Thomas Pralow

A Stitching Algorithm for Automatic Registration of Digital Radiographs

In digital radiography oversized radiographs have to be assembled from multiple spatially overlapping exposures. We present an algorithm for fast automatic registration of these radiographs. An external feature is brought into the radiographs to facilitate the reconstruction. Pivotal for this algorithm is an actual interpretation of this feature instead of a simple detection. It possesses strong robustness against noise, feature masking and feature displacement. Evaluation has been performed on 2000 pairs of clinical radiographs. The proposed algorithm proved to be a powerful enhancement of established automatic registration algorithms.

In-vivo X-ray Dark-Field Chest Radiography of a Pig

X-ray chest radiography is an inexpensive and broadly available tool for initial assessment of the lung in clinical routine, but typically lacks diagnostic sensitivity for detection of pulmonary diseases in their early stages. Recent X-ray dark-field (XDF) imaging studies on mice have shown significant improvements in imaging-based lung diagnostics. Especially in the case of early diagnosis of chronic obstructive pulmonary disease (COPD), XDF imaging clearly outperforms conventional radiography. However, a translation of this technique towards the investigation of larger mammals and finally humans has not yet been achieved. In this letter, we present the first in-vivo XDF full-field chest radiographs (32 × 35 cm2) of a living pig, acquired with clinically compatible parameters (40 s scan time, approx. 80 µSv dose). For imaging, we developed a novel high-energy XDF system that overcomes the limitations of currently established setups. Our XDF radiographs yield sufficiently high image quality to enable radiographic evaluation of the lungs. We consider this a milestone in the bench-to-bedside translation of XDF imaging and expect XDF imaging to become an invaluable tool in clinical practice, both as a general chest X-ray modality and as a dedicated tool for high-risk patients affected by smoking, industrial work and indoor cooking.

Robust and Fast Estimation of Signal-Dependent Noise in Medical X-Ray Image Sequences

We present a practice-oriented, i.e. fast and robust, estimator for strong signal-dependent noise in medical low-dose X-ray images. Structure estimation by median filtering has shown to be superior to linear binomial filtering. Falsifications due to remaining structure in the estimated noise image are significantly reduced by iterative outlier removal.

Real-Time denoising of medical x-ray image sequences: three entirely different approaches

Low-dose X-ray image sequences exhibit severe signal-dependent noise that must be reduced in real-time while, at the same time, preserving diagnostic structures and avoiding artifacts. We propose three different methods with applications beyond medical image processing. Major contributions are innovative motion detection based on independent binarization of positive and negative temporal differences, real-time multiscale nonlinear diffusion in the presence of severe signal-dependent noise, and multi-resolution inter-scale correlation in shift-dependent pyramids. All methods exhibit excellent performance over a broad range of noise, detail, and contrast levels. As performance in medical imaging depends to a large degree on the type of intervention and individual preferences of medical staff, no method is generally superior and all methods are considered for the next generation of fluoroscopy systems.

Medical x-ray image enhancement by intra-image and inter-image similarity

In medical X-ray examinations, images suffer considerably from severe, signal-dependent noise as a result of the effort to keep applied doses as low as possible. This noise can be seen as an additive signal that degrades image quality and might disguise valuable content. Lost information has to be restored in a post-processing step. The crucial aspect of filtering medical images is preservation of edges and texture on the one hand and removing noise on the other hand. Classical smoothing filters, such as Gaussian or box filtering. are data-independent and equally blur the image content. State-of-the-art methods currently make use of local neighborhoods or global image statistics. However, exploiting global self-similarity within an image and inter-image similarity for subsequent frames of a sequence bears an unused potential for image restoration. We introduce a non-local filter with data-dependent response that closes the gap between local filtering and stochastic methods. The filter is based on the non-local means approach proposed by Buades1 et al. and is similar to bilateral filtering. In order to apply this approach to medical data, we heavily reduce the computational costs incurred by the original approach. Thus it is possible to interactively enhance single frames or selected regions of interest within a sequence. The proposed filter is applicable for time-domain filtering without the need for accurate motion estimation. Hence it can be seen as a general solution for filtering 2D as well as 2D+t X-ray image data.

An Algorithm for Automatic Stitching of CR X-ray Images

Stitching of X-ray images is of interest in cases of disease patterns like scoliosis [1] or asymmetries in the structure of leg bones. Under those circumstances a measurement of the leg or the spine as a whole is necessary. As these objects of interest exceed the maximum size of the normal radiographic image recording device, two or three radiographs have to be joined. The manual procedure typically is very time consuming and needs training. Therefore, an algorithm has been developed that stitches the radiographs automatically. In the project presented here, it has been concentrated on CR images (CR = computed radiography).

Ruler-Based Automatic Stitching of Spatially Overlapping Radiographs

We present an algorithm for fast automatic registration of spatially overlapping radiographs. It possesses strong robustness against noise, feature masking and feature displacement. Pivotal for this algorithm is an actual interpretation of the stitching feature instead of a simple detection. The proposed method has been evaluated on 3000 clinical radiographs and proved to be a powerful enhancement of established automatic registration algorithms.

Intelligent feature selection for model-based bone segmentation in digital radiographs

In this paper we propose a method to enhance Active Shape Model based bone segmentation. One major weakness of the classic algorithm is the use of a single dedicated image feature. However to model the variation of image content along the object boundaries it is more suitable to use different features for different regions. We derive an automatic intelligent selection of these features and integrate it into the classic Active Shape Model segmentation. We evaluated the proposed algorithm on the task of delineating bone structures in more than 150 clinical radiographs of the lower extremity and achieve superior accuracy compared to previously published approaches.

Noise Reduction with Edge Preservation by Multiscale Analysis of Medical X-Ray Image Sequences

Real-time visualization of digital X-ray image sequences requires the reduction of severe noise while preserving diagnostic details. We introduce a noise reduction method for X-ray image sequences using products of Laplacian pyramid coefficients. The method features SNR improvement comparable to the Wiener filter, however, being superior in the preservation of fine structures and generating a more stable image impression in sequences.

Motion Detection for Adaptive Spatio-temporal Filtering of Medical X-Ray Image Sequences

Spatio-temporal filters are used to improve the quality of X-ray image sequences exhibiting severe noise in real-time. The spatial and temporal ratios have to be adapted locally in order to avoid artifacts. We propose a method processing the positive and negative pixel values of difference images independently in order to detect regions dominated by motion and single pixels dominated by noise. In the context of noise-adaptive binarization using Euler numbers, the influence of noise and motion on Euler curves is investigated.