Ivan Bajla - Academia.edu (original) (raw)

Papers by Ivan Bajla

Proceedings of SPIE, Apr 27, 2001

ABSTRACT

Proceedings of SPIE, Sep 10, 1999

ABSTRACT

In the paper we describe basic functions of a Hierarchical Temporal Memory (HTM) based on a novel... more In the paper we describe basic functions of a Hierarchical Temporal Memory (HTM) based on a novel biologically inspired network model of the overall large-scale structure of human neocortex. It appeared in a form of research release of the system NuPIC (Numenta Platform for Intelligent Computing) in 2007. In the design of the HTM, hierarchical structure and spatio-temporal relations serve for generation of invariant representations of the outer world (e.g. world of visual patterns), similar to those in human neocortex. There are several open issues for a research into HTM, in particular those applied to pattern recognition tasks. In the paper we report our results of the HTM architecture design and optimization of the network parameters for the task of recognition of the handwritten digits from the well benchmarked USPS database.

The research in the domain of content-based image retrieval (CBIR) is concentrated on several pro... more The research in the domain of content-based image retrieval (CBIR) is concentrated on several problems, among which two problems are essential:-development of efficient image classification algorithms, and selection of suitable image features. Recently, a biologically inspired Hierarchical Temporal Memory (HTM) network demonstrated promising results in image classification tasks. The focus of this paper is to explore possibilities of this network to be applied to CBIR. In particular, we study the performance of the HTM network when, instead of conventional grey-level images and features, color features are used. The results of our experiments show that using color texture features, defined for a reduced range of color quantization values, performs comparably well as the grey-scale image features.

The poster gives an overview of the theoretical and simulation research carried out by the Image ... more The poster gives an overview of the theoretical and simulation research carried out by the Image Processing Department at IMS during the recent 5 years. The results achieved include the following topics: image reconstruction from projections (optimal parameters of the Direct Fourier Method, noise suppression and its relation to image object detection), rank algorithms as a class of local image transforms (novel formalism, effective computational approaches, processing of CT images, especially aimed to differ between liver pathologies), synthesis of 2,5D image representation based on segmentation of the CT slices (investigation of surface tracking and volume rendering algorithms, implementation issues in medical practice). The basic description of the software system MAN designed as a working environment for PC-based research in CAR will be provided. The user interface incorporated into this system is window-oriented and utilizes a graphic layout of the videomemory content. The software system MAN being developed in C language runs under MS-DOS operating system. Due to the virtual image processor philosophy, the system can be easily adapted to various image processing hardware. The system is designed to be extended permanently by user-written programs. The experience with an introduction of the ACR-NEMA standard into our system will be discussed briefly.

2017 11th International Conference on Measurement, 2017

In the paper, the controversial claim of the authors [3] that: “Clustering of time series subsequ... more In the paper, the controversial claim of the authors [3] that: “Clustering of time series subsequences, which are generated by the sliding window principle, is meaningless” is addressed and thoroughly tested. The test results confirmed the cited claim in respect with the synthetic pattern data, although, minor deviations depending on different window length have been obtained. For the retrospective set of real Intracranial Pressure (ICP) data acquired for the patients with the severe traumatic brain injury, we proposed an alternative of the sliding window approach consisting in the definition of specific segmentation of ICP records and in introducing six quantitative features. We show that clusterization in the corresponding feature vector space does not possess the property claimed by Keogh in [3].

Advances in Biomedical Measurement, 1988

Nowadays, tomographic imaging methods based on various physical principles represent an attractiv... more Nowadays, tomographic imaging methods based on various physical principles represent an attractive tool for medical diagnosis. In contemporary X-ray tomography systems the convolution filtered backprojection method still remains the leading mathematical approach implemented. However, continually increasing demands, namely computational speed in multiplanar or 3-D spatial methods and dynamic tomographic studies, have evoked a renewed interest in earlier proposed methods. One of the most promising of these has proved to be the Direct Fourier Inversion method (the so-called DFI method). This method was proposed by Ramachandran (Ramachandran and Lakshminarayanan, 1971), but unsatisfactory reconstruction quality was attained and, as a result, the method was not pursued further. Several papers were published from the late 1970s onwards (Mersereau, 1976; Stark et al., 1981; Niki et al., 1983), which exhibited a new interest in the DFI method. Although certain possibilities of error suppression were described in these papers, some important questions still remained to be tackled. We therefore considered it appropriate to re-evaluate discrete mathematical aspects of the DFI method and to develop effective computational algorithms. The Fourier spectral character of the measured data in Magnetic Resonance Tomography (Cho et al., 1982) made such research still more interesting.

Physiological Measurement, 2017

Objective: The conventional application of intracranial pressure (ICP) monitoring of traumatic br... more Objective: The conventional application of intracranial pressure (ICP) monitoring of traumatic brain injury (TBI) patients consists merely in the acquisition of ICP values in discrete time and their comparison to the established ICP threshold. An exceeding of this threshold triggers a special emergency treatment protocol. This paper addresses the possibility of making use of the rich information latent in the ICP records of known vital and fatal outcomes gathered during real clinical practice of treating TBI patients. Our assumption was that the proposed algorithmic procedure derived from this information could, in addition to ICP monitoring itself, provide a complementary added value. This might help clinicians to make better decisions during a patient's treatment. Approach: We concentrated on studying specific clustering schemes for subsequences of ICP time series. The clusterization problem was formulated for feature vectors which are introduced to represent ICP time subsequences. The ICP transformation to a feature space uses global and local definitions of time subsequences. For clusterization itself, we adopted hierarchical Gaussian mixture models (hGMMs). By using posterior probabilities of the clusters, we introduced three novel alarm functions. We explored two alternative methods of searching for optimum alarm function thresholds (ROC analysis and a novel efficiency measure). Main results: We performed extensive cross-validation experiments on a clinical retrospective data set. The results of the optimization over several hGMMs, various feature space dimensionality and all the types of the novel alarm functions show the potential of the novel alarm functions for supplementing conventional ICP monitoring. Significance: In conclusion, the paper provides a prospective extended ICP monitoring technique for real TBI patients, based on the

2017 11th International Conference on Measurement, 2017

The topic of our research is based on the perspective of electromagnetic principles of living sys... more The topic of our research is based on the perspective of electromagnetic principles of living system operation. The overall objective of this research is to expand knowledge of the impact of weak external low-frequency electromagnetic fields (EMF) on the functioning of selected biological objects. The particular aim of the research is to design a new experimental approach enabling efficient scanning through EMF parameters while searching for a specific response of the investigated biosystem. Methods based on the impedance spectroscopy are being developed to determine the response in the growth rate of cell cultures in an aqueous medium as well as in electrical characterization of cell structures. For this purpose, a complex experimental platform for quantitative characterization of the biological response to specific parameters of weak low-frequency electromagnetic fields is under development.

Proceedings of the 2nd International Conference on Pattern Recognition Applications and Methods, 2013

The aim of this paper is to report on a pilot application of a bio-inspired intelligent network m... more The aim of this paper is to report on a pilot application of a bio-inspired intelligent network model, called Hierarchical Temporal Memory (HTM), for recognition (detection) of untrustworthy manipulation with an Automatic Teller Machine (ATM). HTM was used as a crucial part of an anomaly detection system to recognize hard-to-identifiable faces, i.e., faces with a mask, covered with a scarf, or wearing sunglasses. Those types of face occlusion can be a good indicator of potentialy malicious intentions of an ATM user. In the presented system, the Kinect camera was used for acquisition of video image sequences. The Kinect's depth output along with skeleton tracking was used as a basis of the color image segmentation. To test the proposed system, experiments have been carried out in which several participants performed normal and untrustworthy actions using an ATM simulator. The output of the face classification system can assist a security personnel in surveillance tasks.

2019 12th International Conference on Measurement, 2019

Within the pipeline of microscopic yeast cell image operations, a problem of decoloring is addres... more Within the pipeline of microscopic yeast cell image operations, a problem of decoloring is addressed. The approach proposed by Grundland et al. is analyzed and improved. A special color phantom for yeast cell images is constructed. An optimum combination of two controlling parameters of this decoloring method is found for phantom segmentation, by precision and recall characteristics in comparison to Ground truth.

Pattern Analysis and Applications, 2008

Erythropoietin (Epo) is a hormone which can be misused as a doping substance. Its detection invol... more Erythropoietin (Epo) is a hormone which can be misused as a doping substance. Its detection involves analysis of images containing specific objects (bands), whose position and intensity are critical for doping positivity. Within a research project of the World Anti-Doping Agency (WADA) we are implementing the GASepo software that serves for Epo testing in doping control laboratories worldwide. For identification of the bands we have developed a segmentation procedure based on a sequence of filters. Whereas all true bands are properly segmented, the procedure generates a number of false positives (artefacts). To separate these artefacts we suggested a post-segmentation supervised classification using real-valued geometrical measures of objects. The method is based on a fuzzy modification of Ross Quinlan's ID3 method, included in the mlf TM software (Machine Learning Framework). It provides a framework that generates fuzzy decision trees, as well as fuzzy sets for input data. Initially used training set of segmented objects has been replaced by a new one prepared by more accurate expertise using the latest release of the GASepo software. The new fuzzy decision trees (FDT) have been generated for a set of five and nine fuzzy sets. The comparison of the results on testing set of segmented objects shows that the classification based on the new FDTs outperforms other classification methods.

Computer Methods and Programs in Biomedicine, 2005

A novel methodology of electrophoretic gel image analysis has been proposed that is based on two-... more A novel methodology of electrophoretic gel image analysis has been proposed that is based on two-dimensional image processing methods instead of previously used one-dimensional Gaussian deconvolution. The crucial problem of the analysis of imperfect gels, that consists in band detection, is solved using the algorithms of band boundary detection and intensity homogeneity indication. The template approach represents the core element of the developed algorithms. The GelMaster software system has been developed in which the novel algorithms are implemented. It involves two-stage interaction with the user: detection of the true bands and deleting the false band detections. The main features of the GelMaster system and the most important algorithms are described.

IEEE Transactions on Medical Imaging, 1993

Absfrucf-A great number of tomographic systems, especially those equipped with fast data acquisit... more Absfrucf-A great number of tomographic systems, especially those equipped with fast data acquisition techniques, scan the objects investigated by divergent (fan) X-ray beams. Fan-beam projection data require special reconstruction techniques to be implemented. Among reconstruction techniques from parallel projection data, the Direct Fourier Method (DFM) proved to be one of the most promising ones, especially for high-speed image reconstruction in the high-end 3-D and dynamic tomographic systems. The goal of our paper is to answer the topical question: how would direct use of the DFM influence the quality of image reconstruction from the fan-beam data? The formula describing the error caused by such an approximation is derived in the paper. The conclusions deduced from the formula are confirmed by computer simulations. Furthermore, the boundary values of data acquisition geometry parameters have been estimated for the case of using the DFM without recalculating the fan-beam data.

Eighth International Conference on Quality Control by Artificial Vision, 2007

In recent years non-negative factorization (NMF) methods of a reduced image data representation a... more In recent years non-negative factorization (NMF) methods of a reduced image data representation attracted the attention of computer vision community. These methods are considered as a convenient part-based representation of image data for recognition tasks with occluded objects. In ...

Image Processing: Machine Vision Applications, 2008

Non-negative matrix factorization of an input data matrix into a matrix of basis vectors and a ma... more Non-negative matrix factorization of an input data matrix into a matrix of basis vectors and a matrix of encoding coefficients is a subspace representation method that has attracted attention of researches in pattern recognition in the recent period. We have explored crucial aspects of NMF on ...

Proceedings of SPIE, Apr 27, 2001

ABSTRACT

Proceedings of SPIE, Sep 10, 1999

ABSTRACT

In the paper we describe basic functions of a Hierarchical Temporal Memory (HTM) based on a novel... more In the paper we describe basic functions of a Hierarchical Temporal Memory (HTM) based on a novel biologically inspired network model of the overall large-scale structure of human neocortex. It appeared in a form of research release of the system NuPIC (Numenta Platform for Intelligent Computing) in 2007. In the design of the HTM, hierarchical structure and spatio-temporal relations serve for generation of invariant representations of the outer world (e.g. world of visual patterns), similar to those in human neocortex. There are several open issues for a research into HTM, in particular those applied to pattern recognition tasks. In the paper we report our results of the HTM architecture design and optimization of the network parameters for the task of recognition of the handwritten digits from the well benchmarked USPS database.

The research in the domain of content-based image retrieval (CBIR) is concentrated on several pro... more The research in the domain of content-based image retrieval (CBIR) is concentrated on several problems, among which two problems are essential:-development of efficient image classification algorithms, and selection of suitable image features. Recently, a biologically inspired Hierarchical Temporal Memory (HTM) network demonstrated promising results in image classification tasks. The focus of this paper is to explore possibilities of this network to be applied to CBIR. In particular, we study the performance of the HTM network when, instead of conventional grey-level images and features, color features are used. The results of our experiments show that using color texture features, defined for a reduced range of color quantization values, performs comparably well as the grey-scale image features.

The poster gives an overview of the theoretical and simulation research carried out by the Image ... more The poster gives an overview of the theoretical and simulation research carried out by the Image Processing Department at IMS during the recent 5 years. The results achieved include the following topics: image reconstruction from projections (optimal parameters of the Direct Fourier Method, noise suppression and its relation to image object detection), rank algorithms as a class of local image transforms (novel formalism, effective computational approaches, processing of CT images, especially aimed to differ between liver pathologies), synthesis of 2,5D image representation based on segmentation of the CT slices (investigation of surface tracking and volume rendering algorithms, implementation issues in medical practice). The basic description of the software system MAN designed as a working environment for PC-based research in CAR will be provided. The user interface incorporated into this system is window-oriented and utilizes a graphic layout of the videomemory content. The software system MAN being developed in C language runs under MS-DOS operating system. Due to the virtual image processor philosophy, the system can be easily adapted to various image processing hardware. The system is designed to be extended permanently by user-written programs. The experience with an introduction of the ACR-NEMA standard into our system will be discussed briefly.

2017 11th International Conference on Measurement, 2017

In the paper, the controversial claim of the authors [3] that: “Clustering of time series subsequ... more In the paper, the controversial claim of the authors [3] that: “Clustering of time series subsequences, which are generated by the sliding window principle, is meaningless” is addressed and thoroughly tested. The test results confirmed the cited claim in respect with the synthetic pattern data, although, minor deviations depending on different window length have been obtained. For the retrospective set of real Intracranial Pressure (ICP) data acquired for the patients with the severe traumatic brain injury, we proposed an alternative of the sliding window approach consisting in the definition of specific segmentation of ICP records and in introducing six quantitative features. We show that clusterization in the corresponding feature vector space does not possess the property claimed by Keogh in [3].

Advances in Biomedical Measurement, 1988

Nowadays, tomographic imaging methods based on various physical principles represent an attractiv... more Nowadays, tomographic imaging methods based on various physical principles represent an attractive tool for medical diagnosis. In contemporary X-ray tomography systems the convolution filtered backprojection method still remains the leading mathematical approach implemented. However, continually increasing demands, namely computational speed in multiplanar or 3-D spatial methods and dynamic tomographic studies, have evoked a renewed interest in earlier proposed methods. One of the most promising of these has proved to be the Direct Fourier Inversion method (the so-called DFI method). This method was proposed by Ramachandran (Ramachandran and Lakshminarayanan, 1971), but unsatisfactory reconstruction quality was attained and, as a result, the method was not pursued further. Several papers were published from the late 1970s onwards (Mersereau, 1976; Stark et al., 1981; Niki et al., 1983), which exhibited a new interest in the DFI method. Although certain possibilities of error suppression were described in these papers, some important questions still remained to be tackled. We therefore considered it appropriate to re-evaluate discrete mathematical aspects of the DFI method and to develop effective computational algorithms. The Fourier spectral character of the measured data in Magnetic Resonance Tomography (Cho et al., 1982) made such research still more interesting.

Physiological Measurement, 2017

Objective: The conventional application of intracranial pressure (ICP) monitoring of traumatic br... more Objective: The conventional application of intracranial pressure (ICP) monitoring of traumatic brain injury (TBI) patients consists merely in the acquisition of ICP values in discrete time and their comparison to the established ICP threshold. An exceeding of this threshold triggers a special emergency treatment protocol. This paper addresses the possibility of making use of the rich information latent in the ICP records of known vital and fatal outcomes gathered during real clinical practice of treating TBI patients. Our assumption was that the proposed algorithmic procedure derived from this information could, in addition to ICP monitoring itself, provide a complementary added value. This might help clinicians to make better decisions during a patient's treatment. Approach: We concentrated on studying specific clustering schemes for subsequences of ICP time series. The clusterization problem was formulated for feature vectors which are introduced to represent ICP time subsequences. The ICP transformation to a feature space uses global and local definitions of time subsequences. For clusterization itself, we adopted hierarchical Gaussian mixture models (hGMMs). By using posterior probabilities of the clusters, we introduced three novel alarm functions. We explored two alternative methods of searching for optimum alarm function thresholds (ROC analysis and a novel efficiency measure). Main results: We performed extensive cross-validation experiments on a clinical retrospective data set. The results of the optimization over several hGMMs, various feature space dimensionality and all the types of the novel alarm functions show the potential of the novel alarm functions for supplementing conventional ICP monitoring. Significance: In conclusion, the paper provides a prospective extended ICP monitoring technique for real TBI patients, based on the

2017 11th International Conference on Measurement, 2017

The topic of our research is based on the perspective of electromagnetic principles of living sys... more The topic of our research is based on the perspective of electromagnetic principles of living system operation. The overall objective of this research is to expand knowledge of the impact of weak external low-frequency electromagnetic fields (EMF) on the functioning of selected biological objects. The particular aim of the research is to design a new experimental approach enabling efficient scanning through EMF parameters while searching for a specific response of the investigated biosystem. Methods based on the impedance spectroscopy are being developed to determine the response in the growth rate of cell cultures in an aqueous medium as well as in electrical characterization of cell structures. For this purpose, a complex experimental platform for quantitative characterization of the biological response to specific parameters of weak low-frequency electromagnetic fields is under development.

Proceedings of the 2nd International Conference on Pattern Recognition Applications and Methods, 2013

The aim of this paper is to report on a pilot application of a bio-inspired intelligent network m... more The aim of this paper is to report on a pilot application of a bio-inspired intelligent network model, called Hierarchical Temporal Memory (HTM), for recognition (detection) of untrustworthy manipulation with an Automatic Teller Machine (ATM). HTM was used as a crucial part of an anomaly detection system to recognize hard-to-identifiable faces, i.e., faces with a mask, covered with a scarf, or wearing sunglasses. Those types of face occlusion can be a good indicator of potentialy malicious intentions of an ATM user. In the presented system, the Kinect camera was used for acquisition of video image sequences. The Kinect's depth output along with skeleton tracking was used as a basis of the color image segmentation. To test the proposed system, experiments have been carried out in which several participants performed normal and untrustworthy actions using an ATM simulator. The output of the face classification system can assist a security personnel in surveillance tasks.

2019 12th International Conference on Measurement, 2019

Within the pipeline of microscopic yeast cell image operations, a problem of decoloring is addres... more Within the pipeline of microscopic yeast cell image operations, a problem of decoloring is addressed. The approach proposed by Grundland et al. is analyzed and improved. A special color phantom for yeast cell images is constructed. An optimum combination of two controlling parameters of this decoloring method is found for phantom segmentation, by precision and recall characteristics in comparison to Ground truth.

Pattern Analysis and Applications, 2008

Erythropoietin (Epo) is a hormone which can be misused as a doping substance. Its detection invol... more Erythropoietin (Epo) is a hormone which can be misused as a doping substance. Its detection involves analysis of images containing specific objects (bands), whose position and intensity are critical for doping positivity. Within a research project of the World Anti-Doping Agency (WADA) we are implementing the GASepo software that serves for Epo testing in doping control laboratories worldwide. For identification of the bands we have developed a segmentation procedure based on a sequence of filters. Whereas all true bands are properly segmented, the procedure generates a number of false positives (artefacts). To separate these artefacts we suggested a post-segmentation supervised classification using real-valued geometrical measures of objects. The method is based on a fuzzy modification of Ross Quinlan's ID3 method, included in the mlf TM software (Machine Learning Framework). It provides a framework that generates fuzzy decision trees, as well as fuzzy sets for input data. Initially used training set of segmented objects has been replaced by a new one prepared by more accurate expertise using the latest release of the GASepo software. The new fuzzy decision trees (FDT) have been generated for a set of five and nine fuzzy sets. The comparison of the results on testing set of segmented objects shows that the classification based on the new FDTs outperforms other classification methods.

Computer Methods and Programs in Biomedicine, 2005

A novel methodology of electrophoretic gel image analysis has been proposed that is based on two-... more A novel methodology of electrophoretic gel image analysis has been proposed that is based on two-dimensional image processing methods instead of previously used one-dimensional Gaussian deconvolution. The crucial problem of the analysis of imperfect gels, that consists in band detection, is solved using the algorithms of band boundary detection and intensity homogeneity indication. The template approach represents the core element of the developed algorithms. The GelMaster software system has been developed in which the novel algorithms are implemented. It involves two-stage interaction with the user: detection of the true bands and deleting the false band detections. The main features of the GelMaster system and the most important algorithms are described.

IEEE Transactions on Medical Imaging, 1993

Absfrucf-A great number of tomographic systems, especially those equipped with fast data acquisit... more Absfrucf-A great number of tomographic systems, especially those equipped with fast data acquisition techniques, scan the objects investigated by divergent (fan) X-ray beams. Fan-beam projection data require special reconstruction techniques to be implemented. Among reconstruction techniques from parallel projection data, the Direct Fourier Method (DFM) proved to be one of the most promising ones, especially for high-speed image reconstruction in the high-end 3-D and dynamic tomographic systems. The goal of our paper is to answer the topical question: how would direct use of the DFM influence the quality of image reconstruction from the fan-beam data? The formula describing the error caused by such an approximation is derived in the paper. The conclusions deduced from the formula are confirmed by computer simulations. Furthermore, the boundary values of data acquisition geometry parameters have been estimated for the case of using the DFM without recalculating the fan-beam data.

Eighth International Conference on Quality Control by Artificial Vision, 2007

In recent years non-negative factorization (NMF) methods of a reduced image data representation a... more In recent years non-negative factorization (NMF) methods of a reduced image data representation attracted the attention of computer vision community. These methods are considered as a convenient part-based representation of image data for recognition tasks with occluded objects. In ...

Image Processing: Machine Vision Applications, 2008

Non-negative matrix factorization of an input data matrix into a matrix of basis vectors and a ma... more Non-negative matrix factorization of an input data matrix into a matrix of basis vectors and a matrix of encoding coefficients is a subspace representation method that has attracted attention of researches in pattern recognition in the recent period. We have explored crucial aspects of NMF on ...