The artificial intelligence model could be used to enable more effective care for skin cancer patients and could lead to similar advances in the diagnosis and treatment of other cancers.
Researchers from the University of Helsinki, HUS Comprehensive Cancer Center, Aalto University and Stanford University have developed an artificial intelligence model that predicts which skin cancer patients will benefit from a treatment that activates the immune defense system. In practice, the AI model makes it possible to diagnose skin cancer with a blood test, determine prognosis and target therapies with ever greater precision.
The skin cancer-related study was published in the esteemed journal Nature Communications.
The right medicine for the right patient
Boosting the body’s own defense system has proven to be a particularly effective therapy for skin cancer. The problem with therapies that activate the immune system is the differences between groups of patients: while some patients can be said to be cured, others do not derive any benefit from the treatment.
Previous research has failed to provide doctors with tools to predict who will benefit from treatment that activates the immune system. Correct targeting of therapies is extremely important, as drug therapies are expensive and serious adverse effects are quite common.”
Jani Huuhtanen, PhD and PhD researcher, University of Helsinki and Aalto University
A complex AI model for a simple question
The international research group hypothesized that the immune cells of patients for whom the therapy was ineffective do not recognize the skin cancer as an enemy, which is why the patients do not benefit from the treatment.
Using the AI model, the group analyzed samples from nearly 500 skin cancer patients and compared them to samples from nearly 1,000 healthy individuals. To aid interpretation, the researchers used another AI model developed by the Mark M. Davis lab at Stanford University. From these samples, the researchers simply calculated the number of immune cells that recognized the skin cancer.
As expected, more skin cancer-sensitive defensive cells were found in melanoma patients than in healthy patients.
“This finding may in the future make it possible to identify skin cancer from a blood sample,” says professor of translational hematology Satu Mustjoki from the University of Helsinki.
In addition, skin cancer patients who had more defensive cells that recognized the skin cancer were more likely to benefit from therapies that activate the immune system than those who did not have these cells.
Focusing the AI model on other types of cancer
The use of AI models in medicine has grown exponentially, but applying them to patient care requires long-term collaboration between clinicians and researchers specializing in artificial intelligence.
“In future studies, our goal is to explore the use of the AI model developed now and investigate whether it can predict treatment responses also for new drug therapies against cancer still in development,” says the associate professor of computational biology and machine learning Harri Lähdesmäki from Aalto. University
“Our AI model is agile and adaptable, allowing it to calculate the number of cancer-sensitive defensive cells also in other cancers, such as breast, lung and blood cancers,” adds Jani Huuhtanen.
“All of our research is based on open source software, which makes our AI model available to other researchers and clinicians, and also enables its development,” says Huuhtanen.