Tomasz Hachaj

Rule-based approach to recognizing human body poses and gestures in real time

In this paper we propose a classifier capable of recognizing human body static poses and body gestures in real time. The method is called the gesture description language (GDL). The proposed methodology is intuitive, easily thought and reusable for any kind of body gestures. The very heart of our approach is an automated reasoning module. It performs forward chaining reasoning (like a classic expert system) with its inference engine every time new portion of data arrives from the feature extraction library. All rules of the knowledge base are organized in GDL scripts having the form of text files that are parsed with a LALR-1 grammar. The main novelty of this paper is a complete description of our GDL script language, its validation on a large dataset (1,600 recorded movement sequences) and the presentation of its possible application. The recognition rate for examined gestures is within the range of 80.5–98.5xa0%. We have also implemented an application that uses our method: it is a three-dimensional desktop for visualizing 3D medical datasets that is controlled by gestures recognized by the GDL module.

A system for detecting and describing pathological changes using dynamic perfusion computer tomography brain maps

This paper presents a novel method of detecting and describing pathological changes that can be visualized on dynamic computer tomography brain maps (perfusion CT). The system was tested on a set of dynamic perfusion computer tomography maps. Each set consisted of two perfusion maps (CBF, CBV and TTP for testing the irregularity detection algorithm) and one CT brain scan (for the registration algorithm) from 8 different patients with suspected strokes. In 36 of the 84 brain maps, abnormal perfusion was diagnosed. The results of the algorithm were compared with the findings of a team of two radiologists. All of the CBF and CBV maps that did not show a diagnosed asymmetry were classified correctly (i.e. no asymmetry was detected). In four of the TTP maps the algorithm found asymmetries, which were not classified as irregularities in the medical diagnosis; 84.5% of the maps were diagnosed correctly (85.7% for the CBF, 85.7% for the CBV and 82.1% for the TTP); 75% of the errors in the CBF maps and 100% of the errors in the CBV and the TTP maps were caused by the excessive detection of asymmetry regions. Errors in the CBFs and the CBVs were eliminated in cases in which the symmetry axis was selected manually. Subsequently, 96.4% of the CBF maps and 100% of the CBV maps were diagnosed correctly.

Application of neural networks in detection of abnormal brain perfusion regions

In this paper we modify the existing image processing schema for computed tomography perfusion (CTP) maps analysis in order to increase its efficiency in detection and classification of perfusion abnormalities by adding multilayer perceptron neural network diagnostic module and new neural network based image processing procedure. The main new contribution of this paper is description of self-organizing Kohonens map (SOM) architecture that is used in the image processing step. CTP is based on generation of time series images that shows the flow of contrast material in vascular system of the brain. Despite the fact that SOM was previously reported as a reliable tool for time series classification we decided to utilize it as the extension of our existing methodology. We also present methodology of CTP - based diagnostic system for brain stroke diagnosis updated with our latest results and never before presented validation results. Our proposition is consisted of two steps: initial image processing after which potential asymmetry regions between hemispheres are detected and the second step during which position, type and prognostic map for potentially infarcted tissues are generated. We also made comparison of efficiency of our new approach with existing one. The test set for our algorithm validation was consisted of 75 CTP images triplets (one cerebral blood flow, cerebral blood volume and CT in each) from 30 patients (both man and woman) with suspicion of ischemia/stroke. The maps were previously diagnosed by an expert. Forty two CTP maps were described as normal (no abnormalities), 33 maps showed syndromes of ischemic strokes with various severities. Finally over 82% of CTP maps from our test set were rightly segmented and classified.

CAD system for automatic analysis of CT perfusion maps

In this article, authors present novel algorithms developed for the computer-assisted diagnosis (CAD) system for analysis of dynamic brain perfusion, computer tomography (CT) maps, cerebral blood flow (CBF), and cerebral blood volume (CBV). Those methods perform both quantitative analysis [detection and measurement and description with brain anatomy atlas (AA) of potential asymmetries/lesions] and qualitative analysis (semantic interpretation of visualized symptoms). The semantic interpretation (decision about type of lesion: ischemic/hemorrhagic, is the brain tissue at risk of infraction or not) of visualized symptoms is done by, so-called, cognitive inference processes allowing for reasoning on character of pathological regions based on specialist image knowledge. The whole system is implemented in.NET platform (C# programming language) and can be used on any standard PC computer with.NET framework installed.

Novel Applications of VR: Visualization of perfusion abnormalities with GPU-based volume rendering

This article presents an innovative GPU-based solution for visualization of perfusion abnormalities detected in dynamic brain perfusion computer tomography (dpCT) maps in an augmented-reality environment. This new graphic algorithm is a vital part of a complex system called DMD (detection measure and description), which was recently proposed by the authors. The benefit of this algorithm over previous versions is its ability to operate in real time to satisfy the needs of augmented reality simulation. The performance speed (in frames per second) of six volume-rendering algorithms was determined for models with and without semi-transparent pixels.

Real time exploration and management of large medical volumetric datasets on small mobile devices—Evaluation of remote volume rendering approach

Abstract In this paper I present the architecture of system that can be used for real time exploration and management of large medical volumetric datasets. The new state of the art solution presented in this paper is an example of visual data management system. System prototype evaluation proved that it is possible to use low-powered (and cheap) up-to-date mobile devices with programmable GPUs as the remote interfaces for exploration of large volumetric medical data. The implementation was done with high-level programming language that enables portability between different hardware models. The lack of lossy compression enables to display high quality medical images visualizations without any simplifications and noises in frequency domain. The prototype of system is capable to remotely render and send to a client (for example cell phone or tablet) rendered data with frequency 30xa0fps with limited resolution during interaction. One second after the interaction is finished client machine receives full resolution image. The evaluation of the system was performed on volumetric computed tomography angiography image with approximate size 512 3 xa0voxels.

Automatic Detection and Lesion Description in Cerebral Blood Flow and Cerebral Blood Volume Perfusion Maps

The paper presents a new approach of description and analysis of potential lesions in cerebral blood flow and cerebral blood volume perfusion maps. To perform such a computer analysis at first the axial position in patient’s brain must be chosen. In next step the generation of brain perfusion maps connected with detection of asymmetries indicating selected pathological states, and allowing supporting diagnosis of the visible lesions are all done automatically. The constructed system uses the unified algorithm for detection of asymmetry in cerebral blood flow and cerebral blood volume perfusion maps, as well as a registration algorithm created by the authors and based on free form deformation. The tests were performed on set of dynamic perfusion computer tomography maps. Algorithms presented in this paper enable detection of pathological states like head injuries, epilepsy, brain vascular disease, ischemic and hemorrhagic stroke.

Application of Assistive Computer Vision Methods to Oyama Karate Techniques Recognition

In this paper we propose a novel algorithm that enables online actions segmentation and classification. The algorithm enables segmentation from an incoming motion capture (MoCap) data stream, sport (or karate) movement sequences that are later processed by classification algorithm. The segmentation is based on Gesture Description Language classifier that is trained with an unsupervised learning algorithm. The classification is performed by continuous density forward-only hidden Markov models (HMM) classifier. Our methodology was evaluated on a unique dataset consisting of MoCap recordings of six Oyama karate martial artists including multiple champion of Kumite Knockdown Oyama karate. The dataset consists of 10 classes of actions and included dynamic actions of stands, kicks and blocking techniques. Total number of samples was 1236. We have examined several HMM classifiers with various number of hidden states and also Gaussian mixture model (GMM) classifier to empirically find the best setup of the proposed method in our dataset. We have used leave-one-out cross validation. The recognition rate of our methodology differs between karate techniques and is in the range of 81% ± 15% even to 100%. Our method is not limited for this class of actions but can be easily adapted to any other MoCap-based actions. The description of our approach and its evaluation are the main contributions of this paper. The results presented in this paper are effects of pioneering research on online karate action classification.

Semantic Description and Recognition of Human Body Poses and Movement Sequences with Gesture Description Language

In this article we introduce new approach for human body poses and movement sequences recognition. Our concept is based on syntactic description with so called Gesture Description Language (GDL). The implementation of GDL requires special semantic reasoning module with additional heap-like memory. In the following paragraphs we shortly describes our initial concept. We also present software and hardware architecture that we created to test our solution and very promising early experiments results.

Automatic segmentation of the carotid artery bifurcation region with a region-growing approach

Abstract. The original contribution is to propose an intensity-based segmentation algorithm for extracting the carotid artery bifurcation region and validate the proposed solution on real patients’ CTA data. The proposed homogeneity criteria allow the production of locally smooth segmentations and prevent excessive growth into neighboring tissues of similar densities. The obtained segmentation results are compared to manual findings of a radiologist and measured with the Dice similarity coefficient (Dsi). This technique has been shown to be a reliable tool as effective as top state-of-the-art methods (Dsi=93.6%±3.5%).