Analysis of Dynamic Brain Imaging Data

P. P. Mitra,B. Pesaran
DOI: https://doi.org/10.1016/S0006-3495%2899%2977236-X
1998-04-21
Abstract:Modern imaging techniques for probing brain function, including functional Magnetic Resonance Imaging, intrinsic and extrinsic contrast optical imaging, and magnetoencephalography, generate large data sets with complex content. In this paper we develop appropriate techniques of analysis and visualization of such imaging data, in order to separate the signal from the noise, as well as to characterize the signal. The techniques developed fall into the general category of multivariate time series analysis, and in particular we extensively use the multitaper framework of spectral analysis. We develop specific protocols for the analysis of fMRI, optical imaging and MEG data, and illustrate the techniques by applications to real data sets generated by these imaging modalities. In general, the analysis protocols involve two distinct stages: `noise' characterization and suppression, and `signal' characterization and visualization. An important general conclusion of our study is the utility of a frequency-based representation, with short, moving analysis windows to account for non-stationarity in the data. Of particular note are (a) the development of a decomposition technique (`space-frequency singular value decomposition') that is shown to be a useful means of characterizing the image data, and (b) the development of an algorithm, based on multitaper methods, for the removal of approximately periodic physiological artifacts arising from cardiac and respiratory sources.
Neurons and Cognition,Quantitative Methods
What problem does this paper attempt to address?
The problem that this paper attempts to solve is that the data sets generated by modern brain - function imaging techniques (such as functional magnetic resonance imaging (fMRI), intrinsic and extrinsic contrast optical imaging, and magnetoencephalography (MEG)) are huge and complex in content. How to effectively analyze and visualize these data to separate signals from noise and characterize signals? Specifically, the paper focuses on two main aspects: 1. **Characterization and Suppression of Noise**: The data collected by modern imaging techniques contains a large amount of noise components, which may be derived from instruments or physiological factors. The paper has developed specific techniques to identify and remove these noises, thereby improving the quality of data. 2. **Characterization and Visualization of Signals**: After removing the noise, how to extract useful signals from the data and effectively characterize and visualize them? The paper proposes the method of multivariate time - series analysis, especially using the multi - window spectral analysis framework, to process these complex brain - imaging data. The focus of the paper is on developing analysis and visualization techniques applicable to dynamic brain - imaging data. These techniques can handle non - stationary data within a short - time window and can efficiently represent and analyze data in the frequency domain. It is particularly worth mentioning that the paper proposes a "spatial - frequency singular - value decomposition" technique for characterizing image data and has developed an algorithm based on the multi - window method for removing periodic physiological artifacts caused by the heart and respiration. Keywords include: spectral analysis, functional magnetic resonance imaging (fMRI), optical imaging, magnetoencephalography (MEG), multivariate time - series analysis, singular - value decomposition.