Johann Drexl

Integrated wavelets for enhancement of microcalcifications in digital mammography

This paper presents a new algorithm for enhancement of microcalcifications in mammograms. The main novelty is the application of techniques we have developed for construction of filterbanks derived from the continuous wavelet transform. These discrete wavelet decompositions, called integrated wavelets, are optimally designed for enhancement of multiscale structures in images. Furthermore, we use a model based approach to refine existing methods for general enhancement of mammograms resulting in a more specific enhancement of microcalcifications. We present results of our method and compare them with known algorithms. Finally, we want to indicate how these techniques can also be applied to the detection of microcalcifications. Our algorithm was positively evaluated in a clinical study. It has been implemented in a mammography workstation designed for soft-copy reading of digital mammograms developed by IMAGETOOL, Germany.

GPU Accelerated Image Registration in Two and Three Dimensions

Medical image registration tasks of large volume datasets, especially in the non-rigid case, often put a heavy burden on computing resources. GPUs are a promising new approach to address computational intensive image processing tasks. We investigate recently introduced GPU hardware features that accelerate 2D and 3D rigid and nonrigid registration tasks. Our implementation is entirely GPU based.

Prevalence of Potential Retrograde Embolization Pathways in the Proximal Descending Aorta in Stroke Patients and Controls

Background: Retrograde diastolic blood flow in the proximal descending aorta (DAo) connecting complex plaques (≥4 mm thick) with brain-supplying supra-aortic arteries may constitute a source of stroke. Yet, data only from high-risk populations (cryptogenic stroke patients with aortic atheroma ≥3 mm) regarding the prevalence of this potential stroke mechanism are available. We aimed to quantify the frequency of this mechanism in unselected patients with cryptogenic stroke after routine diagnostics and controls without a history of stroke. Methods: 88 patients (67 stroke patients, 21 cardiac controls) were prospectively included. 3D T1-weighted bright blood MRI of the aorta was applied for the detection of complex DAo atheroma. ECG-triggered and navigator-gated 4D flow MRI allowed measuring time-resolved 3D blood flow in vivo. Potential retrograde embolization pathways were defined as the co-occurrence of complex plaques and retrograde blood flow in the DAo reaching the outlet of (a) the left subclavian artery, (b) the left common carotid artery, or/and (c) the brachiocephalic trunk. The frequency of these pathways was analyzed by importing 2D plaque images into 3D blood flow visualization software. Results: Complex DAo plaques were more frequent in stroke patients (44 in 31/67 patients (46.3%) vs. 5 in 4/21 controls (19.1%); p = 0.039), especially in older patients (29/46 (63.04%) patients ≥60 years of age with 41 plaques vs. 2/21 (9.14%) patients <60 years of age with 3 plaques; p < 0.001). Contrary to our assumption, retrograde diastolic blood flow at the DAo occurred in every patient irrespective of the existence of plaques with a similar extent in both groups (26 ± 14 vs. 32 ± 18 mm; p = 0.114). Therefore, only the higher prevalence of complex DAo plaques in stroke patients resulted in a three times higher frequency of potential retrograde embolization pathways compared to controls (22/67 (32.8%) vs. 2/21 (9.5%) controls; p = 0.048). Conclusions: This study revealed that retrograde flow in the descending aorta is a common phenomenon not only in stroke patients. The existence of potential retrograde embolization pathways depends mainly on the occurrence of complex plaques in the area 0 to ∼30 mm behind the outlet of the left subclavian artery, which is exposed to flow reversal. In conclusion, we have shown that the frequency of potential retrograde embolization pathways was significantly higher in stroke patients suggesting that this mechanism may play a role in retrograde brain embolism.

Rapid image recognition of body parts scanned in computed tomography datasets

AimAutomatic CT dataset classification is important to efficiently create reliable database annotations, especially when large collections of scans must be analyzed.MethodAn automated segmentation and labeling algorithm was developed based on a fast patient segmentation and extraction of statistical density class features from the CT data. The method also delivers classifications of image noise level and patient size. The approach is based on image information only and uses an approximate patient contour detection and statistical features of the density distribution. These are obtained from a slice-wise analysis of the areas filled by various materials related to certain density classes and the spatial spread of each class. The resulting families of curves are subsequently classified using rules derived from knowledge about features of the human anatomy.ResultsThe method was successfully applied to more than 5,000 CT datasets. Evaluation was performed via expert visual inspection of screenshots showing classification results and detected characteristic positions along the main body axis. Accuracy per body region was very satisfactory in the trunk (lung/liver >99.5% detection rate, presence of abdomen >97% or pelvis >95.8%) improvements are required for zoomed scans.ConclusionThe method performed very reliably. A test on 1,860 CT datasets collected from an oncological trial showed that the method is feasible, efficient, and is promising as an automated tool for image post-processing.

HWT - hybrid watershed transform: optimal combination of hierarchical interactive and automated image segmentation

In quantitative medical imaging and therapy planning, the optimal combination of automated and interactively defined information is crucial for image segmentation methods to be both efficient and effective. We propose to combine an efficient hierarchical region merging scheme that collects per-region statistics across hierarchy levels with a trainable classification engine that facilitates automated region labeling based on an arbitrary number of reference segmentations. When applying the classification engine, we propose to use a corridor of non-classified regions resulting in a sparse labeling with extremely low false-classification rate, and to attribute labels to the remaining basins through successive merging with ready-labeled basins. The proposed hierarchical region merging scheme also permits to efficiently include interactively defined labels. We denominate this general approach as Hybrid Hierarchical Interactive Image Segmentation Scheme (HIS2). More specifically, we present an extension of the Interactive Watershed Transform, which we combine with a trainable two-class Support Vector Machine based on Gaussian radial basis functions. Finally, we present a novel asymmetric marker scheme, which provides a powerful means of regionally correcting remaining inaccuracies while preserving full detail of the automatic labeling procedure. We denominate the complete algorithm as Hybrid Watershed Transform (HWT), which we apply to one challenging segmentation problem in clinical imaging, namely efficient bone removal in large computed tomography angiographic data sets. Efficiency and accuracy of the proposed methodology is evaluated on multi-slice images from nine different sites. As a result, its ability to rapidly and automatically generate robust and precise segmentation results in combination with a versatile manual correction mechanism could be proven without requiring specific anatomical or geometrical models.

Visualization of microcalcifications by full-field digital mammography using a wavelet algorithm.

Abstract Purpose : A method is presented for visualizing microcalcifications by full-field mammography using wavelet frames, an enhancement operator, and a suitable reconstruction technique. Material and methods : The computer algorithm is tested on 24 digital mammographies (Senographe 2000D, GE Medical Systems) containing microcalcifications. This algorithm highlights microcalcification-like structures by means of a filter bank that is based on wavelet decomposition. What is new about this filter bank is that the discretization of the continuous wavelet transformation is specifically adjusted to the problem of image analysis, allowing for a better discrimination of microcalcifications from the image background. Results : In all cases (24/24), microcalcifications were depicted with a markedly higher contrast. Conclusion : Special computer algorithms improve the visualization of microcalcifications. This means in principle, that there is a potential to improve diagnostic assessment.

Computerassistierte Radiologie (CAR) in der Mammographie

Die Treffsicherheit bei der Beurteilung pathologischer Zustande an Hand radiologischer Bilder ist von apparativen Einflusgrosen und subjektiven Wahrnehmungen gepragt. In der vorliegenden Arbeit stellen wir eine die Befundung objektivierende Methode vor, mit deren Hilfe computerassistiert speziell mikrokalkformige Strukturen in Mammographien hervorgehoben werden konnen. Die Uberprufung der Methode an uber 100 in einem DICOM-Archiv des IT-ASSIST gespeicherten Bildern mit gesicherten Befunden zeigt, dass sie unterstutzend beim Auffinden von Mikrokalk im Rahmen der Mammographie vorteilhaft einsetzbar ist.

MammoInsight Computer Assisted Detection: Performance Study with Large Database

We extend our previous publications by presenting results on a large number of mammograms digitized with a laser scanner and analyzed by our MammoInsight CAD-system for the detection of clustered microcalcifications. We measure sensitivities on a per breast basis and on a per cluster basis for our system and show how they interrelate. We compare the performance of our system to the state-of-the art in the research literature. Finally, we show that our CAD-system is able to obtain a high sensitivity per breast, justifying it’s use in a clinical environment.

A fast and noise-robust method for computation of intravascular pressure difference maps from 4d PC-MRI data

Noninvasive blood flow measurements by 4D flow-sensitive MRI can be used to compute the intravascular distribution of blood pressure. In this work, we present an efficient algorithm for this task, based on the Navier-Stokes equations including zero-divergence condition for the velocity field. Its accuracy and robustness is investigated on two different CFD-based software phantoms. The method has been integrated into research software for analysis of clinical 4D flow measurements, and is tested on six patients with aortic coarctation. The pressure drop across the stenosis is quantified and coincides well with published results from a previously validated solution algorithm.

A tool for the interactive analysis and exploration of in-vivo haemodynamics from 4D PC MRI

Background Time-resolved 3D (4D) phase contrast (PC) MRI allows deriving anatomical as well as functional information in the cardiovascular system. Progress in 4D MR techniques now facilitates volumetric, 3-directional, cine PC MRI data in reasonable scan times. The analysis of these data is, however, a challenging task because of the data complexity (3 spatial dimensions, 3 velocity directions, time) and the many processing steps required. Current evaluations of 4D PC MRI data often use a combination of home built and commercial tools, which are tailored to often time consuming (up to several hours) workflows for special research questions. The purpose of this study was to develop and evaluate a novel analysis tool that allows a fast interactive exploration of patient-specific 4D-PC hemodynamics.