Andrzej Skalski

Data Processing Tasks in Wireless GI Endoscopy: Image-Based Capsule Localization & Navigation and Video Compression

The paper addresses data processing support that is required in capsule gastrointestinal endoscopy. First, capsule position estimation method using standard MPEG-7 image features (descriptors) is discussed. The proposed approach makes use of vector quantization, principal component analysis and neural networks. Next, new algorithms dedicated for virtual colonoscopy (VC) human body inspection are described. The VC images can be registered with endoscopic ones and help in capsule localization and navigation. Finally, an original, low- complexity, efficient image compression method, based on integer-to-integer 4x4 DCT transform, is presented and experimentally verified.

Acoustic Analysis Assessment in Speech Pathology Detection

Abstract Automatic detection of voice pathologies enables non-invasive, low cost and objective assessments of the presence of disorders, as well as accelerating and improving the process of diagnosis and clinical treatment given to patients. In this work, a vector made up of 28 acoustic parameters is evaluated using principal component analysis (PCA), kernel principal component analysis (kPCA) and an auto-associative neural network (NLPCA) in four kinds of pathology detection (hyperfunctional dysphonia, functional dysphonia, laryngitis, vocal cord paralysis) using the a, i and u vowels, spoken at a high, low and normal pitch. The results indicate that the kPCA and NLPCA methods can be considered a step towards pathology detection of the vocal folds. The results show that such an approach provides acceptable results for this purpose, with the best efficiency levels of around 100%. The study brings the most commonly used approaches to speech signal processing together and leads to a comparison of the machine learning methods determining the health status of the patient

Automatic tracking of implanted fiducial markers in cone beam CT projection images

PURPOSE This paper describes a novel method for simultaneous intrafraction tracking of multiple fiducial markers. Although the proposed method is generic and can be adopted for a number of applications including fluoroscopy based patient position monitoring and gated radiotherapy, the tracking results presented in this paper are specific to tracking fiducial markers in a sequence of cone beam CT projection images. METHODS The proposed method is accurate and robust thanks to utilizing the mean shift and random sampling principles, respectively. The performance of the proposed method was evaluated with qualitative and quantitative methods, using data from two pancreatic and one prostate cancer patients and a moving phantom. The ground truth, for quantitative evaluation, was calculated based on manual tracking preformed by three observers. RESULTS The average dispersion of marker position error calculated from the tracking results for pancreas data (six markers tracked over 640 frames, 3840 marker identifications) was 0.25 mm (at iscoenter), compared with an average dispersion for the manual ground truth estimated at 0.22 mm. For prostate data (three markers tracked over 366 frames, 1098 marker identifications), the average error was 0.34 mm. The estimated tracking error in the pancreas data was < 1 mm (2 pixels) in 97.6% of cases where nearby image clutter was detected and in 100.0% of cases with no nearby image clutter. CONCLUSIONS The proposed method has accuracy comparable to that of manual tracking and, in combination with the proposed batch postprocessing, superior robustness. Marker tracking in cone beam CT (CBCT) projections is useful for a variety of purposes, such as providing data for assessment of intrafraction motion, target tracking during rotational treatment delivery, motion correction of CBCT, and phase sorting for 4D CBCT.

Analysis of vocal folds movement in high speed videoendoscopy based on level set segmentation and image registration

Computer analysis of high-speed videoendoscopy (HSV) recordings is becoming increasingly useful in functional evaluation of vocal fold pathology. In this work we present two new approaches for the vocal fold HSV-based analysis that have not been previously reported. First, segmentation of vocal fold edges is performed in our method using the level set algorithm. Second, having two vocal fold contours (for two images) we find point-to-point matching of them using the image registration method based on B-spline free form deformation. Presented results from analysis of normal vocal folds and folds after laser treatment confirm usefulness of the method.

Computed tomography-based radiotherapy planning on the example of prostate cancer: application of level-set segmentation method guided by atlas-type knowledge

In this invited paper an overview of the Computed Tomography-based cancer radiotherapy planning is given. All planning steps are described with details, i.e. its goals, existing solutions and typical realizations. On this background, as an example, a complete procedure of prostate cancer radiotherapy planning is presented in which application of Level-Set segmentation method guided by a priori atlas-type knowledge is proposed. Developed procedure was verified on data base consisting of 266 CT slices of 4 patients.

Quantification of Linear and Non-linear Acoustic Analysis Applied to Voice Pathology Detection

Present development of digital registration and methods of recorded voice processing are useful in detection of most pathologies and diseases of a human vocal tract. The recognition of the voice condition requires the creation of a model which is comprised of different acoustic parameters of speech signal. In this study a vector consisting of 31 parameters for analysing the speech signal was created. The speech parameters were extracted from time, frequency and cepstral domains. Using Principal Components Analysis the number of the parameters was reduced to 17. In order to validate the detection of the pathological voice signal, a tenfold cross-validation and confusion matrix were used. The goal and novelty of this work was the analysis of applicability of the parameters selectively used to assess the pathology.

Left Ventricle USG Image Segmentation Using Active Contour Model

In the paper novel application of active contour without edges method to left ventricle segmentation in ultrasound echocardiographic images is presented. In the proposed approach detection of informative image part and limitation of segmentation area is done by means of Hough transform what guarantees stability and correctness of the segmentation algorithm working on border between object and background with small absolute value of image gradient. Additionally, in order to increase image quality, speckle noise anisotropic diffusion filtering was applied to input noisy data. Preliminary quantitative results for artificial USG-like images and visual ones for real echocardiographic data are presented in the paper.

Voice data mining for laryngeal pathology assessment

The aim of this study was to evaluate the usefulness of different methods of speech signal analysis in the detection of voice pathologies. Firstly, an initial vector was created consisting of 28 parameters extracted from time, frequency and cepstral domain describing the human voice signal based on the analysis of sustained vowels /a/, /i/ and /u/ all at high, low and normal pitch. Afterwards we used a linear feature extraction technique (principal component analysis), which enabled a reduction in the number of parameters and choose the most effective acoustic features describing the speech signal. We have also performed non-linear data transformation which was calculated using kernel principal components. The results of the presented methods for normal and pathological cases will be revealed and discussed in this paper. The initial and extracted feature vectors were classified using the k-means clustering and the random forest classifier. We found that reasonably good classification accuracies could be achieved by selecting appropriate features. We obtained accuracies of up to 100% for classification of healthy versus pathology voice using random forest classification for female and male recordings. These results may assist in the feature development of automated detection systems for diagnosis of patients with symptoms of pathological voice.

Colon Cleansing for Virtual Colonoscopy Using Non-linear Transfer Function and Morphological Operations

The virtual colonoscopy (VC) techniques try to simulate a real colonoscopy. A doctor who makes real colonoscopy examination does not have optimal information about anatomical structures which he looks at. He sees the inner colon structure only. 3D visualization of the colon segmented from computed tomography (CT) data allows him to see the whole organ, its inner and outer part. The VC helps doctors during diagnostic processes in identification and localization of pathological changes and offers computer support for endoscopic procedures. In this paper we present new colon cleansing method based on non-linear transfer function and morphological operations. Colon cleansing is required when we receive non-clean CT data or when a patient had administered contrast before the CT scan. It allows to see the whole colon even this one lying under fluid and to compute colon centerline correctly.

Usage of ICP Algorithm for Initial Alignment in B-Splines FFD Image Registration in Breast Cancer Radiotherapy Planning

Estimation of a resected tumor lodge localization after a breast cancer surgery is a demanding task for the radiotherapy planning. The image registration techniques can be used to improve the radiotherapy. The initial alignment of two volumes is an important aspect of medical image registration procedure. We propose usage of the iterative closest point in two different scenarios: as a initial alignment, replacing intensity based rigid registration and as a initial transform to speed-up traditional rigid registration process. Two versions of the algorithm are presented: a point matching between bone structures and a line matching between volume edges. The correctness and usefulness are evaluated using: a target registration error, comparison of the computation time and convergence ratios, and visual inspection. The results demonstrate that the usage of iterative closest point algorithm significantly improve the initial alignment process in terms of the computation time.