Patent No. 5218530 Method of displaying and analyzing nonlinear, dynamic brain signals

 

Patent No. 5218530

Method of displaying and analyzing nonlinear, dynamic brain signals (Jastrzebski, et al., Jun 8, 1993)

Abstract

The invention is a method to aid analysis of signals, such as electroencephalograms, pursuant to modern mathematical theories of nonlinear dynamical processes, sometimes referred to as chaotic dynamics or chaos theory. It employs graphic display and visual inspection of relatively less filtered, non-averaged raw test data, including raw data heretofore considered random or asynchronous `noise`. The invention enables reversible decomposition of selected elements of graphic portraits of raw signal data to identify subsets of the depicted raw data which correspond to visually-identified, manually-selected patterns from within the graphic portrait. The identified subsets of raw data can be segregated even though a precise mathematical description of the visually identified pattern is unknown. The invention further comprises a variety of techniques for displaying four or more variables and for enhancing visual discrimination of patterns within computer generated graphic phase space portraits, and conceptions for overlaying symbols onto graphic points representing stimulus and response events concurring with particular signal samples in the phase space portrait. The invention also comprises subsets of pattern-generating signal data identified by the method of the invention and thus made available for further computer or other operations separately from the full data set.

Notes:

There are no related applications. No federally-sponsored research and development is involved.

SUMMARY OF THE INVENTION

The invention is a method for applying the new theories of chaotic dynamics to brain signals. It enables a computer operator to automatically identify in a recorded stream of brain signal data the subsets of signal data which correspond to manually-selected patterns observed in multi-variable, reversibly-transformable, phase space portraits. The data selection is enabled even though no algorithm describing the observed pattern is known. This data selection capability is combined with enhanced capabilities to display and manipulate multi-variable phase space portraits, thereby increasing the ability to visually discriminate patterns in the stream of data for selection.

The invention draws a large number of variables into a graphic phase space portrait for display on a computer monitor. It then enables manual selection of subsets of drawing elements from within the displayed portrait corresponding to visually identified patterns, and automatically decomposes the selected drawing elements to identify the subset of signal data corresponding to the visually-identified patterns. Each phase space portrait can simultaneously display multiple variables, including three spatial coordinates corresponding to three separate signal detectors, plus scalar magnitude, direction of change, and color-coded time sequence. In principle, such portraits also could display another variable corresponding to drawing line type and could superpose symbols to depict stimulus and response events relative to the time sequence of recorded signals, though the current model does not implement these features.

The invention also creates and depicts composite phase space portraits of larger sets of four or more signal detectors by means of overlays of simultaneous phase space portraits from a plurality of three-detector subsets. Variable colors serve to visually distinguish the contributions of each layer to the composite, thus enabling visual discrimination of the particular subsets of detectors which provide the signals of most interest. In principle, variable line types also could be used to distinguish layers.

One aspect of the invention may be viewed as a computer-assisted manual sieve to identify from a stream of data the subsets of data which correspond to an observed pattern where no algorithm describing the observed pattern is known. The resulting subsets of pattern-containing data then are available for more intensive analysis and other operations, including efforts to define a descriptive algorithm for the observed pattern and efforts to create a `template` for automated computer-aided pattern recognition.

CONCEIVED USES OF THE INVENTION

One of the Inventors' conceptions is that a human subject can be given a switch to insert flags or markers into the signal data stream corresponding to the moment the subject experiences conscious thoughts, perceptions of external stimuli, or other conscious processes. Similarly, the host computer can trigger a stimulus and simultaneously insert a flag denoting the stimulus into the collected data stream. Thus, both stimulus and response events can be recorded within the data stream in conjunction with simultaneously collected brain signals.

Another of the Inventors' conceptions is that the invention enables computer-assisted manual identification of pattern-generating subsets of raw data. The identified subsets of raw data can then be segregated to serve as empirical data patterns against which computer-aided pattern comparison can be made, even though no precise mathematical description of the segregated raw data can be programmed into the computer. The Inventors further conceive, for example, that such segregated subsets of pattern-generating raw data can be used to `train` computer neural networks to recognize the patterns in such raw data subsets without first defining a descriptive mathematical formula for such raw data subsets.

The Inventors conceive that an empirical library of subsets of raw data comprising brain activity reflecting responses to stimuli, perceptions, and thought processes, can be created by so selecting such visually distinctive, patterned subsets of raw data out of the displayed phase space portraits. The Inventors further conceive that such subsets can be employed as computer-recognizable patterns pattern recognition programs.

The Inventors conceive that the development of computer recognizable patterns in brain signals, in turn, will enable `on-the-fly` or real time analysis of brain signals by computer-aided pattern comparison to such an empirical library of brain signal patterns.