Grzegorz Ostrek

Stroke Slicer for CT-Based Automatic Detection of Acute Ischemia

Computed understanding of CT brain images used for assisted diagnosis of acute ischemic stroke disease was the subject of reported study. Stroke slicer was proposed as computer aided diagnosis (CAD) tool that allows extraction and enhancement of direct early ischemia sign - subtle hypodense of local tissue damage. Hypoattenuation of selected CT scan areas was visualized distinctly in a form of semantic maps. Moreover, brain tissue texture was characterized, analyzed and classified in multiscale domain to detect the areas of ischemic events. As the results of slice-oriented processing, the automatically indicated regions of ischemia and enhanced hypodensity maps were proposed as additional view for computerized assisted diagnosis. The experimental verification of stroke slicer was concentrated on diagnostic improvement in clinical practice by using semantic maps as additional information for interpretation procedure. Reported results indicate possible improvement of diagnostic output for really challenging problem of as early as possible CT-based ischemic stroke detection.

Method of Brain Structure Extraction for CT-Based Stroke Detection

A fully automated method for extracting brain structures from computed tomography images by employing adaptive filtering and finite Gaussian Mixture Modeling (GMM) with context-based enhancement is proposed. Generally, the method is composed of two phases. First, adaptive partial mean filter for noise removal and edge sharpening is used. The second phase is the multistage segmentation. Initial segmentation step concerning brain extraction from skull and non-brain tissue defines a region of interest (ROI) for further processing. Each pixel in ROI is assigned to one of three semantically fundamental classes - white matter (WM), gray matter (GM) and cerebrospinal fluid (CBF) and two extended classes of specific tissue. GMM with expectation-maximization algorithm (EM) is employed to assign initial class labels to image pixels and followed by context information modeling through Contextual Bayesian Relaxation Labeling (CBRL). The CBRL algorithm incorporates local neighborhood information and iteratively refines the outcome of GMM classification. The results of proposed approach have been verified by extracting susceptible-to-stroke regions (SSR) processed for hypodensity distribution estimation. The extracted structures are more smooth and reliable in comparison to region growing segmentation results.

Improving Diagnostic Value of CT Examinations in Hyperacute Ischemic Stroke

This paper presents a computer assisted support of ischemic stroke diagnosis based on nonenhanced CT examinations acquired in the hyperacute phase of stroke. Computational analysis, recognition, and image understanding methods were used for extraction of the subtlest signs of hypodensity in diagnostically important areas. Starting from perception improvement, suggestive and coarse image data visualization was designed as a complement of the standard diagnosis procedure based on CT scan soft-copy review. The proposed method includes an evidence-based description of ischemic conditions and changes, de-skulling and segmenting of unusual areas, the analysis of hypodensity signs across scales and subbands with noise reduction, and hypodensity extraction. Following visualization, forms of empowered hypodensity symptoms localize suggested ischemic areas in source brain image space. Increased visibility of cerebral ischemia for difficult-to-diagnose cases was experimentally noticed and improved diagnostic value of CT was concluded.

Automatic early stroke recognition algorithm in CT images

The subject of the reported study was automatic recognition of early ischemic stroke lesions in CT scans. Proposed extraction method was based on the investigated specificity of tissue texture features in hypothetical penumbra regions. Prediction of such regions was estimated by initial hypodensity enhancement procedure. Block-oriented areas of selected brain tissue were analyzed in both source and multiscale-processed data domains. The extraction and selection of well differentiating features were fundamental effort to verify research hypothesis that acute ischemic tissue is noticeably altered in CT imaging. Moreover, various classifiers were examined on large feature data sets. Limitations and shortcomings caused by a class imbalance problem were considered. Experimental verification of designed and implemented recognition procedures is the main input of this paper.

Stroke Monitor as a Device Improving Diagnostic Value of Computed Tomography in Hyperacute Stroke

Stroke Monitor as a computer-assisted support of ischemic stroke diagnosis was presented in this paper. Computer analysis, processing and image understanding was based on nonenhanced CT examinations acquired in hyperacute phase of stroke. The multiscale extraction of the subtlest signs of hypodensity in diagnostically important areas was designed to improve standard diagnosis procedure based on CT scan soft-copy review. Proposed method includes evidence-based description of ischemic conditions and changes, deskulling and segmenting of unusual areas, the analysis of hypodensity signs across scales and subbands with noise reduction and hypodensity extraction. Following visualization forms of empowered hypodensity symptoms localizes suggested ischemic areas in source brain image space. Increased visibility of cerebral ischemia for difficult-in-diagnosis cases was experimentally noticed and improved diagnostic value of CT was concluded.

Stroke Display Extensions: Three Forms of Visualization

The computer-assisted support of acute ischemic stoke detection was the subject of our research reported in this paper. The conditioning of early stroke diagnosis based on CT examinations was analyzed. The multiscale extraction of the subtlest signs of hypodensity which were often undetected in standard CT scan review was presented. Proposed method was as follows: evidence-based description of ischemic changes, the analysis of hypodensity signs across scales, noise reduction and hypodensity extraction, and following display of ischemic changes localized in source brain image space. Important issues were: –extension of the brain tissues for marginal and missing space after deskulling and segmenting of unusual areas; –multiscale transform selection; –denoising in scale-space domain; –visualization conditions fixing. Three forms of extracted stroke sings visualization were proposed. Increased visibility of cerebral ischemia for difficult-in-diagnosis cases was experimentally noticed.

Hypodensity extractor

We report on the extraction procedures of low-contrast symptomatic hypodensity optimized for a computed tomography-based diagnosis. The specific application is brain imaging with enhanced perception of hypodense areas which are direct symptoms of acute ischemia. A standard low-contrast phantom, as commonly employed in dosimetry and imaging quality evaluation, was used to derive numeric criteria for assessing the extraction effectiveness. Our proposed procedure is based on multiscale analysis of the image data expanded over the frames of wavelets, curvelets or complex wavelets, followed by nonlinear approximation of the symptom signatures. Apparent subtle density changes in the phantom were evaluated using computational metrics and subjective ratings. We discuss the advantages and disadvantages of our proposed optimized hypodensity extraction procedures.

Approximation of Subtle Pathology Signs in Multiscale Domain for Computer-Aided Ischemic Stroke Diagnosis

Computed understanding of CT images used for aided stroke diagnosis was the subject of reported research. Subtle hypodense changes of brain tissue as direct ischemia signs was estimated and extracted to improve diagnosis. Fundamental value of semantic content representation approximated from source images was studied. Nonlinear approximation of subtle pathology signatures in multiscale domain was verified for several local bases including wavelets, curvelets, contourlets and wedgelets. Different rationales for best bases selection were considered. Target pathology estimation procedures were optimized with a criterion of maximally clear extraction of diagnostic information. Visual expression of emphasized hypodenstity was verified for a test set of 25 acute stroke examinations. Suggested methods of stroke nonlinear approximation in many scales may facilitate the early CT-based diagnosis.

Stroke Tissue Pattern Recognition Based on CT Texture Analysis

The main objective of this paper is a texture-based solution to the problem of acute stroke tissue recognition on computed tomography images. Our proposed method of early stroke indication was based on two fundamental steps: (i) segmentation of potential areas with distorted brain tissue (selection of regions of interest), and (ii) acute stroke tissue recognition by extracting and then classifying a set of well-differentiating features. The proposed solution used various numerical image descriptors determined in several image transformation domains: 2D Fourier domain, polar 2D Fourier domain, and multiscale domains (i.e., wavelet, complex wavelet, and contourlet domain). The obtained results indicate the possibility of relatively effective detection of early stroke symptoms in CT images. Selected normal or pathological blocks were classified by LogitBoost with the accuracy close to 75 % with the use of our adjusted cross-validation procedure.

Perception Enhancement of Bronchoscopic Video

The perception enhancement of bronchovideoscopy examination was the subject of reported study. Proposed methods enrich diagnosis and treatment giving real-time additional view based on following adjusted concepts: contrast enhancement, histogram equalization, sharpening, color correction. Additionally wavelets, edge preserving and morphological filters were involved in the image processing. Wavelet texture extraction and sharpening were considered as the most useful and included into complex system of bronchoscopy storage, review and analysis called “Bronchovid”. Extended capabilities of classical white light bronchoscopy which has higher resolution than other autofluorescence and spectroscopy based diagnostic methods can be complementary for them in case of any endobronchial biopsy suspicious findings.