Researchers have successfully used two different machine learning algorithms to predict three common symptoms—sleep disturbance, anxiety and depression—experienced by cancer patients undergoing chemotherapy. The team's findings were published in PLOS One.
“While some patients experience very few symptoms, other patients undergoing the same treatment experience multiple co-occurring symptoms that are severe and extremely distressing,” wrote author Nikolaos Papachristou, of the University of Surrey in the U.K., and colleagues.
All 799patients had a diagnosis of breast, gastrointestinal, gynecological or lung cancer and were approached between February 2010 and December 2013. They had all received chemotherapy within the four weeks prior to vetting and were scheduled to receive at least two more cycles of treatment.
The algorithms—Support Vector Regression (SVR) and Non-linear Canonical Correlation Analysis by Neural Networks (n-CCA)—were compared and tested for their ability to predict sleep disturbance, anxiety and depression.
Papachristou et al. noted that SVR did not show “meaningful differences” between the real and predicted values. Additionally, the researchers reported obtaining “fairly similar results” with the n-CCA algorithm.
“The ability to predict the severity of future symptoms in oncology patients will be a powerful tool for oncology clinicians,” the researchers wrote. “Developing computational tools using machine learning techniques will assist clinicians to risk profile patients and implement pre-emptive symptom management interventions.”
Future work, the authors noted, will highlight defining a set of predictors and improving the efficacy of both algorithms.