By Emily Sokol, Health IT Analytics | July 11, 2019
Artificial intelligence (AI) developed at the Cleveland Clinic uses machine learning to combine medical scans and electronic health records (EHR) generating personalized radiation therapy doses for cancer patients.
Traditional radiation therapy is delivered to patients uniformly. Dosing guidelines do not take specific information about a patient’s individual risk factors or tumor characteristics into consideration when recommending treatment. However, the Cleveland Clinic’s novel method marks a change towards a more personalized approach.
By using a patient’s medical imaging data and clinical risk factors, their model generates a unique radiation dose for each patient. This strategy can ultimately help minimize negative side effects for patients and reduce treatment failures to less than 5%.
Published in The Lancet Digital Health, the study used pre-treatment scans from 944 lung cancer patients scheduled to be treated with high-dose radiation. The patient’s computerized tomography (CT) scans were extracted into a deep-learning model that analyzed the images and created image signatures to predict future treatment outcomes. Mathematical modeling combined the image signature with the patient’s clinical risk factors from medical records to generate 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,” explained the study’s lead author, Mohamed Abazeed, MD, PhD, in an earlier press release. “The framework can ultimately be used to deliver radiation therapy tailored to individual patients in everyday clinical practices.”
The new approach sets itself apart from other AI technologies by expanding traditional machine learning algorithms and methodologies. The technology combines machine learning with artificial neural networks. The networks then determine how much previous knowledge will be used to guide predictions about future treatment failure. The network can modify the levels of prior knowledge used when generating predictions based on the neural network, making it adaptable to different clinical settings.
Because the framework is adaptable to different clinical care settings, each hospital or clinic can modify the model to best fit its available datasets and optimize the response for its specific patient population. Not only does the model customize radiation treatment therapies for individual patients, but it can also adjust its structure for the unique patient populations each care setting sees.
The framework was developed using one of the largest datasets of patients receiving lung radiotherapy. Therefore, it has greater accuracy than other models and the likelihood of false findings is low.
“While highly effective in many clinical settings, radiotherapy can greatly benefit from dose optimization capabilities,” stated Abazeed. “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.”
This announcement comes shortly after the American Medical Association endorsed several new policies in support of artificial intelligence in medical practice.
“Machine learning tools, including deep learning, are poised to play an important role in healthcare,” stated Abazeed. “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.”
The study was conducted at the Cleveland Clinic in collaboration with Siemens Healthcare to advance cancer research. The work was funded by the National Institutes of Health, the National Cancer Institute, the American Lung Association, Siemens Healthcare, and VeloSano, the Cleveland Clinic’s philanthropic initiative.