Abstract:
Lung ventilation and perfusion analyses using chest imaging methods require a correct segmentation of the lung to offer anatomical landmarks for the physiological data. An automatic segmentation approach simplifies and accelerates the analysis. However, the segmentation of the lungs has shown to be difficult if collapsed areas are present that tend to share similar gray values with surrounding non-pulmonary tissue. Our goal was to develop an automatic segmentation algorithm that is able to approximate dorsal lung boundaries even if alveolar collapse is present in the dependent lung areas adjacent to the pleura. Computed tomography data acquired in five supine pigs with injured lungs were used for this purpose. First, healthy lung tissue was segmented using a standard 3D region growing algorithm. Further, the bones in the chest wall surrounding the lungs were segmented to find the contact points of ribs and pleura. Artificial boundaries of the dorsal lung were set by spline interpolation through these contact points. Segmentation masks of the entire lung including the collapsed regions were created by combining the splines with the segmentation masks of the healthy lung tissue through multiple morphological operations. The automatically segmented images were then evaluated by comparing them to manual segmentations and determining the Dice similarity coefficients (DSC) as a similarity measure. The developed method was able to accurately segment the lungs including the collapsed regions (DSCs over 0.96).
Abstract:
Radiofrequency ablation is the gold standard for treating cardiac arrhythmias. However, the success rate of this procedure depends on numerous parameters. Wet lab experiments provide the opportunity to investigate cardiac electrophysiology under reproducible conditions. To evaluate the electrophysiological changes of ablated myocardium in these studies it is necessary to consider the three-dimensional (3D) geometry of the lesions. For this purpose, we investigated the usage of different magnetic resonance imaging (MRI) sequences as well as an image processing procedure to analyze in-vitro preparations. To differentiate signal intensities between nonablated and ablated tissue we evaluated FISP (fast imaging with steady-state precession; delivering dominantly T1-weighted images) and RARE (rapid acquisition with relaxation enhancement; delivering dominantly T2-weighted images). After image processing, the ablated tissue was segmented in each image slice forming a 3D volume. The geometry of the lesion was modeled by the boundary of this volume. It was generally feasible to distinguish between healthy myocardium and ablated tissue as well as to determine lesion transmurality. The analysis of the reconstructed lesion geometries from FISP and RARE MRI showed a high agreement, however T2-weighted sequences showed larger lesion volumes as well as higher variations in segmentation compared to T1- mapping. FISP with higher quality may be used to reconstruct the 3D geometry of the ablation lesions.
Abstract:
This thesis discusses methods developed for two separate medical image processing problems.Many cardiac arrhythmias, such as atrial fibrillation (AF), are currently treated with radiofrequency ablation (RFA). However, this therapy method has high recurrence rates. The success of the ablation depends on the formation of reliably transmural lesions. In order to improve the success rate of the RFA procedure, the electrophysiological changes in cardiac tissue are investigated which occur with lesion formation. Electrical and fluorescence-optical measurements are made on myocardial tissue from rats. In this thesis, magnetic resonance imaging (MRI) data sets of rat ventricles containing acute ablation lesions were used to reconstruct the three-dimensional (3D) shape of the lesions and to determine their transmurality. The lesion areas in the MRI slices were manually segmented after preprocessing was performed to increase lesion visibility. The 3D geometry of the lesions was then modeled by creating alpha shapes which comprised the segmented regions of the image slices. The analysis of slice segmentation and lesion volume results showed that the low image quality led to high variations in segmentation and lesion volumes. The implemented methods can be applied on new data sets with higher quality to obtain better results.For the analysis of the perfusion in injured swine lungs using electrical impedance tomography (EIT), the study animals were captured in a computed tomography (CT) scanner simultaneously with the EIT measurements to obtain reference data for the validation of the results. The goal of this thesis was to implement an algorithm for the segmentation of the lungs including lesions from the CT images. A semi-automatic segmentation method was developed which was able to detect lung boundaries even if lesions were present in the dorsal regions of the lungs touching the pleura. For this purpose, a region growing algorithm was combined with a spline interpolation approximating the pleura through points detected on the rib surfaces. The developed method delivered accurate segmentation results for the lung regions which where captured with the EIT measurements. The attained segmentation masks had Dice similarity coefficients over 0.96 in reference to manual segmentations performed by physicians.