Vetology's AI Guardian Software
As medical equipment and technology has improved over time, the performance and results that are available to veterinarians has helped significantly to improve the quality of patient care. Almost everyone in the veterinary industry has benefited from these technological advancements: pet owners, patients, DVMs, clinic staff, practice managers, and practice owners.
However, there is a group of care providers that have received a bit of a burden along with the advancements. Modern equipment has provided better quality images and improved diagnosis. Pet owners along with their furry companions are grateful, but the steady - and lately sharp - increase in the number of digital images has created a significant demand for expertise and reports from radiologists. We rely upon radiologists for interpretation, guidance, and direction, so clinics can continue to deliver the best quality care.
All of this has created a level of demand that will soon far exceed the availability of radiologists, so we decided to do something to proactively address a few issues we anticipated years ago.
In the beginning, we recognized that we needed a way to harness the radiologist's knowledge and expertise while helping scale the volume of images they could interpret in a work day. This was essential if the industry was going to keep pace with the projected rise in demand for image interpretations. In Phase 1 we created the first veterinary image database that was open for keyword searching by anyone, and it was always planned to be free to use. We trained this artificial intelligence (AI) to find similar image matches based on keyword searching and visual characteristics found in the source image. Keyword searching is still available today at no charge on our web site here.
As a Vetology® AI monthly subscriber, you could upload your own radiograph images, enter keywords, and the system would show you computer-matched cases with accompanying radiologist reports from our database. The initial feedback was fantastic.
Soon after we released our image match solution in 2017, we heard from radiologists and clinics that it would be nice if they could receive a preliminary report showing whether there were any abnormalities in their source images. This led our team deeper into the world of augmented and artificial intelligence and our Vetology AI Guardian® service was born.
With the rise of machine learning and faster processors, we started to train the AI to actually interpret the radiographic images. This is when things started to get exciting. With machine learning and neural network technology, we learned that we could train the AI to identify and report abnormalities in the images, just like radiologists would. Using AI technology, we completed algorithms to analyze canine thorax, heart, and lungs. Canine abdomen and feline algorithms are in development and will be released soon.
At first, we tracked how effective the AI was when compared to radiologists. We asked questions like, "In what percent of cases did the Vetology AI Guardian software report match what the radiologist wrote?" Many clinics and hospital groups were also interested in the answer to that question. Most appeared to be interested in AI as compared with the standard of the radiologist. Over time we learned that we should be asking a different question. Since radiologists are not perfect, they sometimes miss things. Human radiologists get tired after long work days, so we started thinking, "How does AI perform in patients with clinical signs?"
The more we trained the AI and honed our process, the better our results became. We strove to ask better questions and in the end, we got better results because of it. Now, we deliver an AI radiology report in under 5 minutes that serves clinic owners, DVMs, and radiologists well, providing a 'second set of eyes' to support decision-making, and to keep pet care moving forward in the clinic's day-to-day operations. Many of our users now tell us they can't imagine working without Vetology AI Guardian software.
Today, when you consider the differences in digital imaging systems and the effect that the specific environment has on an animal, it is clear to us that there can be no 1-size-fits-all approach when it comes to delivering the very best image interpretations and AI reports. Each digital system has its own unique image processing which creates an image appearance that is unique to that particular system. That uniqueness between digital x-ray systems means they produce objectively different nuanced attributes and 'looks' in the images they produce, and we believe that means the AI evaluations should certainly take this into account.
Consider an example of two similar animals living in two parts of a country with different climates. Animals living in hot and dusty climates have more bronchial markings without clinical signs than animals who live in cooler, higher elevation and forested locations. AI should account for those variables and adjust the interpretations accordingly.
This is the level of sophistication we are busy creating today. This is the next-generation of Vetology AI Guardian software capability we are producing, and we believe that veterinarians should expect this higher level of customized AI service.