Virginia Tech
Use of Artificial Intelligence to Predict Histologic Grade of Canine Mast Cell Tumors from Cytologic Specimens
Abstract
dc:description.abstractBackground: Cutaneous mast cell tumors (MCT) are the most common skin tumors in dogs and display variable biological behavior which can be predicted via histological grade. Although diagnosis is usually straightforward using minimally invasive fineneedle aspiration and cytology, accurate prognostication continues to rely on histopathologic evaluation, which requires surgical excision and may increase patient morbidity. Cytologic grading schemes have been proposed; however, their clinical utility remains limited due to low specificity for identifying high-grade tumors. This study evaluates the use of artificial intelligence to predict MCT grade from archived cytology samples using matched histopathologic grading as the reference standard. Objective: The primary objective was to apply a previously described convolutional neural network (CNN) to differentiate cutaneous low-grade (CLG) and cutaneous highgrade (CHG) MCTs using cytology images. A secondary objective was to use PCR amplification of cytologic MCT specimens to detect mutations in the c-kit gene associated with aggressive tumor behavior. Methods: Forty-five cytologic specimens from canine MCTs were retrospectively identified, representing 15 of each histologically confirmed CLG, CHG, and subcutaneous (SC) MCTs. Whole-slide images from the CLG and CHG specimens were obtained and 10 representative images per slide were extracted for analysis. These images were then used to train, validate, and test the CNN's ability to grade MCTs accurately. Additionally, DNA was extracted from all cytologic specimens, followed by PCR amplification of c-kit exon 11. PCR products were visualized via gel electrophoresis, and Sanger sequencing was used to confirm canine c-kit amplification. Results: The CNN achieved 94% accuracy when distinguishing CLG and CHG MCTs using cytologic images. We also show successful DNA extraction from cytologic specimens and identification of previously described mutations in canine MCTs. These findings support the feasibility of mutation detection via cytology and its potential application in non-surgical prognostication of canine MCTs.
Degree
thesis:*- Name thesis:degree_name
- Master of Science
- Level thesis:degree_level
- masters
- Discipline thesis:degree_discipline
- Biomedical and Veterinary Sciences
- Department dc:contributor.department
- Biomedical and Veterinary Sciences
- Grantor dc:publisher
- Virginia Tech
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Nightengale, Marlie
- Chairs dc:contributor.committeechair
-
- Pacholec, Christina
- Strandberg, Natalia
- Committee members dc:contributor.committeemember
-
- Zimmerman, Kurt L.
- Lahmers, Kevin K.
Subjects
dc:subject × 5Rights
dc:rights- Statement dc:rights
-
- In Copyright
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Other
- vt_gsexam:46757