Project Profile
Spectral Analysis of Biomedical Signals
Signal Analysis
Project Director: Julie Greenberg
jgreenbe@mit.edu
Project Participants: Lori Breslow, Sean Brophy, Leonardo Cedolin, John Newman, Natalie Smith, Minnan Xu
Contents
Audience
Learning objectives
Sequence of learning activities
Innovations
Placement in the taxonomy
Assessment
Dissemination timetable
References
Go back to the list of projects
Revised: February 23, 2004
If you are interested in adopting this material, please contact Julie Greenberg (jgreenbe@mit.edu)
Audience
This module is intended for advanced undergraduates or graduate students familiar with the basics of signal processing. It is suitable for use near the end of an introductory signals and systems course or in a subsequent signal processing course. Before attempting this module, students should be familiar with the following topics: sampling, the discrete-time Fourier transform, and convolution and multiplication in time and frequency. Familiarity with windowing is also helpful, although that topic is thoroughly reviewed in the module.
Go back to the top of this profile
Learning Objectives
Major Learning Objective
Students can analyze and interpret the frequency content of biomedical signals
Detailed Learning Objectives
Students are able to:
· analyze the effects of multiple variables on a frequency-domain representation
· select parameters to perform frequency analysis of a signal, given desired specifications
· interpret a given frequency-domain representation, given the parameters used
· infer the parameters used, given a frequency-domain representation
Key Concepts
These learning objectives require understanding of:
factors affecting amplitude resolution
factors affecting frequency resolution
effect of changing window length
effect of changing window shape
effect of changing DFT length
effect of changing multiple parameters simultaneously
Go back to the top of this profile
Sequence of Learning Activities
This module consists of a combination of on-line and in-class activities organized in a Legacy cycle:
The challenge is to design a system for monitoring a patient’s electrocardiogram (ECG) signals in a hospital setting and sounding an alarm when a life-threatening ventricular arrhythmia occurs. The background is introduced in a prior lecture on cardiac electrophysiology, and the actual challenge is presented to the students at the start of a class session.
Immediately after presentation of the challenge in class, students break up into small groups to generate ideas. With access to sample ECG signals and MATLAB(R), they formulate an initial approach to solving the challenge. Next, the class reconvenes to hear multiple perspectives. Each small group presents their ideas. The instructor solicits student reactions to the various ideas and moderates a discussion to identify their strengths and weaknesses. The instructor may also suggest additional approaches and/or contribute expert reactions and insights to the students’ ideas.
Students research spectral analysis and revise their understanding by completing online exercises and by attending a lecture on Fourier spectral analysis. The online exercises consist of a web-based tutorial, centered on an interactive demonstration of Fourier spectral analysis. The demonstration allows students to select key parameters and perform spectral analysis of cosine, electrocardiogram, and speech signals (http://web.mit.edu/6.555/www/matweb/demo.html). The tutorial consists of a series of questions, accompanied by links to text summaries of key concepts, a glossary of terms, and hints specific to particular questions. The questions are structured to guide students as they progressively develop an understanding of the key concepts in spectral analysis, making explicit connections between phenomena observed in the demonstration and the learning objectives. (Access to the tutorial requires a username and password; please contact jgreenbe@mit.edu for a guest account.) The lecture on Fourier spectral analysis is deliberately presented late in the Legacy cycle, after students have been exposed to the topic and have had extensive hands-on experience with the online exercises.
Students test their mettle the following week in a four-hour lab session dedicated to designing and testing their ventricular arrhythmia detector, implemented in MATLAB (R). Finally, they go public by completing a detailed lab report and taking a quiz that includes questions on Fourier spectral analysis.
Go back to the top of this profile
Innovation
Prior to the development of this module, students studying Fourier spectral analysis attended a lecture on the topic and completed a similar lab project on the design of a ventricular arrhythmia detector. In many cases, laboratory reports indicated that students had selected the parameters of the Fourier spectral analysis processing by trial and error, rather than by applying the principles covered in lecture. Performance on quizzes furthered our impression that students were not obtaining the desired understanding of the material.
In the current module, having an explicit challenge provides students with clear motivation for the subsequent activities. The interactive demonstration allows students to engage in detailed manipulations of the relevant parameters, while the structured tutorial questions encourage constructive use of the simulation, as opposed to “fiddling” with parameters. The relatively late placement of the lecture prepares students to learn the details of Fourier spectral analysis and provides an opportunity to further revise their developing understanding of the topic.
Go back to the top of this profile
Placement in the Taxonomy
VaNTH’s Signal Analysis taxonomy is currently under revision..
Go back to the top of this profile
Assessment
We collected performance data for test and control groups (that is, students who studied spectral analysis with and without the module) and developed a standardized scoring system for assessing student understanding of the key concepts in Fourier spectral analysis. We found that students in the test group demonstrated significantly better understanding than students in the control group. In addition, survey results and comments indicate a postive reaction to the interactive tutorial and demo, as well as the structure provided by the HPL framework.
Go back to the top of this profile
Timetable
Since 2001, this module has been used each spring in a course titled Biomedical Signal and Image Processing (HST582J/6.555J) offered by the Harvard-MIT Division of Health Sciences and Technology. It is currently available for use at other institutions; please contact the Julie Greenberg (jgreenbe@mit.edu) for details.
Go back to the top of this profile
References
Greenberg, J.E., Smith, N.T., and Newman, J.H. “Instructional Module in Fourier Spectral Analysis, Based on Principles of ‘How People Learn’,” J. Engineering Education 92, pp. 155-165, April 2003.
Greenberg, J.E., Delgutte, B., and Gray, M.L. “Hands-On Learning in Biomedical Signal Processing,” IEEE Engineering in Medicine and Biology Magazine 22, pp. 71-79, July/Aug. 2003.
Need to insert appropriate links to VaNTH website.
Go back to the top of this profile