AI in Companion Animal Oncology: From Cancer Diagnostics to Precision Treatment to AI-Enabled Drug Discovery

Artificial intelligence is moving veterinary oncology beyond conventional image interpretation towards data-driven cancer characterization, individualized treatment decisions and potentially new therapeutic discoveries. Peer-reviewed studies in canine lymphoma, mast cell tumors, brain tumors and molecular diagnostics demonstrate genuine progress—but also expose the gap between promising algorithms and proven clinical benefit.

Cancer remains one of the most consequential health challenges in companion animals, particularly as dogs and cats live longer and owners increasingly seek advanced diagnostic and treatment options. Yet veterinary oncology continues to face practical constraints: specialist availability, diagnostic variability, treatment costs, limited molecular data and uncertainty about how an individual tumor will respond to therapy.

Artificial intelligence (AI) offers a potential way to address several of these challenges. Machine-learning systems can analyze pathology slides, identify patterns in medical images, integrate molecular information and estimate the probability of treatment response. More advanced systems may eventually help clinicians choose between therapeutic options and support the development of new anticancer drugs.

The scientific evidence is encouraging, but uneven. Diagnostic image analysis has a growing research base; treatment-response prediction has produced important clinical proof-of-concept findings; and AI-assisted drug discovery remains a longer-term opportunity requiring substantial experimental validation.

A central question for the animal-health industry is therefore not simply whether AI can recognize cancer. It is whether AI can make cancer care more accurate, more personalized, more accessible and demonstrably better for the animal.

1. The evidence: where veterinary oncology AI stands today

A scoping review by Li, Yu and Gao, published online in 2025 and appearing in the 2026 volume of BMC Veterinary Research, identified 69 studies applying AI or machine learning to companion animal oncology. The literature covered diagnostic classification, prognosis, molecular analysis and treatment-related applications. Fifty-four studies focused exclusively on dogs, illustrating the comparatively limited evidence base for cats.

Key Evidence at a Glance

Sources: Li et al.; Shum et al. Figures refer to particular studies and tasks, not pooled clinical performance or universal veterinary AI accuracy.

The research points to five broad developments:

• Pathology AI: automated analysis of tumor morphology, cell counts, mitotic figures and grading-related features

• Diagnostic AI: AI-assisted detection, segmentation, classification and quantitative characterization of tumors

• Molecular diagnostics AI: computational interpretation of genetic, epigenetic and circulating biomarker data

• Precision Therapy AI: prediction of drug sensitivity, response and potentially treatment resistance

• Drug Discovery AI: target discovery, drug repurposing and more efficient clinical research

These areas are not equally mature. Evidence is strongest for selected diagnostic tasks. Predicting treatment benefit and demonstrating improved survival require more demanding studies.

2. AI in Pathology: one of the strongest near-term applications

Digital pathology converts conventional tissue slides into high-resolution digital images. Computer-vision algorithms can analyze these images to identify tumor patterns, quantify cellular features and support pathologists in classification and grading.

The clinical rationale is straightforward: pathology remains central to cancer diagnosis, but some tasks are time-consuming, quantitative and subject to interobserver variation.

Canine lymphoma: a 2026 proof of concept

A study published in Veterinary Oncology in February 2026 investigated deep-learning-assisted cytological analysis of canine lymphoma using a ResNet-50 convolutional neural network. The study included 184 dogs, 260 lymph-node slides and 2,600 digital images.

The findings support the potential use of AI to flag suspicious specimens, assist preliminary classification and prioritize cases for specialist review. They do not establish that an AI model can independently confirm lymphoma or replace immunophenotyping and other appropriate tests.

Mast cell tumors: improving quantitative consistency

Canine cutaneous mast cell tumors are another important application because grade and morphology can influence treatment decisions. The veterinary oncology scoping review describes research in which:

• An algorithm classified selected cutaneous round-cell tumors with 91.66% accuracy and reported 100% accuracy for mast cell tumor grading within its study dataset

• Computer-assisted mitotic figure detection increased a pathologist’s F1-score from 0.68 to 0.79, alongside a reported 38% reduction in false negatives

• Automated nuclear morphometry achieved an AUC of 0.943 for predicting tumor-specific survival in one study, compared with 0.839 for the manual comparator

The most useful near-term workflow would have AI identify relevant tissue regions, quantify mitotic figures and nuclear characteristics, flag uncertainty and provide measurements for the pathologist’s final interpretation.

Industry implication: AI-assisted pathology is a promising entry point for veterinary diagnostics companies because it addresses a defined workflow and can be evaluated against established pathology assessments.

3. AI in Veterinary Imaging: Beyond Detecting a Visible Mass

Radiography, ultrasound, CT and MRI play important roles in detecting tumors, staging disease and planning surgery or radiation therapy. AI can potentially improve the consistency of image interpretation and extract quantitative features that are difficult to assess visually.

Radiomics, for example, measures image characteristics such as shape, texture, intensity and spatial heterogeneity. Machine-learning models can combine these features to estimate disease characteristics or outcomes.

Brain tumors: classification and grading

The scoping review summarizes canine MRI research reporting approximately 85% accuracy in distinguishing glial neoplasia from noninfectious inflammatory meningoencephalitis. Other studies reported classification performance of up to 94% for selected glioma types or grades and up to 96.8% correct classification for canine meningioma histological grade.

These findings suggest that MRI-based AI may support diagnostic prioritization where biopsy is invasive or difficult.

However, study-specific performance should not be interpreted as evidence that MRI AI can replace histopathology. Scanner protocols, image quality, case selection and disease prevalence can substantially affect model performance.

Pulmonary metastases: why specificity is not enough

The scoping review reports an evaluation of Vetology AI software for pulmonary nodules and masses with 69.3% accuracy and 93.75% specificity.

AI imaging tools should therefore be evaluated on clinically relevant measures, including sensitivity by lesion size, false-negative rates, incremental benefit when used by a radiologist, and the effect on downstream clinical decisions.

A 2026 example: CT radiomics and lymph-node metastasis

A study published online in July 2026 evaluated CT-derived radiomic features to predict cervical lymph-node metastasis in dogs with oral malignancies. It included 49 dogs and 195 lymph-node observations, with 107 radiomic features extracted per node.

The results show the difficulty of developing a reliable metastasis classifier from a relatively small dataset. The random forest model missed many metastatic observations at the evaluated threshold, while XGBoost produced numerous false positives. Patient-level performance was also modest.

The lesson is important for the commercial market: more features and more complex algorithms do not automatically produce better clinical decisions. External validation and careful threshold selection are essential.

4. Liquid Biopsy and Early Cancer Detection

Liquid biopsy seeks evidence of cancer from biological material in blood or other fluids. Depending on the test, this may include circulating tumor DNA, DNA methylation patterns, microRNA or protein biomarkers.

AI can help combine multiple signals into a risk score, particularly when no single biomarker is sufficiently informative on its own.

The veterinary oncology scoping review summarizes a canine multicancer detection test combining cell-free DNA measurements, next-generation sequencing and logistic regression. Its validation study reported 71.3% overall sensitivity and 98.7% specificity for seven common cancer types. The review also notes that the commercial PetDx operation closed in 2024.

This is an instructive example of both scientific promise and commercial uncertainty: promising diagnostic performance does not guarantee sustainable adoption.

Why prevalence changes screening performance

Consider a hypothetical screening population of 10,000 dogs with cancer prevalence of 1%. If a test achieved 71.3% sensitivity and 98.7% specificity in that population, the expected results would be approximately:

Illustrative calculation using the reported sensitivity and specificity and an assumed 1% prevalence; this is not the study’s observed screening result.

Even a highly specific test can generate substantial follow-up testing when used in a low-prevalence population. A negative result also cannot exclude every cancer, particularly if sensitivity varies by tumor type or stage.

The future of AI-enabled screening may therefore lie in risk-adapted testing that combines age, breed, clinical history, previous cancer, laboratory findings and other relevant information to determine which animals are most likely to benefit from further investigation.

Such a strategy requires prospective evaluation of diagnostic benefit, false positives, overdiagnosis, cost and animal welfare before it can be recommended as a routine screening programme.

5. AI-Guided Chemotherapy: Transition from Diagnosis to Treatment

One of the most consequential developments in veterinary AI is research that connects computational predictions with actual treatment outcomes.

A 2024 study published in Frontiers in Oncology—Multimodal machine learning models identify chemotherapy drugs with prospective clinical efficacy in dogs with relapsed B-cell lymphoma—investigated models designed to predict response to 10 commonly used chemotherapy drugs.

The investigators reported clinical use of prediction-guided treatment recommendations across multiple US clinics. The reports were sent seven days after the laboratory received live tumor biopsies. Clinicians could use the results alongside their own judgment rather than being required to follow a particular prediction.

The study reported better outcomes on measures including overall response, complete response, duration of complete response and survival among dogs receiving drugs predicted to be effective, compared with relevant comparison groups.

Nevertheless, the findings should not be interpreted as definitive proof of a universal survival benefit. Treatment-selection studies can be affected by differences in patient fitness, prior therapies, disease severity, clinician decisions and other confounding factors. Larger prospective controlled studies are needed to establish the incremental benefit of AI-guided treatment over standard oncology care.

What precision oncology could deliver

A more advanced treatment-response system could combine tumor characteristics, molecular results, disease stage, previous treatment, organ function and patient-specific safety constraints.

Its goal would be to estimate the probability of response to each relevant therapy, provide uncertainty estimates and identify supporting evidence.

Crucially, predicting response to a drug is not the same as proving that the drug is the best treatment for that patient. Comparative treatment benefit must be established through appropriately designed clinical studies.

6. AI in Surgery, Radiation Therapy and Longitudinal Monitoring

AI could support cancer treatment well beyond diagnosis and chemotherapy selection.

The scoping review describes AI research in CT-based tumor segmentation, MRI-derived synthetic CT generation and optical coherence tomography for distinguishing healthy from cancerous tissue during canine soft-tissue sarcoma procedures.

These applications could reduce repetitive work and improve measurement consistency. They should not be assumed to reduce recurrence or improve survival until patient-level outcome studies demonstrate those benefits.

7. AI Drug Discovery: Longer-Term Therapeutic Frontier

AI may eventually contribute to new veterinary cancer treatments through target discovery, drug repurposing, molecular design, combination therapy optimization and clinical-trial prioritization.

Five potential pathways deserve attention:

1. Target discovery: identifying molecular pathways associated with aggressive tumors or treatment resistance

2. Drug repurposing: prioritizing existing compounds for laboratory testing against newly identified targets

3. Biomarker discovery: identifying molecular features associated with response to targeted therapy

4. Combination therapy research: ranking drug combinations for experimental evaluation

5. Trial optimization: identifying suitable patients, improving recruitment and analyzing response patterns

These are credible research directions, but they do not establish that an AI-designed drug has already demonstrated improved outcomes in companion animals.

Comparative Oncology: Opportunity for Veterinary and Human Medicine

Dogs develop spontaneous cancers in natural environments, with intact immune systems and the biological complexity of real patients. Some canine cancers share molecular and pathological characteristics with human cancers, including selected lymphomas, osteosarcomas and mammary tumors.

Comparative oncology can therefore support bidirectional research: human oncology informs candidate targets and experimental methods, while veterinary patients can provide insight into cancer biology and therapeutic response in naturally occurring disease.

However, canine and feline cancers must be studied on their own biological terms. Species differences in tumor genetics, drug metabolism, tolerability and clinical practice mean that human models cannot simply be transferred without validation.

8. Commercial Opportunity: Where Should Animal-Health Industry Invest?

The most attractive entry point is likely to be a narrow, clinically useful diagnostic product rather than a universal autonomous oncology system.

A defensible platform would combine curated, multicentre data; reliable links between diagnosis, treatment and outcomes; validated performance across clinical settings; and integration with laboratory and hospital workflows.

The algorithm alone is unlikely to be the strongest competitive advantage. Data quality, clinical evidence, workflow integration and trusted veterinary partnerships are more durable assets.

9. Critical Barriers: Accuracy, Generalizability and Trust

The American College of Veterinary Radiology and European College of Veterinary Diagnostic Imaging issued a joint position statement in 2025 emphasizing transparency, validation, safety, secure handling of patient data and continued veterinary involvement in imaging AI. The same principles apply across veterinary oncology.

A credible AI oncology product must be evaluated at the operating threshold appropriate to its intended use. Sensitivity, specificity, calibration, predictive values, external validation and clinical consequences are all important; headline accuracy alone is insufficient.

10. Road to 2030: A Realistic Outlook

These are strategic projections, not a published industry consensus or guaranteed milestones. Adoption will depend on data access, funding, regulatory requirements and the strength of clinical evidence.

The most defensible near-term expectation is greater use of AI in diagnostic workflows. The potentially more transformative opportunity is treatment-response prediction. Routine autonomous cancer diagnosis and AI-designed veterinary cancer drugs should not be assumed to be standard care by 2030.

11. Conclusion: From Cancer Detection to Better Treatment Decisions

AI is establishing a credible role in companion animal oncology, but its value will ultimately depend on what it enables clinicians to do better.

Digital pathology and imaging offer practical starting points for improving consistency, quantification and workflow efficiency. Liquid-biopsy analytics may support earlier detection, provided that screening benefits outweigh false positives and downstream harms. Treatment-response prediction represents a particularly important frontier because it addresses the uncertainty that remains when a cancer has relapsed or multiple therapeutic options are available.

The 2024 canine lymphoma study provides an important proof of concept for linking machine-learning predictions with clinical treatment outcomes. At the same time, the more modest results of independent radiomics research demonstrate why rigorous validation remains indispensable.

For animal-health companies, laboratories and investors, the most promising strategy is to develop validated diagnostic assistance first, outcome-linked precision oncology second, and AI-enabled therapeutic discovery as a longer-term programme.

The real measure of success will not be how accurately a model labels an image. It will be whether it helps veterinarians reach better decisions, avoid ineffective interventions, preserve quality of life and ultimately improve outcomes for dogs and cats.

References and further reading

1. Li S, Yu F, Gao Y. The application of artificial intelligence in veterinary oncology: a scoping review. BMC Veterinary Research. 2025; journal volume 2026, 22:28. https://doi.org/10.1186/s12917-025-05192-y

2. Multimodal machine learning models identify chemotherapy drugs with prospective clinical efficacy in dogs with relapsed B-cell lymphoma. Frontiers in Oncology. 2024. https://doi.org/10.3389/fonc.2024.1304144

3. Shum C et al. Deep learning-assisted cytological image analysis for canine lymphoma. Veterinary Oncology. 2026. https://doi.org/10.1186/s44356-026-00056-5

4. Appleby RB et al. American College of Veterinary Radiology and European College of Veterinary Diagnostic Imaging position statement on artificial intelligence. Journal of the American Veterinary Medical Association. 2025;263(6):773–776. https://doi.org/10.2460/javma.25.01.0027

5. Clinical Use of Molecular Biomarkers in Canine and Feline Oncology: Current and Future. Veterinary Sciences. 2024;11(5):199. https://doi.org/10.3390/vetsci11050199

6. Deep Learning for Diagnostic and Clinical Decision Support in Companion Animals: A Systematic Review of CNN-and LSTM-Based Approaches. Frontiers in Veterinary Science. 2026. https://doi.org/10.3389/fvets.2026.1952589

7. Pinard CJ et al. CT-Based Radiomic Features Predict Cervical Lymph Node Metastasis in Dogs With Oral Malignancy: A Machine Learning Study Using Leave-One-Patient-Out Cross-Validation. Veterinary and Comparative Oncology. 2026. https://doi.org/10.1111/vco.70092

8. Generative AI in Veterinary Pathology: Feasibility of a GPT-Based Assistive Tool for Gross, Cytologic, and Histopathologic Assessment of Canine Cutaneous Neoplasms—A Pilot Study. 2026. Full text