GE HealthCare said its artificial intelligence programs were able to help predict cancer patients’ responses to immunotherapies by finding patterns within routinely collected clinical data.
Developed through a yearslong collaboration with Vanderbilt University Medical Center (VUMC), the models were able to parse electronic medical records and digest real-world information such as diagnosis codes and certain medication regimens; additional, manually entered inputs included the patient’s smoking history and the number of previous immune checkpoint inhibitor drugs they had taken.
According to the company, the algorithms were able to deliver 70% to 80% accuracy in forecasting efficacy outcomes and the likelihood of unwanted side effects—across a range of different cancer types, including melanoma and lung or genitourinary cancers—by analyzing deidentified demographic, genomic, tumor, cellular, proteomic and imaging data collected from more than 2,200 VUMC patients.
That includes rates of hepatitis, colitis and pneumonitis—the three most commonly severe adverse events related to the immune system—as well as one-year overall survival.
The researchers said current methods for predicting the effectiveness of different immunotherapies rely on collecting novel biomarkers that may not have been thoroughly studied in large, diverse populations.
Meanwhile, the current standard of care may include trial-and-error approaches and monitoring for potentially dangerous toxicities. The research team’s work was published in the Journal of Clinical Oncology Clinical Cancer Informatics.
“We focused primarily on this routinely collected structured data to build predictive models with the goal that these models would be able to be implemented in any clinical setting,” Travis Osterman, director of cancer clinical informatics at Vanderbilt-Ingram Cancer Center, said in a statement.
GE HealthCare said it is evaluating plans to commercialize the AI technology, pending regulatory approvals, to help support everyday patient care as well as in biopharma drug development programs.
“We want to use AI to personalize predictions and provide decision support for the clinician in determining appropriate therapies,” said Jan Wolber, global digital product leader in GE HealthCare’s pharmaceutical diagnostics division. The company also said its scalable AI methodology could have the potential for use in other areas such as neurology or cardiology.
GE HealthCare signed on to a five-year partnership with Vanderbilt to develop these AI diagnostic models in early 2019, alongside plans to explore new PET imaging tracers to help physicians select cancer patients for clinical trials and monitor the success of immunotherapies.