information about your pet – history, clinical signs etc, to arrive at a diagnosis. In the past few years though computers have started to be used in the interpretation of these images, in the form of artificial intelligence (AI).
Artificial intelligence means using computers to perform tasks which mimic human intelligence. In the case of interpreting images, we are referring to “computer vision”. The use of a computer to process images in a manner similar to the way people view objects and apply reasoning to them to interpret what they see. The computer will perform a sort of “pattern recognition” to the images. It will assess data points on an image, and compare them with a large reference library of “normal” and “abnormal” images. In the case of veterinary images, the computer will assess whether an image is normal or showing signs of a disease process. This is similar to the way in which vets already assess images – we have an idea of what a healthy animal looks like on an x-ray, and we compare our idea of the healthy animal with what the x-ray is showing us. If there is a difference to our own “reference library”, then we flag it up as perhaps showing signs of disease.
I’m sure the reader can appreciate the advantages of using computers in pattern recognition. Modern computer power, combined with the internet, means that a simple PC or phone can access millions of anonymised medical images to use as a reference library when investigating a disease process. That will be infinitely more x-ray images than even the most experienced medical practitioner could hope to assess in their entire working lifetime. Furthermore, the more images reviewed by the AI, the “smarter” it gets. The system will learn, through human instruction, what data points are correlated with a “correct” diagnosis. Another advantage is speed – the computer can read many more images than a human can in the same time period. So the aim of AI in veterinary imaging is to provide quick, accurate, and potentially cheaper diagnoses.
So, will AI replace human intelligence in veterinary medicine? No, not yet, and not for a long time. At the moment, AI is a useful additional tool in aiding diagnosis. AI can be an aid in decision making, providing probabilities of results, rather than exact findings. It is like having an additional clinician offering their opinion on the interpretation of images. At the end of the day, there is a clinician combining case history, test results and examination findings to form a diagnosis and treatment plan. We, at Downe Vets, have been using AI as an aid to our diagnosis of conditions for the past year. We have found it to be a useful tool in getting the right diagnosis for your pet, and thus improving patient care. It will certainly have a place at the cutting edge of clinical practice.