You are here

F7AMBPMZD - Advanced Methods of Data Analysis and Processing

Code Completion Credits Range Language
F7AMBPMZD KZ 3 1P+1C English
Garant předmětu:
Václava Piorecká
Lecturer:
Václava Piorecká
Tutor:
Václava Piorecká, Marek Piorecký, Jan Štrobl
Supervisor:
Department of Biomedical Technology
Synopsis:

This course comprehends/deals methods of biosignal generation, biosignal acquisition and basic parameters of biosignals required for diagnostics. Methods and algorithms for biosignal processing, analysis and evaluation used for biological signals, mainly electrophysiological signals. Preprocessing, filtering, time and frequency analysis. Use of modern spectral analysis methods. Visualisation of results, topographic mapping, method of compressed spectral arrays (CSA). Adaptive segmentation of non-stationary signals is discussed. Application of methods using artificial intelligence. Methods of automated signal classification - supervised/unsupervised, cluster analysis, learning classifier. Artificial neural networks (ANN). Practical application of biosignal processing. Case studies of ANN application on epileptogenic recordings and neural recordings in general. Genetic algorithms and simulated annealing is presented.

Requirements:

Active participation in exercises is mandatory. One absence is allowed without giving a reason. Other absences must be documented. During the exercises, student will complete scored tasks, 45 points can be obtained from the scored tasks. At the end of the semester, student will write a final test with value of 55 points. Credit is given if the student obtains at least 50% of the points from the exercise (ECTS grading scale).

Syllabus of lectures:

1. Introduction to biosignal processing. Motivation. Characteristics of basic biosignals: EEG, EOG, EMG, EKG, fMRI. Simultaneous recording. Artefacts. Genesis, source, diagnostic application. Application fields.

2. Fixed and adaptive segmentation. Stationary and non-stationary signals. Basic methods. Multichannel on-line adaptive segmentation. Feature extraction. Parameter settings. Advantages and limitations of the methods. Other segmentation algorithms.

3. Spectral analysis of biosignals - application. Basic methods. Parametric and non-parametric methods. Periodogram, AR model, LDR algorithm. Practical problems of spectral estimation. Cross-spectrum, coherence and phase spectrum. Methods of compressed spectral arrays (CSA). Digital filtering of biosignals, design of IIR and FIR filters.

4. Case studies of biosignal processing in practice.

5. Topographic mapping of electro-physiological activity. Visualisation. Principle of brain mapping. Amplitude and frequency mapping. Interpolation. Iterative method for generating maps. Animation. 3D-spline interpolation. Isopotential mapping. Application in clinical diagnostics. Dynamic mapping. LORETA.

6. Methods of automated classification I.: Unsupervised learning. Metrics. Normalisation. Basic algorithms of cluster analysis. K-means algorithm. Density based methods of classification. Examples of classification on simulated data and on real EEG recordings. Application of fuzzy set approach for increasing homogeneity of the classes. Optimal number of classes. Limits and restrictions of cluster analysis.

7. Methods of automated classification II.: Learning classifiers. Comparison of supervised and unsupervised learning. Examples of application of basic learning algorithms for classification on simulated and real data. Neural networks and methods of artificial intelligence.

Syllabus of tutorials:

1. Hidden information extraction methods and reduction of dimension.

2. EEG signal acquisition.

3. Spectrum estimation using parametric methods / ARMA models for EEG prediction.

4. Spectral analysis and synthesis. Digital filtering of the signal.

5. Visualisation of results of spectral analysis, method of compressed spectral arrays (CSA).

6. Amplitude and frequency brain mapping.

7. Learning classifiers, simulated data.

Study Objective:

Comprehend and be able to apply methods of biosignal processing.

Study materials:

Mandatory:

1.SÖRNMO, Leif a Pablo LAGUNA. Bioelectrical signal processing in cardiac and neurological applications. Amsterdam: Elsevier Academic Press, ©2005. xiii, 668 s. ISBN 0-12-437552-9

2.SANEI, Saeid a Jonathon A. CHAMBERS. EEG signal processing. Chichester: Wiley, ©2007. xxii, 289 s. ISBN 978-0-470-02581-9

3.COHEN, Mike X. Analyzing neural time series data: theory and practice. Cambridge, Massachusetts: MIT, ©2014. xv, 578 s. Issues in clinical and cognitive neuropsychology. ISBN 978-0-262-01987-3

Recommended:

4.MALMIVUO, Jaakko. a Robert PLONSEY. Bioelectromagnetism: principles and applications of bioelectric and biomagnetic fields. New York: Oxford University Press, 1995. xxii, 482 s., příl. ISBN 0-19-505823-2

5.PRINCIPE, José C., Neil R. EULIANO a W. Curt LEFEBVRE. Neural and adaptive systems: fundamentals through simulations. New York: Wiley, c2000. ISBN 0-471-35167-9.

6.HAYKIN, Simon S. Neural networks and learning machines. 3rd ed. New York: Pearson, c2009. ISBN 978-0-13-147139-9.

Note:
The course is a part of the following study plans:
Downloads: