Canine cancer study shows machine learning models trained with different data types yield better clinical outcome prediction.
Heterogeneity is a significant issue when we talk about failed cancer treatments. And that is why precision medicine promises a revolution in cancer therapy.
Studies on cancer heterogeneity have enabled scientists to categorize different tumors into subtypes. Cancer treatments that are tailored according to subtypes are bound to yield better clinical outcomes.
ImpriMed, a veterinary precision medicine company, has used the power of AI and machine learning to pinpoint drugs that can be the most effective against particular cancer subtypes in individual patients. A study published in a top veterinary journal showed that ImpriMed’s novel AI training method resulted in improved drug response prediction.
The cancer study objective
The goal of the study was to gauge the effectiveness of a proposed methodology for predicting the likelihood of clinical remission and progression-free survival in individual patients following treatment. Three different types of data were used to train machine learning models in the study. The data included live cancer cell drug sensitivity, immunophenotype, and patient information
How the study was conducted
A total of 242 dogs suffering from canine lymphoma were selected from a pool of diseased animals. The selected canines were given chemotherapy by board-certified veterinary oncologists. Patients who received at least three of the five drugs that comprise L-(CHOP) chemotherapy were screened within the first four weeks of diagnosis. A subset was then chosen that had data on drug sensitivity, immunophenotype, patient information, and a prognosis of the first 12 weeks after the administration of chemotherapy.
The categorical data obtained were converted into numerical values and rescaled. The compiled data was then used to develop three machine learning models. For each model, the data was split into a train set and test set. The outputs of ML models (classifiers) were the probabilities of achieving clinical remission by the 4th, 8th, or 12th week since the administration of chemotherapy.
The progression-free survival over time for each patient was predicted using the Cox proportional hazard model.
Finally, the performance of the three machine learning models was assessed.
What were the results
The predictive accuracy of the machine learning models was significantly improved by utilizing features from the three different types of data. Superior performance and utility in predicting survival were also observed when compared to the conventional stratification method. It became evident that this novel training methodology can help to improve and personalize cancer care.
What is the inference for veterinarians
The likelihood of clinical remission at various time points with different cancer drugs can aid in making informed treatment decisions. The individualized progression-free survival prospects will also facilitate better preparation for monitoring prognosis and planning follow-up visits after chemotherapy is completed.
ImpriMed’s technology should be an integral part of your lymphoma and leukemia cancer treatment. ImpriMed’s AI-based cancer treatment solution improves clinical outcomes while saving time and money.