Automated and standardized quality planning for prostate cancer radiotherapy
Alejandro Berlin, MD, MSc, Radiation Medicine Program, Princess Margaret Cancer Centre
Thomas Purdie, PhD., Department of Radiation Oncology, University of Toronto
Lead Investigator Bio:
Dr. Alejandro Berlin: I am a radiation oncologist and clinician-investigator at the Princess Margaret Cancer Centre – UHN and the Department of Radiation Oncology at the University of Toronto. My main clinical practice and focus is in the treatment of genitourinary malignancies with personalized approaches combining systemic, external beam radiation and brachytherapy treatments. Particularly interested in innovative clinical trials, novel imaging modalities, translational oncology, and the development and implementation of transformative technologies aiming to disrupt the status quo. At this early stage of my career, my overarching goal is to establish new individualized treatments for patients with prostate cancer.
Radiotherapy (RT) planning uses a 3D image of the region to be treated (i.e. CT scan), and a complex computational process for defining the RT beams’ characteristics (i.e. energy, shape) required to ensure the cancer gets the prescribed radiation dose while surrounding normal tissues get as little as possible. Generating high-quality RT plans demands specialized expertise and significant time investment (i.e. hours to days) for each individual plan. However, variability of these factors results in discrepancies within/between Institutions, exposing patients to sub-optimal plans that could jeopardize their treatment’s results (i.e. decreased cure, increased side-effects). We have developed a unique method that capitalizes hundreds of peer-reviewed high-quality plans, and progressively learns without being explicitly programmed (i.e. machine-learning). The method collapses RT planning to only 15-25 minutes without necessitating user input, rendering RT plans of consistent and reproducible quality. In this proposal, we will generate prostate cancer (PCa) specific libraries and train specific planning algorithms. Subsequently, we will use the automated planning framework for upcoming patients and compare its performance to the conventional user-based method. We aim to unveil a new RT planning method for PCa that makes the most out of reference centers’ expertise, reduces variability and errors, and enables highly individualized and cost-effective PCa care.
Radiation therapy (RT) planning (RTp) is one of the key determinants of the quality and success of curative RT. However, the current RTp practice is technically challenging and resource-intensive, requiring hours of iterative processes to generate a plan. Both experience and invested time impact resultant RT plan quality, while variability of these causes discrepancies within/between Institutions. As a result, sub-optimal plans can be used clinically which in turn may jeopardize clinical outcomes (i.e. decrease cure, increase side-effects). Current pressing need for RT capacity creates value for novel efficient, consistent and sustainable frameworks for high quality RTp. Our group has developed a unique machine learning-based method that capitalizes large libraries of previously approved and quality-assured RT plans, collapsing the planning time to only 15-25 minutes without necessitating user’s input. Importantly, it provides a systematic solution of consistent and reproducible plan quality. Herein, we aim to train prostate cancer-specific algorithms for different clinical scenarios (i.e. primary treatment, post-surgery RT) and prospectively validate this automated framework, its integration into clinical routine, and its potential for dissemination. Our work has various potential impacts: i) enabling a workflow that meets increasing demand and makes better use of limited resources; ii) standardizing the RTp process based on the expertise of established reference centres, reducing variability and errors; and iii) providing a readily translatable and sustainable framework for highly personalized and cost-effective cancer care.
Impact on prostate cancer patients:
The proposed work will generate a sustainable, expert learning framework for delivering high-quality and cost-effective cancer care. We will have qualitative and quantitative prospective data to guide the broader implementation of the automated planning method, serving as a proof-of-concept and providing a process map for seamless wider application both within and across institutions. We expect to have a direct positive benefit for patients by increasing capacity and reducing costs of delivering RT, therefore expanding access to modern state-of-the-art therapeutics for prostate cancer patients, regardless of their geographical location. As well, evidence-based adoption of a validated tool as the one proposed will help standardizing planning based on expert’s previous experience and consensus, and could potentially reduce errors and lower treatment-related patient side effects. The latter will be our group’s working hypothesis for subsequent efforts after the present initiative. We envisage that this pivotal study will be followed by similar endeavors, expanding to other GU treatment scenarios as well as other treatment sites and cancer centers. If successful, this germinal effort will have a pollinizing effect in the RT field by showing value for the development, translation and validation of advanced technologies into routine clinical practice for patients with prostate cancer.