Can Computers Read Radiographs?
Can computers read radiographs? Learn how veterinary radiology AI supports veterinarians with faster image interpretation, improved diagnostic confidence, and AI-assisted workflows without replacing radiologists.
What Veterinary AI Really Means for Modern Radiology
Based on a CE webinar by Dr. Eric van Eerde, DVM, DACVR, Director of Radiology at SignalPET

Veterinary medicine is entering a new era.
Artificial intelligence is no longer a futuristic concept reserved for Silicon Valley or human hospitals. It’s already being used in veterinary clinics every day — helping veterinarians write medical records, identify disease trends, analyze cytology, and assist in radiograph interpretation.
But one question still sits at the center of the conversation:
Can computers actually read radiographs?
The answer is more nuanced than a simple yes or no.
AI is not replacing veterinarians. It is not replacing radiologists. And despite the hype surrounding artificial intelligence, no responsible veterinary professional should treat AI as an infallible black box.
What AI can do is something incredibly valuable:
It can provide a second set of eyes.
And in a profession where nearly 90% of veterinary radiographs never receive radiologist review, that can be a valuable tool.
The Reality of Veterinary Radiology
Veterinary radiology has always faced a structural problem: a limited number of radiologists available and financial limitations for clients who do not have pet insurance.
In North America and Europe combined, there are fewer than 2,000 veterinary radiologists serving tens of thousands of veterinary clinics. The result is that the vast majority of radiographs are interpreted by general practitioners working alone.
That does not reflect a lack of care.
It reflects the reality of modern veterinary medicine.
Radiologist review can be expensive, turnaround times may not fit within the appointment window, and many pet owners cannot justify the added cost for every patient.
So veterinarians do what they have always done: they rely on their own training, experience, and clinical judgment with varying degrees of comfort.
Even experienced clinicians understand the anxiety that comes with interpreting films independently. Every veterinarian remembers cases where they wondered later:
“Did I miss something?”
That is the gap AI is beginning to fill.
Not by replacing doctors.
By supporting them.
What Artificial Intelligence Actually Is

Artificial intelligence is often misunderstood because the term itself sounds far more magical than the reality.
At its core, AI is pattern recognition.
Modern machine learning systems are trained to identify relationships, detect abnormalities, and recognize patterns across massive amounts of data.
And that matters because radiology itself is fundamentally pattern recognition.
Radiologists spend years training their eyes and brains to identify subtle abnormalities:
- A tiny pulmonary nodule
- Mild cardiomegaly
- A faint interstitial pattern
- A small amount of free abdominal gas
- Early left atrial enlargement
AI excels at identifying outliers and repeating tasks consistently across enormous datasets.
That makes imaging one of the most natural applications for machine learning.

Why AI Fits Radiology So Well
One of the biggest strengths of AI is its ability to process large amounts of visual information without fatigue.
Humans get tired.
Computers do not experience fatigue.
They do not become distracted.
They do not rush because 40 more studies are waiting in the queue.
AI evaluates every image systematically.
That consistency creates enormous value as a clinical support tool.
In human medicine, AI has already demonstrated strong performance in areas such as mammography and pulmonary nodule detection.
Veterinary medicine is beginning to see similar advances.
How Veterinary Radiology AI Is Trained
Most veterinary radiology AI uses what is called supervised learning.
This means radiographs are labeled by experts — typically veterinary radiologists — and fed into the AI system so the model can learn what abnormalities look like.
The quality of the training data matters tremendously.
AI is only as good as the information used to train it.
Poor labels create poor outcomes.
High-quality datasets create stronger and more reliable models.
This is one reason large datasets are so important.
SignalPET’s models, for example, have been trained using more than 20 million veterinary images and clinical cases.
That scale matters because rare diseases, subtle findings, breed differences, and technical variations all become easier for AI systems to recognize when exposed to enough real-world data.
AI Does Not Read “The Entire Radiograph”
One of the most important concepts veterinarians need to understand is this:
AI does not magically understand every possible abnormality in an image.
Modern veterinary radiology AI is actually a collection of highly specific tests bundled together.
For example, an AI model may evaluate:
- Cardiomegaly
- Pulmonary nodules
- Pleural effusion
- Pneumothorax
- Hepatomegaly
- Renomegaly
- Fractures
- Pulmonary patterns
But if the system has not been trained to detect a specific abnormality, it cannot reliably identify it.
That distinction is critical.
A “normal” AI report does not mean the image is universally normal.
It means the AI did not detect abnormalities within the set of conditions it was trained to evaluate.
This is why veterinarians must remain actively involved in the interpretation process.
AI should never replace clinical judgment.
The Veterinarian Must Stay “In the Loop”
The American College of Veterinary Radiology (ACVR) and the European College of Veterinary Diagnostic Imaging (ECVDI) both emphasize an important principle regarding AI:
A veterinarian should always remain involved in the diagnostic process.
That principle is essential.
AI is not designed to function independently.
It functions best when paired with veterinary expertise.
The ideal workflow is collaborative:
- The veterinarian reviews the radiographs independently.
- The AI report provides additional analysis and support.
- The clinician compares findings.
This creates a safer, more reliable diagnostic process.
Rather than replacing the veterinarian, AI augments clinical confidence.
The radiologist review remains the clinical gold standard, but when a radiologist review is not feasible for every case, AI can help when used in conjunction with a veterinarian’s expertise.
The Most Important Advancement: Knowing When AI Is Uncertain

One of the biggest misconceptions about AI is the belief that the goal is perfection.
It is not.
The most important AI systems are the ones that understand their own limitations.
Modern AI platforms increasingly use confidence scoring and escalation workflows.
At SignalPET, cases that exceed AI confidence thresholds can automatically escalate to a board-certified radiologist for review.
That changes the role of AI entirely.
Instead of pretending the model is always correct, the system becomes a safety net:
- Routine cases can be handled quickly and affordably.
- Complex or uncertain cases receive specialist oversight.
- Veterinarians gain support without sacrificing quality.
This hybrid model is likely the future of veterinary radiology.
AI as a way to triage difficult cases into the hands of radiologists while reducing or improving the efficiency of the overall workload.
Where AI Is Already Helping Veterinarians
Radiology is only one part of the broader AI transformation happening in veterinary medicine.
Across the industry, veterinarians are already using AI tools to improve efficiency and reduce administrative burden.
Medical Record Generation
AI-powered medical scribes can automatically generate SOAP notes from recorded conversations.
Instead of spending hours completing records after appointments, clinicians can focus on editing and validating summaries.
Clinical Pathology
AI systems are increasingly being used to identify trends across enormous laboratory datasets.
Tools like Antech RenalTech demonstrated that machine learning models can predict chronic kidney disease in cats years before traditional diagnosis by analyzing massive historical datasets.
Cytology
AI can assist with slide analysis by evaluating large areas of a sample without fatigue, helping identify mitotic figures and cellular abnormalities more consistently.
Population-Level Insights
Because AI can analyze millions of data points simultaneously, it may help uncover associations between disease, geography, breed, age, and other variables that would otherwise remain hidden.
The common thread across all of these applications is efficiency.
AI handles repetitive, data-heavy tasks so veterinary professionals can focus more on clinical care.
What the Current Research Shows

Veterinary radiology AI is still relatively young.
As of 2022, there were only about 40 peer-reviewed publications focused on machine learning within veterinary imaging.
But the early results are promising.
Research has demonstrated encouraging performance in areas such as:
- Hip dysplasia classification
- Vertebral heart score measurements
- Left atrial enlargement detection
- Pulmonary pattern recognition
- Pleural effusion identification
- Thoracic screening
- Fracture detection
Interestingly, many studies show that AI performs best in well-defined tasks with strong pattern consistency.
For example:
- AI performs extremely well in measuring VHS.
- AI performs well in identifying pleural disease.
- AI struggles more with subtle, unstructured interstitial patterns.
That mirrors human learning curves.
Certain abnormalities are easier to standardize than others.
The takeaway is not that AI is perfect.
The takeaway is that AI is becoming clinically useful.
Transparency Matters
As AI becomes more integrated into veterinary workflows, transparency becomes critically important.
Veterinarians should also disclose AI usage to clients whenever it contributes to patient care.
Trust in AI will not come from marketing claims.
It will come from transparency, validation, and consistent clinical outcomes.
The Future of Veterinary AI

AI will not replace veterinarians.
It will not replace radiologists.
But veterinary professionals who understand how to work effectively alongside AI will likely gain significant advantages in efficiency, confidence, and diagnostic support.
The future of veterinary imaging is not about choosing between human expertise and artificial intelligence.
It is about combining both.
The clinics that succeed will be the ones that use AI thoughtfully:
- To improve confidence
- To reduce missed findings
- To support clinical decisions
- To improve workflow efficiency
- To make higher-quality care more accessible
That is the real promise of veterinary AI.
Not replacing medicine.
Strengthening it.
About the Webinar
This article is based on a continuing education webinar presented by Dr. Eric van Eerde, DVM, DACVR, Director of Radiology at SignalPET and board-certified veterinary radiologist. The webinar explored the foundations of artificial intelligence, its current role in veterinary medicine, and the future of AI-assisted radiology.
FAQs
No.
AI should support veterinarians and radiologists, not replace them.
AI works best as a second set of eyes and a clinical support tool
AI performs very well in certain focused tasks like VHS measurements, cardiomegaly screening, pleural effusion detection, and some fracture evaluations.
More subtle findings can still be challenging, which is why AI should always be used alongside clinical judgment
Not exactly.
AI systems evaluate specific abnormalities they were trained to detect. If the model was never trained on a certain condition, it may not reliably identify it.
A “normal” AI report does not mean every possible abnormality is ruled out.
There are not enough veterinary radiologists to review every study.
AI helps provide additional support for cases that otherwise may receive no second opinion at all.
AI will likely improve workflow efficiency, diagnostic support, and access to radiology insight.
But veterinarians will remain central to the diagnostic process.
Yes.
Dr. van Eerde highlighted the importance of transparency and helping clients understand how AI contributes to patient care.
Yes.
AI is already being used for:
- SOAP note generation
- Medical record summarization
- Clinical pathology analysis
- Cytology support
- Workflow automation
Modern AI systems can recognize lower-confidence cases.
At SignalPET, uncertain or complex studies can automatically escalate to a board-certified radiologist for further review.
Potentially, yes.
AI evaluates images consistently and does not experience fatigue or distraction.
Dr. van Eerde noted this may help identify findings that could otherwise be overlooked.
It means veterinarians remain involved while AI provides additional analysis and support.
The goal is collaboration, not replacing clinician oversight.
Table of Contents








