Application of a novel deformable image registration technique to facilitate classification, tracking and targeting of tumor and normal tissue

Abstract

The integration of 4D information in the radiation treatment process requires a model of the patient during each treatment fraction, over the course of radiation treatment, and through post-radiation follow-up in addition to the integration of this knowledge into the process of dose accumulation. The aims of this study are to develop a deformable image registration technique which facilitates the integration of (a) tissue classification on different imaging modalities, (b) temporal response of tumor and normal tissue following radiation treatment, (c) image guided treatment, and (d) 4D dose accumulation. A novel deformable image registration technique based on finite element modelling (FEM) and surface projection has been developed. Finite element models of the patient are constructed from image contours. The biomechanical models associated with the FEM provides a physical-based method to model the deformation associated with physiological processes, such as respiration, stomach, bladder, and rectal filling, as well as change in patient position and temporal and spatial response of tissue following radiation treatment. Multi-organ deformable registration is achieved by combining surface projection alignment on selected regions of interest (ROIs) with tissue specific biomechanical material models and surface interfaces between organs. Biomechanical properties and interfaces were optimized through iterative testing on several patients to determine the best-fit parameters for the patient population. In addition to modelling physical phenomenon, FEM provides an inherent method of tracking regions of interest using the elements associated with the model. An initial scan is designated as the base scan to create a FEM of the patient. This model is then deformed into the representation of the patient at the next imaging or treatment session. This allows for direct tracking of any specified region over the entire treatment and follow-up time span to determine the dose received, motion, change in volume, difference in tissue classification, and response to treatment. The technique was demonstrated on a metastatic liver cancer patient who had inhale and exhale CT studies, planning MR scans, and multiple follow-up CT scans. The patient’s CT scan at exhale breath hold was used to construct the base model of the patient. Deformable registration related the base model of the patient to the subsequent scans including CT scan at inhale breath hold, MR scans at inhale and exhale breath hold, and a follow-up CT scan at inhale breath hold 6 months after the completion of radiation treatment. Deformable registration of the CT-base model at exhale to the inhale CT determined the extent of motion and deformation of the liver, tumor, and surrounding organs, allowing for accurate calculation of PTV margins, 4D dose calculation, and quantification of the benefit of modified breathing during treatment or treatment delivery. Following the deformable registration of the CT-base model to the exhale MR scan, to account for changes in position as well as differences in exhale position, the spatial difference between the GTV at CT and MR was calculated. The normal tissue and tumor response was quantified by performing deformable registration of the CT-based model of the patient to the follow-up CT. The method appears to be robust, successfully performing deformable registration on all cases tested, and accurate, as the average error in alignment for the demonstrated patient was 2.5 mm in all directions, similar to the reproducibility of the measurement tool. This novel, 4D deformable image registration technique incorporates the ability to compare tissue classifications, track changes in tumor and normal tissue, perform image guided treatment, and compute 4D dose accumulation into one methodology. The technique has widespread applications including investigation of spatial and temporal normal tissue toxicity following radiation and development of deformable organ models for image guidance.