Lung cancer is the most commonly diagnosed cancer in Canada, leading to the most deaths of any other cancer type. Non-small cell lung cancer (NSCLC) is the most common type of lung cancer; for unresectable cases, treatment recommendation involves chemotherapy and radiation therapy (RT). Radiation pneumonitis (RP), a common radiation induced lung injury, is a major dose-limiting side effect in thoracic RT. When high doses of radiation are delivered to lung tissue in individuals susceptible to RP, a number of unfavorable pathological changes can occur including alveolar inflammation and pulmonary fibrosis. In severe RP cases, patients may experience significant respiratory symptoms that require medical intervention, oxygen supplementation or hospitalization, leading to a reduced quality of life and increased risk of death. Radiation pneumonitis can develop into lung fibrosis months to years after radiation treatment, which can potentially leave the patients with permanent impairment of oxygen transfer. Although fatal RP is uncommon, occurring in less than 2% of those suffering with RP, no patient death should be due to side effects of their cancer treatment.


Building a model that is capable of predicting those patients at the greatest risk of developing RP is crucial, as treatments may be altered to lower this risk thus ensuring safer therapy delivery. Currently there are no clinically useful predictive tools to guide dose modifications or to guide alternative approaches which might modify risk of RP. However, research has shown that early diagnosis and treatment of RP can have a positive impact on patient outcome. Radiomics is the conversion of images into mineable high-dimensional data. Radiomics is capable of extracting more complex diagnostic information from routine medical images, including features that are not identifiable or quantifiable by the physician’s eyes alone. Through the use of radiomics, researchers have been able to identify CT-image features that improve differentiation between post-treatment fibrosis and cancer recurrence, after RT for patients with early-stage lung cancer.  Another group of researchers were able to identify a set of image-based texture features that were significantly related to RP development. This demonstrates that the technological advancements in radiomic software are available for the goal of this study to be achieved. Patient outcomes will improve if this model is successful in predicting RP following RT as it will assist clinicians in tailoring the patient’s RT based on the predicted response from the non-invasive CT imaging routinely acquired. It will also allow opportunities for closer follow-up and earlier medical intervention for patients at high risk of developing RP, in order to improve their quality of life and overall outcome.