Peter Faltin

Volume-preserving correction for image registration using free-form deformations

Registration of an image non-rigidly to another one causes deformations, which generally do not preserve the initial volume. Volume preservation is however indispensable for observing tumors in medical images. This paper presents the correction of B-spline based registration to preserve the volume in observed regions. In contrast to other approaches, our solution is not obtained through energy minimization, but by calculating the correction parameters for the deformation directly. Especially for high resolution image data this strategy is very efficient in terms of computational expenses. We derive a closed form solution to optimize the registration with respect to the compression at a single point and then extend the problem to multiple points. Finally we prove also that the correction terms do not have any significant influence on the registration quality.

Markov-Gibbs model based registration of CT lung images using subsampling for the follow-up assessment of pleural thickenings

Examining the growth rate of pleural thickenings in consecutive 3D-CT images requires the matching of identical thickenings in lung images acquired at two different points in time. The thickenings can be subject to strong deformations caused by their growth. This implies that position information should play a major role in finding correspondences. Here, a MGRF approach is presented to determine a rigid transformation. It aligns the lung volumes by maximizing the probability of the regarded lung tissue to fit an offline trained model. To ensure a symmetrical matching of lung surfaces this probability is calculated reciprocally. Using precalculation, strong sub-sampling and a multiscale approach, the required time can be reduced by a factor of about 80, depending on the image resolution. Due to this speed-up, online follow-up assessment is feasible. We show that this approach results in precise registrations which can be used for a reliable matching of lung thickenings.

Projective 3D-reconstruction of uncalibrated endoscopic images

The most common medical diagnostic method for urinary bladder cancer is cystoscopy. This inspection of the bladder is performed by a rigid endoscope, which is usually guided close to the bladder wall. This causes a very limited field of view; difficulty of navigation is aggravated by the usage of angled endoscopes. These factors cause difficulties in orientation and visual control. To overcome this problem, the paper presents a method for extracting 3D information from uncalibrated endoscopic image sequences and for reconstructing the scene content. The method uses the SURF-algorithm to extract features from the images and relates the images by advanced matching. To stabilize the matching, the epipolar geometry is extracted for each image pair using a modified RANSAC-algorithm. Afterwards these matched point pairs are used to generate point triplets over three images and to describe the trifocal geometry. The 3D scene points are determined by applying triangulation to the matched image points. Thus, these points are used to generate a projective 3D reconstruction of the scene, and provide the first step for further metric reconstructions.

Consistent follow-up segmentation of pleural thickenings

Pleural thickenings, which are an indicator for pleural mesothelioma, can be detected and analyzed using thoracic CT-data. They may cover a large area on the pleura surface and can have complex morphologies. Image noise, small deformations and especially the unclear boundary can significantly influence the segmentation process. Two independent segmentations from two points in time might independently suffer from these influences. As a consequence, they cannot be combined for reliable growth rate estimation. Therefore, we suggest a method to introduce consistency between both points in time. The method is carefully designed to estimate small growth from low resolution structures with complex morphologies. First, a surface mesh with segmentation confidence is extracted from one point in time. This mesh is refined considering the segmentation confidence and partial volume. Subsequently, this refined mesh is transferred to the other point in time and growth is tracked with sub-voxel precision.

3D Lung Surface Analysis Towards Segmentation of Pleural Thickenings

Pleural thickenings are connective tissue propagations caused also by a long time exposure to asbestos. They can be an early stage indicator of the malignant pleural mesothelioma. Its diagnosis is timeconsuming and underlies the physician’s subjective judgment. In order to speed-up the diagnosis and to increase the objectivity of the analysis, three fully automatic methods to detect pleural thickenings from CT data are described and compare in this paper. We apply normal vector analysis, second derivate analysis or curvature scale space computation to analyze the lung surface. In the second step we combine a hysteresisthresholding with the principle of the convex hull to segment localized thickenings precisely. These new approaches are presented in order to allow precise and robust detection of pleural mesothelioma in early stage.

Non-rigid surface proximity registration of CT images considering the influence of pleural thickenings

Given two CT thorax images from the same patient taken at two different points in time, a detailed follow-up assessment of pleural thickenings and their growth requires a registration of the regarded image regions. While the spatio-temporal matching of thickenings could be achieved by a rigid registration, the direct visual comparison or the combination of thickening segmentations from different points in time require a more precise registration. We present a new method which provides a non-rigid registration of the 3D image data in the region close to the lung surface, where pleural thickenings are located. A B-spline based approach is used to compensate the non-rigid deformations of the lungs. The control-grid for the B-splines is determined using a non-iterative method, which requires matched feature points from the registered image pair. However, current non-rigid registration methods compensate all changes of the lung surface. This is in our case explicitly undesired for changes caused by pleural thickenings. Therefore, our approach takes the thickenings into account by choosing feature points not directly located on the lung surface. The number of feature points is reduced and only strong features are kept for a 3D block matching.

Registration of Lung Surface Proximity for Assessment of Pleural Thickenings

Follow-up assessment of pleural thickenings requires the comparison of information from different points in time. The investigated image regions must be precisely registered to acquire this information. Since the thickenings’ growth is the target value, this growth should not be compensated by the registration process. We therefore present a nonrigid registration method, which preserves the shape of the thickenings. The deformation of the volume image is carried out using B-splines. With focus on the image regions located around the lung surface, an efficient way of calculating corresponding points combined with the reuse of information from different scale levels leads to the non-rigid registration, which can be performed within a short computation time.

Flattening of the Lung Surface with Temporal Consistency for the Follow-Up Assessment of Pleural Mesothelioma

Malignant pleural mesothelioma is an aggressive tumor of the lung surrounding membrane. The standardized workflow for the assessment comprises an inspection of 3D CT images to detect pleural thickenings which act as indicators for this tumor. Up to now, the visualization of relevant information from the pleura has only been superficially addressed. Current approaches still utilize a slice-wise visualization which does not allow a global assessment of the lung surface. In this publication, we present an approach which enables a planar 2D visualization of the pleura by flattening its surface. A distortion free mapping to a planar representation is generally not possible. The present method determines a planar representation with low distortions directly from a voxel-based surface. For a meaningful follow-up assessment, the consistent representation of a lung from different points in time is highly important. Therefore, the main focus in this publication is to guarantee a consistent representation of the pleura from the same patient extracted from images taken at two different points in time. This temporal consistency is achieved by our newly proposed link of both surfaces during the flattening process. Additionally, a new initialization method which utilizes a flattened lung prototype speeds up the flattening process.

Automated anatomical description of pleural thickening towards improvement of its computer-assisted diagnosis

Pleural thickenings are caused by asbestos exposure and may evolve into malignant pleural mesothelioma. An early diagnosis plays a key role towards an early treatment and an increased survival rate. Today, pleural thickenings are detected by visual inspection of CT data, which is time-consuming and underlies the physicians subjective judgment. A computer-assisted diagnosis system to automatically assess pleural thickenings has been developed, which includes not only a quantitative assessment with respect to size and location, but also enhances this information with an anatomical description, i.e. lung side (left, right), part of pleura (pars costalis, mediastinalis, diaphragmatica, spinalis), as well as vertical (upper, middle, lower) and horizontal (ventral, dorsal) position. For this purpose, a 3D anatomical model of the lung surface has been manually constructed as a 3D atlas. Three registration sub-steps including rigid, affine, and nonrigid registration align the input patient lung to the 3D anatomical atlas model of the lung surface. Finally, each detected pleural thickening is assigned a set of labels describing its anatomical properties. Through this added information, an enhancement to the existing computer-assisted diagnosis system is presented in order to assure a higher precision and reproducible assessment of pleural thickenings, aiming at the diagnosis of the pleural mesothelioma in its early stage.

Extracting and visualizing uncertainties in segmentations from 3D medical data

Assessing surfaces of segmentations extracted from 3D image data for medical purposes requires dedicated extraction and visualization methods. In particular, when assessing follow-up cases, the exact volume and confidence level of the segmentation surface is crucial for medical decision-making. This paper introduces a new processing chain comprising a series of carefully selected and well-matched steps to determine and visualize a segmentation boundary. In a first step, the surface, segmentation confidence and statistical partial volume are extracted. Then, a mesh-based method is applied to determine a refined boundary of the segmented object based on these properties, whilst smoothness, confidence of the surface and partial volume are considered locally. In contrast to existing methods, the proposed approach is able to guarantee the estimated volume for the whole segmentation, which is an important prerequisite for clinical application. Furthermore, a novel visualization method is presented which was specifically designed to simultaneously provide information about 3D morphology, confidence and possible errors. As opposed to classical visualization approaches that take advantage of color and transparency but need some geometric mapping and interpretation from the observer, the proposed scattered visualization utilizes density and scattering, which are much closer and more intuitively related to the original geometric meaning. The presented method is particularly suitable to assess pleural thickenings from follow-up CT images, which further illustrates the potential of the proposed method.