HOW CAN WE… use AI in diagnostics? 

Transformative AI-Assisted Ki-67 Prognostic Scoring for Canine Mast Cell Tumours

CVS’ Finn Pathologists has invested in creating an innovative AI-assisted Ki-67 scoring system to help predict the behaviour of canine mast cell tumours, offering more accurate, efficient, and reliable prognostic evaluation.

Canine cutaneous mast cell tumours (cMCT) are among the most common skin tumours seen in veterinary practice. These tumours can range from benign to highly aggressive in behaviour. Accurate grading of cMCT is crucial for predicting tumour behaviour and therefore guiding treatment strategies and improving patient outcomes. 

The proliferation marker Ki-67 is a valuable adjunctive tool for assessing tumour behaviour, as it provides more accurate information about cell proliferation than a histological mitotic count alone. However traditional methods of assessing Ki-67 scores involve manually counting the proliferating cells, which is subjective, time-consuming, and prone to variability. 

Finn Pathologists’ new AI-assisted Ki-67 scoring test represents a significant advancement in veterinary diagnostic oncopathology. Its AI algorithm analyses Ki-67 labelling across the entire section of tumour. Positively labelled cells are highlighted, scores are calculated, and the results are validated by a pathologist before reporting. 

This more automated analysis enables the efficient evaluation of large volumes of histological data, ensuring consistent and reliable prognostic assessments. It removes the subjectivity and variability associated with identifying areas of highest cell proliferation, which is a crucial aspect of tumour grading.

Over the past year, veterinary pathologists Dr. Richard Fox, Dr. Melanie Dobromylskyj, and Dr. Annalize Ide have developed the AI model, in collaboration with Aiforia® Technologies and Bristol Vet Specialists' Owen Davies and Kate Boyd. 

The development process has involved acquiring high-quality digital images of Ki-67 labelled histological samples of cMTC across all grades, inputting this data into the AI model, and refining the algorithm. The AI test has since been validated using cases with known clinical outcomes, such as tumour recurrence and patient survival, with results planned for scientific publication.

Dr. Richard Fox, Lead AI Pathologist at Finn Pathologists said: “As canine mast cell tumours in dogs’ skin are so common, and their behaviour unpredictable, we wanted a more accurate way to help vets to stratify patients based on their risk of tumour recurrence and survival.

“Our objective was to create a more standardised and reproducible method for Ki-67 scoring, streamline the Ki-67 scoring process, and improve the accuracy of the assessment of tumour sections. This assists pathologists in identifying the optimal areas for counting and achieve the most accurate score for each patient.”

IMAGE: Model results with heatmap overlay highlights regions with the highest Ki-67 positivity (warmest colours). Inset: A high-power view of model results displaying Ki-67 nuclear positivity in red.