3-D lung deformation and function from respiratory-gated 4-D x-ray CT images : application to radiation treatment planning.

Abstract

Many lung diseases or injuries can cause biomechanical or material property changes that can alter lung function. While the mechanical changes associated with the change of the material properties originate at a regional level, they remain largely asymptomatic and are invisible to global measures of lung function until they have advanced significantly and have aggregated. In the realm of external beam radiation therapy of patients suffering from lung cancer, determination of patterns of pre- and post-treatment motion, and measures of regional and global lung elasticity and function are clinically relevant. In this dissertation, we demonstrate that 4-D CT derived ventilation images, including mechanical strain, provide an accurate and physiologically relevant assessment of regional pulmonary function which may be incorporated into the treatment planning process. Our contributions are as follows: (i) A new volumetric deformable image registration technique based on 3-D optical flow (MOFID) has been designed and implemented which permits the possibility of enforcing physical constraints on the numerical solutions for computing motion field from respiratory-gated 4-D CT thoracic images. The proposed optical flow framework is an accurate motion model for the thoracic CT registration problem. (ii) A large displacement landmark-base elastic registration method has been devised for thoracic CT volumetric image sets containing large deformations or changes, as encountered for example in registration of pre-treatment and post-treatment images or multi-modality registration. (iii) Based on deformation maps from MOFID, a novel framework for regional quantification of mechanical strain as an index of lung functionality has been formulated for measurement of regional pulmonary function. (iv) In a cohort consisting of seven patients with non-small cell lung cancer, validation of physiologic accuracy of the 4-D CT derived quantitative images including Jacobian metric of ventilation, VJac and principal strains, ( Vl1,Vl2 ,Vl3 ), has been performed through correlation of the derived measures with SPECT ventilation and perfusion scans. The statistical correlations with SPECT have shown that the maximum principal strain pulmonary function map derived from MOFID, outperforms all previously established ventilation metrics from 4D-CT. It is hypothesized that use of CT-derived ventilation images in the treatment planning process will help predict and prevent pulmonary toxicity due to radiation treatment. It is also hypothesized that measures of regional and global lung elasticity and function obtained during the course of treatment may be used to adapt radiation treatment. Having objective methods with which to assess pre-treatment global and regional lung function and biomechanical properties, the radiation treatment dose can potentially be escalated to improve tumor response and local control.