Industries / Health / Veterinary
Imaging and lab tests
Taking pictures inside the body with X-rays, ultrasound, CT and MRI, and testing blood, urine, feces and cells, either in the clinic or at an outside lab.
5
Milestones achieved, from 1901 to 2026.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1901
X-rays of animal patients
Six papers on X-rays in veterinary medicine appeared in 1896, the year after Wilhelm Röntgen's discovery. The Vienna Veterinary College obtained X-ray equipment in 1897 and in 1901 made the first radiographs of large and small animal patients, by the account of veterinary radiologists Darryl Biery and Harker Rhodes; the same year its staff reported skin damage from radiation in dogs.
Lessonthe rays that showed bone also burned skin, and safe exposure had to be learned.
Credit
Vienna Veterinary College; Professor Eberlein of the Royal Veterinary Academy in Berlin, called the father of veterinary radiology
1980
Ultrasound finds early pregnancy
E. Palmer and M. A. Driancourt in France published the use of ultrasound, sound echoes turned into a picture, to examine a mare's uterus and ovaries. Before it, vets found pregnancy by feeling through the rectum, reliable from about 30 days; with ultrasound, routine checks moved to 14 to 16 days after ovulation, as Patrick McCue described the change in his 2023 Frank J. Milne lecture to the American Association of Equine Practitioners.
Credit
E. Palmer and M. A. Driancourt (Theriogenology, 1980)
2013
Quality rules for clinic analyzers
Many practices run blood and urine tests on their own machines, and a 2013 review in the Journal of the American Veterinary Medical Association noted that no regulations like those for human labs exist for veterinary lab testing. The American Society for Veterinary Clinical Pathology (ASVCP) published quality guidelines for these point-of-care analyzers that year. A survey of Veterinary Information Network members found that 88% ran some quality checks, but only 25% did so on ELISA kits, the rapid antibody-based tests, and not knowing how was the most common reason given.
Lessona machine moved into the clinic needs the same routine checks as one in a lab, and staff need training to run them.
Credit
B. Flatland, K. P. Freeman, L. M. Vap and K. E. Harr for the ASVCP; survey by R. Bell, K. Harr, M. Rishniw and P. Pion
2024
AI finds Giardia in feces
Vetscan Imagyst, sold by Zoetis, scans a fecal slide and uses a deep-learning model to find parasite eggs and cysts; it was first described in a 2020 paper. An outside team at Ludwig-Maximilians-University Munich tested it on samples from 164 young dogs against a lab immunofluorescence test: it found Giardia with 88.4% sensitivity and 98.1% specificity, against 74.4% and 98.1% for IDEXX's SNAP Giardia rapid test. Its cyst counts varied widely between repeat runs (a coefficient of variation of 67%), so the team advised using it only for yes-or-no answers.
Lessonan AI reader can be reliable at finding something and unreliable at counting it.
Credit
Zoetis; S. Kanski, K. Busch, R. Hailmann and K. Weber (Journal of Veterinary Diagnostic Investigation, 2024)
2026
AI readers miss diagnoses
A Murdoch University team sent abdominal radiographs of 53 dogs, each with a confirmed diagnosis, to six commercial AI reading services in late 2024: Picoxia, Radimal, RapidRead, SignalRAY, Vetology AI and X Caliber Vet AI. Sensitivity for findings ranged from 28% to 78%, a blocked small intestine was caught in 23% to 69% of cases, and the authors judged none suitable for clinical use in its current form. A 2025 study supported by SignalPET had found its software as accurate overall as the best of 11 radiologists on 50 studies, but less sensitive than nine of them.
Lessonresults from makers' own data did not hold on general-practice cases, and the services do not disclose their training data or when their models change.
Credit
Doris Ma, Stephen Joslyn and colleagues, Murdoch University (Journal of the American Veterinary Medical Association, 2026)
Watch for
Tumor cells read without slides
IDEXX says its inVue Dx analyzer now reads fine-needle aspirates, cells drawn from a lump through a needle, and detects mast cell tumors in dogs without a glass slide. It also says its Cancer Dx lymphoma blood test is used by more than 5,500 practices. The release gives no sensitivity or specificity, and its only outside voice is a vet paid as an IDEXX consultant.
Achieved whenresearchers not paid by IDEXX compare its mast cell results with a pathologist's diagnosis on a stated number of dogs and publish sensitivity and specificity.
Claimed by
IDEXX, January 2026
AI blood counts reach Europe
Zoetis says Vetscan OptiCell, a cartridge-based analyzer that counts blood cells by imaging them with AI, gives "expert-level accuracy". It went on sale in the US, UK and Australia in late 2024 and in nine European countries from September 2025, and the only study cited is Zoetis's own unpublished data.
Achieved whenan independent study compares its counts with a reference lab analyzer and manual smear counts and publishes the agreement.
Claimed by
Zoetis, September 2025
AI reader publishes its numbers
Vetology released the sensitivity, specificity and sample size of more than 89 of its X-ray classifiers, for dog and cat chests, abdomens and spines, from about 300,000 test cases in all; its heart failure classifier, for example, has 89.5% sensitivity across 10,951 cases. It calls itself the first and only company in veterinary imaging AI to publish this much. The release does not say who labeled the test cases, and no outside party has checked the figures.
Achieved whena group with no tie to Vetology tests the classifiers on cases the company has not seen and publishes figures that match.
Claimed by
Vetology, January 2026
Open
01
A radiograph set with answers
An open collection of at least 500 dog and cat radiograph studies from several clinics, each paired with a diagnosis confirmed by surgery, necropsy (an animal autopsy), CT, ultrasound or cytology, and released with a scoring script. Then submit it to the commercial AI reading services and publish each one's sensitivity and specificity by finding.
The briefCredit · unclaimed02
Open parasite egg images
A public library of microscope images from dog and cat fecal flotation slides, the standard test for worm eggs, with every egg and cyst boxed and named by two parasitologists, plus a scoring script and a baseline detector.
The briefCredit · unclaimed
Open 01 · the brief
A radiograph set with answers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open collection of at least 500 dog and cat radiograph studies from several clinics, each paired with a diagnosis confirmed by surgery, necropsy (an animal autopsy), CT, ultrasound or cytology, and released with a scoring script. Then submit it to the commercial AI reading services and publish each one's sensitivity and specificity by finding.
Why it is still open
The 2026 Murdoch test used 53 cases and called for larger independent tests. A poster by IDEXX researchers says curated open veterinary imaging datasets remain scarce, and a 2026 review of 98 diagnostic AI studies in dogs and cats found that 13.3% shared their data and 14.3% tested on data from outside the developers.
Built with what exists
teaching hospitals' image archives exported as DICOM, the standard medical image format, with owner details removed using open tools such as pydicom; labels from two board-certified veterinary radiologists plus the confirmed outcome; Zenodo or a similar repository for hosting; and the services' standard paid submissions.
First step
Ask one veterinary teaching hospital for a list of cases from the past five years that had both radiographs and a surgical or necropsy diagnosis, and draft the data-sharing terms.
Done looks like
500 studies from at least three clinics, a confirmed diagnosis for every one, two radiologists' reads with their agreement measured, and one published table scoring at least four AI services and the radiologists on the same cases.
Open 02 · the brief
Open parasite egg images
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public library of microscope images from dog and cat fecal flotation slides, the standard test for worm eggs, with every egg and cyst boxed and named by two parasitologists, plus a scoring script and a baseline detector.
Why it is still open
The AI egg readers sold for pets are trained on their makers' own images. In a 2024 comparison on 59 dog and cat samples, Parasight All-in-One counted 10 to 28 times more eggs than Vetscan Imagyst, depending on the worm. The open egg image sets that exist are for cattle, from the 2024 AI-KFM challenge, or for human parasites.
Built with what exists
a microscope camera or slide scanner at a parasitology teaching lab, labeling tools such as CVAT or Label Studio, an open object detector such as YOLO, and Zenodo for hosting.
First step
Ask a veterinary school's parasitology lab to photograph the slides it already prepares for teaching or diagnosis for one week, and label the first 200 images.
Done looks like
at least 3,000 images holding 10,000 labeled eggs and cysts across at least six parasite types, two experts' labels with their agreement measured, and a baseline detector's recall for each parasite published.