Research News

Results from a new study led by Cleveland Clinic and published in Lancet Digital Health show that an artificial intelligence (AI) framework can provide individualized radiation dose delivery based on data from patient computerized tomography (CT) scans and electronic health records. This AI framework is the first to use medical scans to inform radiation dose delivery, moving the field forward from using generic dose prescriptions to more personalized treatments.

Currently, radiation therapy is delivered uniformly. The dose prescribed by the physician does not reflect the broad diversity of tumors or individual-level factors that may affect treatment success. The new AI framework begins to account for this variability and provides personalized radiation doses that can reduce the radiotherapy failure probability to less than five percent.

“While highly effective in many clinical settings, radiotherapy can greatly benefit from dose delivery optimization. This framework will help physicians develop data-driven, personalized dosage schedules that can maximize the likelihood of treatment success and mitigate radiation side effects for patients,” says Mohamed Abazeed, MD, PhD, practicing radiation oncologist and lead author on the study.  

The framework was built using CT scans and the electronic health records of 849 lung cancer patients treated with high-dose radiation. Pre-radiotherapy scans were input into a deep learning model called Deep Profiler, which analyzed the scans to create an image signature indicative of radiation outcome. Using sophisticated mathematical modeling, this image signature was combined with data from patient health records—which describe clinical risk factors—to predict sensitivity or resistance and generate a personalized radiation dose value, termed iGray (the “Gray” is a unit of ionized radiation dose).

“The development and validation of this framework is exciting because not only is it the first to utilize medical images to inform radiation dose prescriptions, but it also has the potential to directly impact patient care,” says Dr. Abazeed. “The framework has a low implementation barrier to inform radiotherapy-based clinical trial design and ultimately can be used to deliver radiation therapy tailored to individual patients in everyday clinical practices.”

There are several other factors that set this framework apart from other similar clinical machine learning algorithms and approaches, including classical radiomics. While radiomics extracts quantitative tumor features from medical scans, these features are human-derived and inelastic, meaning they do not conform to the classification task at hand. Deep Profiler is an artificial neural network that deforms or bends the quantitative tumor feature space that is anchored in classical radiomics. In this way, the new framework is more flexible and can better detect scan features that are germane to predicting failure.

Additionally, this framework was built using one of the largest datasets for patients receiving lung radiotherapy, rendering greater accuracy and limiting spurious findings. Lastly, each clinical center can utilize their own CT datasets to customize the framework and tailor it to their specific patient population.

“We believe this form of human augmentation represents a leap forward in radiation precision medicine and that iGray values will be an important tool that radiation oncologists can incorporate into practice.”

This study was funded by a National Institutes of Health KL2 training grant to Dr. Abazeed, the National Cancer Institute, American Lung Association, Siemens Healthcare and VeloSano, Cleveland Clinic’s flagship philanthropic initiative to advance cancer research. Dr. Abazeed has appointments in Taussig Cancer Institute and Lerner Research Institute and is a member of the Case Comprehensive Cancer Center.

Photo credits: Top, Russell Lee; Bottom, Lancet Digital Health, July 2019