Context Search of Biomedical Images
Development of Methods for Analysis, Indexing and Search of Biomedical Images in Cardiology
Tech Area / Field
- INF-SIG/Sensors and Signal Processing/Information and Communications
- INF-SOF/Software/Information and Communications
- MED-DID/Diagnostics & Devices/Medicine
3 Approved without Funding
Belarussian State University, Belarus, Minsk
- Republican Science-Practical Center Cardiology, Belarus, Minsk
- University of Toronto / Faculty of Applied Science and Engineering, Canada, ON, Toronto\nTampere University of Technology / Institute of Signal Processing, Finland, Tampere\nBoston University / College of Engineering, USA, MA, Boston\nNational Technical University of Athens, Greece, Athens
Project summaryProject purpose: development of biomedical image analysis, indexing and retrieval methods in cardiology.
State-of-art in medicine is characterized by widespread deployment of various diagnostic tools, based on computing and networking information technologies. Cardiologic applications employ merely all the spectrum of diagnostic technologies: radio- and ultrasound tomography, ultrasound scanning, nuclear magnetic resonance, angiography etc. Static and dynamic images are used either as examination results or for online monitoring, followed by digitizing and storage for future diagnosing, surgical assistance and cardiovascular system control of patients. Medical images can be used in educational purposes, as well as for development of automated diagnostic systems, based on expert system technology and methods of intelligent data analysis. Abundance of progress and deployment of telemedicine generate problems of operative transmission of large-sized medical images over communication channels, of their storage in local and distributed databases. Globally, immense volume of medical images is increasing in dramatic scales. Effective utilization of such data is possible only with already existing effective methods of analysis, indexing and context retrieval of medical images [Moncef Gabbouj "Muvis: A Content-Based Multimedia Indexing and Retrieval Framework". Proceedings of the Seventh International Symposium on Signal Processing and its Applications, 1-4 July 2003, Paris, France].
Intensive development of effective methods and software and hardware tools for processing of various types of complex images has been conducted in major research centers all over the world for the last couple of years. There are occasional attempts to develop image compressing and indexing systems and methods of context retrieval in databases. The most significant advances have been achieved in multimedia signal processing field (acoustic signals and video streams) [James Z. Wang. "Integrated Region-Based Image Retrieval" (monograph). Kluwer Academic Publishers, 190 pages, Dordrecht, 2001].
However, such problems with respect to medical images are rudimentary and nonsystematic. Image compression, indexing and retrieval are treated as separate tasks. Dominant method of medical image indexing is a manual one, which associates images with corresponding text descriptions. This low-effective indexing approach can be conducted by highly qualified experts and it is predicted to be used at most for reference database creation. Known methods of automated medical image indexing exploit unified low-level descriptors. That fact decreases their effectiveness in context search systems.
Thus, development of methods, algorithms and software for analysis, indexing and retrieval of medical images must be conducted in comprehensive, system fashion. Adaptability, utmost use of the apriori information about objects of study ought to be fundamental principles of image processing and context search systems.
Impact of the proposed project on progress in the research field.
In the course of the project methods, algorithms and software are to be developed, implementing system approach to image processing technology and creating a necessary basis for development of highly effective tools of disease diagnosis in cardiologic applications.
The research group will be formed of two teams. The first one consists of specialists from Belarussian State University. The second team is with Republican Science and Practice center “Cardiology”. The group will include experts on systems and devices of reconnaissance, detection, tracking and recognition of radar targets and aiming, target designation and self-homing systems. In the course of the project, BSU specialists are to develop methods, algorithms and software for analysis, indexing and retrieval of medical images. RSPC “Cardiology” experts are to form medical image databases, perform reference indexing, conduct clinical tests of the software complexes, which will be based on the developed methods and algorithms.
Project participants have significant experience in development of applied signal processing systems, in creation of software for medical applications, in the fields of theoretical, experimental and clinical cardiology. They have mastered up-to-date instrumental and biomedical diagnostic technologies.
During the accomplishment of the proposed project the following milestone stages are to be completed:
- Development and study of effective wavelet transform-based methods for biomedical image registration and processing.
- Development of algorithms for biomedical image classification, which are based on support vector machines and hierarchical hidden Markov models.
- Segmentation and analysis of biomedical images for the purpose of informative feature extraction in context search.
- Development of methods and algorithms of biomedical image context retrieval in local and distributed databases.
- Development of effective biomedical image retrieval, indexing and storage technologies.
- Creation of the software complex for analysis, indexing and retrieval of biomedical images in cardiology.
As the main result of project fulfillment the complex for analysis, indexing and retrieval of biomedical images in cardiology will be created. New scientific results will be obtained: effective wavelet transform-based methods for medical image registration and processing will be developed and studied, algorithms for medical image classification based on support vector machines and hierarchical hidden Markov models will be developed. From practical point of view effective technologies creation lets to develop highly productive cardio images analysis and search system which will help physicians to find similar clinical presentations, make predictions, fast diagnosis and prescribe treatment at early disease stages. The software complex for analysis, indexing and retrieval of biomedical cardiological images which will be developed can be used in real commercial automated systems for cardiological disease diagnosis and telemedicine.
Application of the project results. Scientific results, obtained during the project fulfillment, can be used for:
- Development of the new generation of local and distributed biomedical image databases with effective tools of context search;
- Development of automated systems for cardiological disease diagnosis;
- Further progress and improvement of facilities in cardiology.
Meeting ISTC Goals and Objectives
Project realization will allow to:
- Providing weapons scientists and engineers from BSU, particularly those who possess knowledge and skills related to development of systems and devices of reconnaissance, detection, tracking and recognition of radar targets and aiming, target designation and self-homing systems, opportunities to redirect their talents to peaceful activities;
- Promoting integration of scientists of CIS states into the international scientific community in order apply the achieved results in partnership projects;
- Supporting basic and applied research and technology development for peaceful purposes, notably in fields of health monitoring of population, considering multiple-factor influence of the Chernobyl APP disaster;
- Contributing to the solution of international technical problems- to increase effectiveness of cardiovascular disease diagnosis;
- Reinforcing the transition to market-based economies responsive to civil needs.
Information on the work volume: Duration of the project will be 36 months. Work volume can be devides to two main blocks: algorithms and methods development for image processing and practical indexing system development and testing for real data. In this connection two institutions will take part at the project. All investigations and works connected with methods and algorithms development for biomedical images indexing and recognition will be held at the Belarussian State University. All works connected with training biomedical data collecting and sorting, clinical tests of developed methods and systems will be held at the Republican Science and Practice Center “Cardiology”.
The whole project is pided into 6 connected tasks:
- development and study of effective methods for biomedical images registration and processing based on wavelet transforms;
- development of biomedical images classification algorithms based on support vectors method and hierarchical hidden Markov models;
- biomedical images segmentation and analysis in order to select informative features for context search;
- methods and algorithms development for context biomedical images search in distributed and local data bases;
- development of effective methods for search, indexing and storing of biomedical images;
- development of hardware-software complex for analysis, search and indexing of biomedical images.
Overall estimated cost of the project is will total 7146 man-days.
Role of foreign collaborators. Cooperation with the project collaborators will further to fruitful exchange of scientific information within the scope of the project research, to objective appraisal and review of the scientific results, to active participation in improvement and application of the achieved results.
Technical approach and methodology. Up-to-date digital signal processing methods will be used for analysis, indexing and context retrieval of biomedical images: wavelet transforms, hidden Markov models (including hierarchical hidden Markov models), support vector machines, busting algorithms. These methods are widely used for image processing and recognition. On this basis new methods of biomedical image context search are to be developed. They will be based on apriori information on relation between a patient’s cardiovascular system condition and his complex examination results. Fast index searxh algorithms are already developed for face identification task and can be applied The project stuff has high qualification in biomedical signal processing and experience in indexing and recognition systems development. Main points of the proposed project were examined at a series of international workshops and conferences and discussed with leading specialists at biomedical signal processing sphere. Due to the above mentioned reasons the project purpose has to be reached.