Using Artificial Intelligence to Personalize the Dose of Radiation Therapy

A new study led by Cleveland Clinic demonstrates that artificial intelligence (AI) can personalize the dose of radiation therapy used to treat cancer patients, by using health records and medical scans.

Reported in The Lancet Digital Health on 27^th June, 2019, the group of scientists created an AI framework based on electronic health records and computerized tomography (CT) scans of the patient. This is the first-ever AI framework to use medical scans to determine radiation dosage, advancing the field from using generic dose prescriptions to more individualized treatments.

At present, radiation therapy is given equally. The dose given does not reflect variations in individual tumor characteristics or patient-specific factors that may impact treatment success. The AI framework starts to justify this variability and offers personalized radiation doses that can mitigate the treatment failure probability to below 5%.

While highly effective in many clinical settings, radiotherapy can greatly benefit from dose optimization capabilities. 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.

Mohamed Abazeed, MD, PhD, Study Lead Author, and Radiation Oncologist, Taussig Cancer Institute, Cleveland Clinic

Abazeed is also a scientist at the Lerner Research Institute.

The framework was developed using electronic health records and CT scans of 944 lung cancer patients treated with high-dose radiation. Pre-treatment scans were fed into a deep-learning model, which examined the scans to produce an image signature that predicts treatment results.

Using advanced mathematical modeling, this image signature was integrated with data from patient health records—which depicts clinical risk factors—to create a personalized radiation dose.

“The development and validation of this image-based, deep-learning framework is exciting because not only is it the first to use medical images to inform radiation dose prescriptions, but it also has the potential to directly impact patient care,” said Dr. Abazeed. “The framework can ultimately be used to deliver radiation therapy tailored to individual patients in everyday clinical practices.”

Many other factors make this first-of-its-kind framework different from other comparable clinical machine learning algorithms and techniques. The novel technology created by the group employs an artificial neural network that combines traditional methods of machine learning with the power of a contemporary neural network.

The network determines the prior knowledge required to direct predictions about treatment failure. Using the network, it is possible to tune the extent that prior knowledge informs the model. This hybrid method is well-suited for clinical applications as most clinical datasets in individual hospitals are simpler in sample size when compared to non-clinical datasets used for making other well-known AI predictions (i.e. online shopping or ride-sharing).

Furthermore, this framework was developed using one of the largest datasets for patients undergoing lung radiotherapy, offering greater accuracy and restricting false findings. Finally, each clinical center can make use of their own CT datasets to modify the framework and adapt it to their particular patient population.

Machine learning tools, including deep learning, are poised to play an important role in healthcare. This image-based information platform can provide the ability to individualize multiple cancer therapies but more immediately is a leap forward in radiation precision medicine.

Mohamed Abazeed, MD., PhD, Study Lead Author, and Radiation Oncologist, Taussig Cancer Institute, Cleveland Clinic