Deformable Image Registration for Radiation Therapy Planning: Algorithms and Applications

Abstract

Approximately 60% of cancer patients are treated with external beam radiotherapy at some point during disease management. Despite the extended time frame of fractionated therapy (4-6 weeks), radiation therapy planning is carried out based on information that is currently limited to a single 3D anatomical computed tomography scan at the onset of treatment. This concept may result in severe treatment uncertainties, including the irradiation of risk organs and reduced tumor coverage. Repeat 3D single or multi-modality imaging acquired at various time intervals during and after a radiation course provides the opportunity to increase treatment accuracy and precision by optimizing treatment in response to anatomical changes; to improve target delineation through modality-specific complementary tumor representations, and to assess treatment response. Integration of multiple imaging sources into a single patient model requires compensation of geometric differences while maintaining modality-specific differences in information content. Deformable image registration aims to reduce such uncertainties by estimating the spatial relationship between the volume elements of corresponding structures across image data. This paper reviews the algorithmic components of deformation algorithms, and their application to treatment sites with evident geometric changes, including mono-and multi-modal image registration for cancer of the head and neck, lung, liver, and prostate.

Learning anatomy changes from patient populations to create artificial CT images for voxel‐level validation of deformable image registration

“…With recent progress in the development of highly conformal radiotherapy techniques, such as intensity‐modulated radiotherapy (IMRT), volumetric‐modulated arc therapy (VMAT), and proton therapy, deformable image registration (DIR) has become very important to radiation therapy in compensating for nonrigid variations, automatically delineating target volumes and normal tissue contours, and determining dose accumulation for plan evaluation (1‐6) . This trend has been clearly demonstrated in recent years as more and more FDA‐approved commercial DIR tools have become available in clinical radiation therapy, such as MIM Maestro (MIM Software, Cleveland, OH), VelocityAI (Velocity Medical Solutions, Atlas, GA), SPICE in Pinnacle 3 treatment planning system (Philips Healthcare, Cleveland, OH), Smart Segmentation in Eclipse treatment planning system (Varian Medical Systems, Palo Alto, CA), and atlas‐based segmentation in RayStation treatment planning system (RaySearch Laboratories AB, Stockholm, Sweden).…”

Diversifying Multi-Objective Gradient Techniques and their Role in Hybrid Multi-Objective Evolutionary Algorithms for Deformable Medical Image Registration

“…DIR is a medical image processing task that can be of great value for healthcare. Its clinical implementation is still limited and presents many challenges [9]. Currently, a registration outcome is computed based on a single combination of different objectives, using predominantly single-objective gradient methods.…”

Patient-specific model of lung deformation using spatially dependent constitutive parameters

“…Such a technique is aimed at maximizing dosage to the target tumour while limiting exposure to neighbouring healthy organs, thus improving the control of tumour growth and limiting side effects [1]. EBRT techniques include computed tomography (CT), magnetic resonance imaging (MRI), single-photon emission tomography (SPET) and photon emission tomography (PET).…”