Approaches for Incorporating Patient-Specific Response in Radiation Therapy

Abstract

In an era of increasingly personalized medical care, radiation therapy (RT) continues to be a prominent and effective focal cancer treatment. However, the standard RT treatment process remains largely uniform, using a one-size-fits-all approach designed to maximize therapeutic benefit to patient populations typically stratified only by tumor type, anatomical location, and extent of the disease. However, additional factors within these populations may lead individual patients to respond differently to RT during and after treatment. Therefore, current RT treatment practices and processes are likely suboptimal for individual patients and could be better tailored to each patient by incorporating additional patient-specific information into the RT planning and treatment evaluation processes. To facilitate this, actionable treatment response models that account for patient variability must be developed and implemented. These models can be formulated though analysis of anatomical and functional changes observed within patient imaging and by evaluating treatment outcomes based on underlying biological factors in addition to the therapeutic intervention. This dissertation includes multiple investigations focused on developing and implementing patient-specific models to support personalized, evidence-based RT. Advancements in anatomical and functional patient imaging have allowed for non-invasive evaluation and analysis of changes throughout the course of RT and follow-up care. Accurate spatial alignment of patient imaging is required to incorporate these new imaging techniques into the RT planning process, evaluate treatment response, and better correlate anatomical and functional changes to therapeutic radiation delivery. However, the ability to link and compare imaging studies before, during, and after treatment is often obscured by significant spatial and volumetric variations caused by physiological and treatment-related factors. To address this, we conducted a series of studies on geometrically-based patient modeling to improve the accuracy of deformable image registration (DIR) and evaluate the clinical impact. First, an intensity-based DIR algorithm was characterized in the context of modern clinical imaging scenarios. Algorithm parameters were evaluated with respect to clinical accuracy metrics and dosimetric impact. Next, radiation dose, in combination with patient factors including tumor location and type, were used to biomechanically model longitudinal liver anatomy changes during RT and follow-up care. A previously developed biomechanical DIR algorithm was modified to incorporate the newly developed liver-response models and was shown to improve spatial and volumetric correlation. Following studies in geometric patient modeling, we studied the direct incorporation of patient-specific response models into the RT planning process. Through a prioritized fluence optimization approach, we implemented the concept of utility-based planning where the optimization objective is to maximize the predicted value of overall treatment utility for a patient, defined by the probability of efficacy (e.g., local control) minus the weighted sum of toxicity probabilities. Implementation of the prioritized utility-based optimization strategy offers an intuitive approach to biological optimization in which planning trade-offs are explicitly optimized. The feasibility of this new planning approach was demonstrated on a cohort of non-small cell lung cancer patients and was shown to improve overall plan utility through situational tumor dose modification and normal tissue dose redistribution.