Cleveland Clinic study seeks to personalize dosages of radiation therapy used to treat cancer by developing artificial intelligence (AI) driven CT scan and health records.
The Current Situation
Today, radiotherapy factors in generic dose prescriptions that preclude an individual’s specific and unique tumor characteristics. Increasingly, researchers consider novel approaches to personalize treatments for greater effectiveness and reduced side effects.
Personalizing the Radiotherapy Dose
The Cleveland Clinic and Siemens Healthineers have developed an AI-infused framework that utilizes a patient’s CT scans and electronic medical records to produce an individualized radiation dose.
The team analyzed the health records of 944 lung cancer patients treated with stereotactic body radiotherapy. 849 patients in an internal study cohort (mean age, 74.1 years; interquartile range, 67.6-80.7; 51% women) and 95 patients in an independent validation cohort (mean age, 76 years; range, 70-82.3; 60% women) were studied.
A high-dose treatment was promoted to achieve local tumor control, while the team simultaneously sought to avoid surgical morbidity in patients afflicted with early-stage lung cancer or lung metastases. The team’s research revealed, that at least in some patient subgroups, studies evidence large numbers of failure rates. There is a growing understanding that a way to potentially mitigate failures would include the ability to personalize adjustment of radiotherapy dosage.
Senior Author Comment
Mohammed Abazeed of the Cleveland Clinic noted, “While highly effective in many clinical settings, radiotherapy can greatly benefit from dose optimization capabilities.” Abazeed continued, “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.”
Deep Profiler: A Deep Neural Network
Deep Profiler is a deep neural network that includes radiomics designed into its algorithmic training logic. The study lead loaded patients’ pre-therapy CT scans into Deep Profiler, which produced a fingerprint, or score, that predicted time-to-event treatment outcomes and estimated classic radiomic features. As reported in Healio, the deep neural network in combination with clinical variables enabled the team to calculate iGray, a patient-specific dose that estimates the probability of treatment failure to be below 5%.
Results revealed significantly higher 3-year cumulative incidence of local radiation treatment failure among patients in the internal cohort with high versus low Deep Profiler Scores.
The image-based deep learning framework, powered by AI technology and a deep neural network called Deep Profiler, appears to be effective for personalizing radiotherapy for patients with lung cancer.
The study was funded by Siemens Healthcare, the National Institutes of Health grant to Dr. Abazeed as well as National Cancer Institute, American Lung Association, Siemens Healthcare and VeloSano (Cleveland Clinic’s flagship philanthropic initiative).
Mohamed E. Abazeed, MD, PhD, radiation oncologist, Cleveland Clinic Taussig Cancer Institute, Researcher at Lerner Research Institute