How AI Can Enhance the Prediction of Immunotherapy

Immunotherapy is using artificial intelligence to predict the progression of the disease with advanced melanoma. 

FREMONT, CA: According to a Cancer Clinical Research study, there is the possibility to predict the outcome of immune checkpoint inhibitors (ICI) therapy by utilizing histology slides and clinical and demographic characteristics of patients. This research aimed to develop a more accessible method for oncologists to identify the reaction to immunotherapy in patients with advanced melanoma.

Many recent studies that show other opportunities for predicting response, but depend on biomarkers, scalability, and demand a considerable amount of resources, are not always plausible. They intend to establish a streamlined process to pre-treatment prognostication by exploiting the available knowledge with the help of routine clinical care.

Two cohort groups were used in the study, the first being a trained group of 121 people who underwent treatment between 2004 and 2018 at New York University. The other cohort is an independent group with 30 patients who received treatment at the University of Vanderbilt.

Researchers created a multivariable classifier that combines clinical data with neutral network predictions, and it was validated on two slide scanners. 

Patients were accurately stratified into the high or low risk of disease development by the classifier. High-risk Vanderbilt patients had worse progression-free survival (PFS) than low-risk ones.

The study also indicated that 50 percent of NYU cohort patients developed disease progression (POD) compared to 64 percent of Vanderbilt cohort patients. A large percentage of the NYU population received monotherapy with anti-CTLA-4, and Vanderbilt patients got anti-PD1-agents.

Researchers have accomplished their hypothesis, and the study shows a suitable method of predicting immunotherapy response. It can be done by incorporating neutral networks classification on histology slides with clinicodemographic information.

With more clinical trials on more extensive databases, the research suggests that the model can be implemented into clinical practice. They note that the shortage of databases is one of the most significant restrictions of this study, and future studies must work on a broader scale.

With such a setting, the quick and readily available method could offer rapid initial assessments to preselect treatment candidates or recognize those who require even more analysis utilizing complementary predictive models.