Empowering veterinary decisions:
A data-driven approach to pet care

Importance of timely and effective treatment for pets with lymphoma

Pet care has improved significantly in the last 25 years. However, how quickly and fully a sick pet recovers relies mainly on the medical services near its owner’s home.
Lymphoma is a type of cancer affecting the lymph system. A diagnosis demands urgent and effective medical care to start the pet on a path to recovery or ease its suffering if the diagnosis is terminal. It’s vital to provide the correct treatment promptly to prevent pets from suffering needlessly or dying prematurely.

Unlocking the potential of asynchronous pet data

Breed diversity and data harmonization
The pet care industry faces a significant challenge due to the wide range of breeds, with over 300 dog breeds and more than 70 cat breeds documented. This complicates the identification and treatment of diseases. Different breeds may respond differently to treatments. Combining pet data from various sources is crucial to address this challenge. This ensures that accurate and comprehensive information is available for analysis and decision-making.

The need for personalized analysis
Our client, a global pet care solution provider in over 100 countries, required personalized analysis to determine the ideal lymphoma treatment plan for individual pets based on their unique attributes and symptoms.

Understanding the correlation between dosage duration and life expectancy The client also sought to understand the correlation between the dosage duration of specific cancer-fighting drugs and individual pets’ life expectancy. This analysis was crucial for optimizing treatment outcomes and improving quality of life.

Revolutionizing pet oncology: AI-driven treatment plans

We received data on over 23 million cats from the client. In several cases, we enhanced the data with other data sets to fit into over 80 feature categories. About 500 cats were chosen for analysis to achieve several goals.

We focused on the following areas to develop the solution:
● Oncology research: Our data scientists and researchers examined published works on pet oncology to gather information.
● Extrapolating data: We used Natural Language Processing to convert veterinarians’ notes into useful digital data, capturing over 80 key attributes of pets. By employing SHAP analysis (Shapley Additive Explanations), we pinpointed essential factors for predicting feline lymphoma.
● Pet persona clustering: Clusterization algorithms grouped similar data points on the cats.
● Predictive Al algorithms: We developed a pet-level algorithm to suggest the best treatment for any specific cat.

We delivered a solution to the client with the following three components:

Saving one cat now
With this solution, the client can predict the occurrence of lymphoma in pets and the effectiveness of different treatments based on each pet’s unique characteristics. Our client can derive individualized pet survival predictions and the potential determinants of their future health status through machine learning and survival analysis techniques.

Saving many pets in the future
Considering the insights drawn from various algorithms, our solution offers a comprehensive understanding that can be applied to similar issues other animals face. Therefore, the client can expand its use to help even more animals.