In 2017 a new technological advancement in high-throughput CBCs was introduced for human hematology. The innovation was centered around a shift from flow cytometry to digital imaging of cells.
While the new technology opened major opportunities for innovations, to date the technology has not been adopted by veterinary medicine. This article explores some of the current and future possibilities made possible by Moichor’s imaging-based CBC technology in conjunction with the company’s application of artificial intelligence.
“This is the forefront; it’s where hematology is headed,” Moichor Director of Pathology Kyle Webb, DVM, DACVP, said. “It's already being used in the human field but in a slightly different context.
We have an opportunity in the veterinary field, not only to make this shift to the cutting edge in technology, but also, to drive the applications for that technology in ways that improve outcomes for veterinary patients,” Dr. Webb said.
Dr. Webb said that while imaging alone isn’t completely new, Moichor’s methodology is effectively automating and perfecting a veterinary pathologist’s skillset in a way that is even more precise and even faster than a pathologist like her could achieve.
“I’m incredibly excited about this because now we have an option to use technology specifically built for veterinary species, rather than adapting technology originally built for human use. And since animals can't tell us how they're feeling, we need the best diagnostics to ensure we don't miss something."
Here are some examples of how image-based CBCs provide a superior quality result than flow cytometry.
As one might expect, image-based CBCs start with automated microscopy identifying the monolayer and imaging the cells within a statistically more accurate subset of cells than a pathologist would examine (300-500 cells).
By using images though, a couple of key opportunities arrive. For one, it becomes possible for an AI model using computer vision to not only count cells, but also analyze morphology. Not only do the cells get segmented by type for interpretation by an AI algorithm and for a pathologist, but also a visual record is created that’s easily calibrated, revisited and communicated between pathologist and clinician.
As Dr. Webb puts it: “It’s literally photographic evidence.”
The model has been built in a way that enables a pathologist to quickly check and confirm any sample and a team of pathologists are regularly reviewing samples to make sure that the models are completely up to date and are completely accurate.
Where things get very exciting is when we can take that data and do bigger things with it.
With the precision enabled by the morphology recognition, a second layer of sophistication becomes possible based on the quality of data that comes with imaging.
For Dr. Webb, it’s a short jump from morphology to training the model to even recognize hemoparasites. “Clinicians will not only get results a lot faster but they will also be a lot more in-depth than the preliminary results clinicians are used to getting,” Dr. Webb said.
“As the AI continues to receive training data, not only will we be able to build the models to pick up on very specific cell types or parasite types, but we can also feed the model information with what to do with the information,” Dr. Webb said. “That could mean the algorithm picking out an inflammatory leukogram without a pathologist even having to have laid eyes on it yet.”
“We can train the AI to differentiate types of leukemia on a preliminary basis. We’re already training the AI to recognize actual blast cells in birds and not just recognize them as big cells the way an automated analyzer would in mammals.
We’re driven by the notion of pathologist review-level results within the timeframe that you would normally be getting just the initial CBC results,” Dr. Webb said.
Currently, reference labs only provide one reference interval for all dogs. But the literature shows, for example, that sighthounds have a higher normal hematocrits than other breeds because they needed to be able to oxygenate better than other dogs when bred for racing.
So, as we continue to build our bank of data for sight hounds we will be able to provide those breed specific reference intervals.
This won’t be unique to just sight hounds, either. Thanks to our quickly growing database, we will have the ability to provide reference intervals specific to each dog breed and cat breed, as we’re currently working on with birds and reptiles.
Whether you’re dealing with a fractious pet, a teacup breed, or a chemo patient where you need to check neutrophil levels regularly, with imaging-based CBCs you don’t need to take a large volume of blood for a high quality CBC. For cases where you’re hesitant to draw a lot of blood and you know you need enough for a chemistry panel too, we only need enough for a smear to run a basic CBC because our artificial intelligence can achieve a superior level of accuracy in counting individual cells with less blood than would be possible with flow cytometry.
It can be challenging for a clinician to communicate with a pathologist what they are seeing on an in-house slide and the cells the clinician wants the pathologist to be looking for.
But with imaging, a clinician can say, This number cell under this column, what is that? And the pathologist can look at that number cell in that column and see exactly which cell they're talking about.
“So it can really improve communication about the case between the clinician and the pathologist and that’s just one more thing that's so great about doing CBCs via imaging analysis,” Dr. Webb said.
There are many other great reasons why clinics love Moichor, including the Internal Messaging System, which is a direct line for instant messaging with the a pathologist, and is a great resource for quick answers.
Here are a just a few of the complementary benefits you’ll find with Moichor:
For more details on the technology, check out our white paper or speak to a member of our team.