From Glass to Pixels
Traditional histopathology requires a pathologist to examine tissue sections on glass microscope slides, rendering a diagnosis based on pattern recognition developed through years of training. This process, while effective, has inherent limitations: inter-observer variability (different pathologists may interpret the same slide differently), fatigue effects during high-volume workloads, and geographic limitations (the slide must physically reach the pathologist).
Digital pathology addresses these limitations through whole-slide imaging (WSI) — high-resolution scanning of entire tissue sections into digital files that can be viewed, analyzed, and shared electronically. Combined with artificial intelligence (AI) algorithms trained on thousands of annotated slides, digital pathology promises more consistent, faster, and potentially more accurate diagnoses.
For veterinary medicine, where pathology expertise is even more concentrated geographically than in human medicine, the implications are particularly significant.
Whole-Slide Imaging in Veterinary Practice
WSI technology captures glass slides at 20x or 40x magnification, generating high-resolution digital images that can be viewed on any computer with appropriate software. A single whole-slide image at 40x magnification may contain over 10 billion pixels and occupy 2-5 GB of storage.
Current applications in veterinary pathology:
- Teleconsultation: Digitized slides can be shared with specialist pathologists anywhere in the world, eliminating shipping delays and enabling same-day expert review
- Education: Digital slide archives enable consistent training across veterinary schools and continuing education programs
- Archival: Digital slides do not degrade over time, unlike glass slides that can fade, break, or develop coverslip artifacts
- Quantitative analysis: Digital images enable precise measurement of tumor dimensions, cell counts, and staining intensity that are difficult to standardize with conventional microscopy
AI-Assisted Diagnosis
Mitotic Count Automation
Bertram et al. (2020) demonstrated that deep learning algorithms can count mitotic figures in canine cutaneous mast cell tumors with accuracy comparable to expert pathologists. This is clinically important because mitotic count is one of the strongest prognostic factors for MCTs, and manual counting is time-consuming and subject to inter-observer variability.
The algorithm evaluates each cell division figure candidate, classifies it as a true mitosis or mimic (apoptosis, karyorrhexis), and generates a total count per defined area. In validation studies, the AI’s counts correlated strongly with expert consensus counts and predicted clinical outcomes at least as well as manual counts.
Tumor Classification
AI models trained on annotated whole-slide images can assist with:
- Differentiating benign from malignant tumors
- Classifying tumor type (carcinoma vs. sarcoma vs. round cell tumor)
- Grading tumors according to established criteria (Patnaik/Kiupel for MCTs, histological grading for soft tissue sarcomas)
- Identifying tumor margins on resection specimens
Aubreville et al. (2020) created a fully annotated dataset of canine mammary tumor whole-slide images, demonstrating both the feasibility and value of veterinary-specific AI training data. Notably, their work showed that AI models trained on canine tumor data could inform human breast cancer research, illustrating the One Health value of veterinary digital pathology.
Pattern Recognition
Pielmeier et al. (2022) showed that deep learning models can identify spatial patterns within tumor microenvironments that are conserved across species and cancer types. This cross-species pattern recognition has implications for:
- Predicting tumor behavior based on microenvironmental features rather than tumor cell morphology alone
- Identifying therapeutic targets based on spatial relationships between tumor cells and immune cells
- Prognostication beyond traditional grading systems
Current State of Implementation
Digital pathology adoption in veterinary medicine is accelerating but remains uneven:
Large reference laboratories (IDEXX, Antech, university veterinary pathology departments): Actively implementing WSI and beginning to integrate AI-assisted tools for quality assurance and diagnostic support
Commercial AI products: Several companies are developing veterinary-specific AI diagnostic tools, including automated mitotic counting and tumor classification algorithms. Most are in validation phases rather than widespread clinical deployment.
Individual practices: Most general veterinary practices continue to rely on conventional glass slide submission to reference laboratories. Direct adoption of digital pathology at the practice level is rare due to scanner cost ($50,000-$200,000+) and workflow integration challenges.
Impact on Canine Cancer Diagnosis
The most immediate clinical impact of digital pathology is in cancer diagnosis and grading:
- More consistent MCT grading: AI-assisted mitotic counting reduces the variability between pathologists that currently creates prognostic uncertainty, particularly for borderline cases
- Faster turnaround: Digital slide sharing and AI pre-screening can reduce pathology report turnaround from 3-7 days to 1-2 days
- Second-opinion access: Digitized slides can be shared with specialist pathologists globally for difficult cases without physical slide transport
- Quantitative biomarkers: AI enables extraction of quantitative features (nuclear-to-cytoplasmic ratio, cell density, vascular density) that are impractical to measure manually but may have prognostic value
- Quality assurance: AI systems can flag slides that may have been misclassified, serving as a safety net for pathologist review
Limitations and Concerns
- Training data bias: AI models trained primarily on samples from specific breeds, tumor types, or laboratories may perform poorly on underrepresented populations
- Validation gaps: Many AI pathology tools lack prospective clinical validation — retrospective accuracy does not guarantee real-world performance
- Regulatory framework: No FDA or equivalent veterinary regulatory pathway exists specifically for AI-assisted veterinary diagnostics
- Cost: WSI scanner acquisition and AI platform licensing costs limit adoption outside large reference laboratories
- Pathologist role: AI is a decision-support tool, not a replacement for veterinary pathologists. Final diagnostic responsibility remains with the pathologist
- Data storage: The massive file sizes of whole-slide images create significant storage and bandwidth challenges
Practical Implications for Dog Owners
While digital pathology operates behind the scenes, its advances affect owners in several ways:
- More accurate tumor grading may lead to better-informed treatment decisions for dogs with cancer
- Faster pathology results reduce the anxiety of waiting for biopsy reports
- Access to specialist pathology review regardless of geographic location
- Research acceleration: Digital pathology datasets enable faster development of new diagnostic markers and treatment approaches
When your dog has a biopsy submitted for pathology, you can ask your veterinarian whether the laboratory uses digital pathology and AI-assisted tools. This information helps contextualize the diagnostic report.
Related Conditions
Related Science
- AI in Veterinary Diagnostics
- Canine Cancer Immunotherapy Advances
- Mast Cell Tumor Grading and Prognosis
Frequently Asked Questions
Does digital pathology replace the pathologist?
No. AI-assisted digital pathology serves as a decision-support tool that improves consistency and efficiency, but final diagnostic interpretation and clinical correlation remain the responsibility of a qualified veterinary pathologist.
Will digital pathology make biopsy results faster?
Yes, in many cases. Digital slide sharing eliminates shipping time, and AI pre-screening can flag urgent cases for priority review. Some reference laboratories are already reporting faster turnaround times for digitized submissions.
Is AI-assisted diagnosis as accurate as a pathologist?
For specific, well-defined tasks (mitotic counting, tumor classification), validated AI algorithms perform comparably to expert pathologists. For complex diagnostic scenarios requiring clinical correlation and nuanced interpretation, human pathologists remain essential.
How does this affect the cost of my dog’s biopsy?
Currently, digital pathology does not significantly change biopsy costs to clients. The technology costs are absorbed at the laboratory level. As adoption scales, efficiency gains may eventually reduce per-sample costs.
Bottom Line
Digital pathology is transforming veterinary diagnostics through whole-slide imaging and AI-assisted analysis. The most immediate clinical impact is more consistent, faster cancer diagnosis — particularly for tumors like mast cell tumors where grading determines treatment. While full adoption is still underway, the technology is already improving diagnostic quality at major reference laboratories and will increasingly influence the accuracy and speed of pathology results for companion animals.
References
- Aubreville M et al. Annotated canine breast cancer whole slide image dataset. Sci Data. 2020;6:274.
- Bertram CA et al. Computerized mitotic count in canine MCTs. Vet Pathol. 2020;57(2):214-226.
- Pielmeier RM et al. Deep learning spatial behaviors in tumor microenvironments. Nat Commun. 2022.
- Salvi M et al. Image processing for digital pathology. Comput Biol Med. 2021;138:104912.