Abstract: Deformable image registration (DIR) is a cross-cutting technology with diagnostic and therapeutic medical applications. DIR algorithms were first developed in computer vision research to estimate motion between a source and target image, the resulting registered image visually appears similar to the target image. For medical applications the goal in applying DIR is to obtain an accurate spatial registration of the underlying anatomy and not simply image similarity. We developed a statistical framework for quantitative evaluation of DIR spatial accuracy based on large samples of expert-determined landmark features. Central to this framework is the statistical relationship between the number of landmark points required to assess spatial accuracy, the desired uncertainty range of the mean error, and an a priori estimated behavior of the DIR. DIR is at the heart of our strategy to quantify COPD small airway disease air-trapping and four dimensional computed tomography (4D CT) ventilation. The optimal DIR algorithm and its spatial accuracy in registering the underlying anatomy should be assessed for each application. We will develop and test new DIR algorithms for exhale and inhale breath-hold CT (eBH-CT &iBH-CT) images pairs (COPD air trapping evaluation) and for 4D CT images (4D CT ventilation). Current CT image analysis methods for COPD evaluation focus on the separate anatomic evaluation of the eBH-CT &iBH-CT images. They are unable to find air-trapping due to bronchiolitis alone. We propose to evaluate the eBH- &iBH CT image pairs simultaneously using DIR to link the two to identify regions of air-trapping due to both emphysema and bronchiolitis. Next, to continue our development of ventilation imaging derived from 4D CT, we will test the ability of 4D CT ventilation image guidance to reduce pulmonary function loss after radiotherapy in a randomized phase II trial for non-small cell lung cancer patients. Public Health Relevance: This study will develop novel image registration methods and their application, with an emphasis on application specific validation. With this technology we will develop and test methods to find air-trapping in chronic obstructive pulmonary disease patients. We will test our novel ventilation imaging method in radiation treatment planning to reduce normal lung injury after treatment for lung cancer.
Showing the most recent 10 out of 16 publications
