WO2012042036A1

WO2012042036A1 - Method and device for converting time-varying signals acquired over a large number of recording channels

Info

Publication number: WO2012042036A1
Application number: PCT/EP2011/067155
Authority: WO
Inventors: Marco De Curtis; Vadym Gnatkovskyy
Original assignee: Fondazione I.R.C.C.S. Istituto Neurologico "Carlo Besta"
Priority date: 2010-10-01
Filing date: 2011-09-30
Publication date: 2012-04-05
Also published as: IT1401999B1; ITMI20101801A1

Abstract

The method for converting time-varying signals (Aj) acquired over a large number of recording channels comprises the steps of: acquiring all the signals (Aj) for an acquisition time that is equal for all the signals computing, by means of the Fourier transform, the power spectrum (Fij) of each signal (Aj) in consecutive time windows (Wi) memorising the power spectrum (Fij) of each signal (Aj) in each time window (Wi) attributing and memorising a correlation between an intensity value of a colour and the value of the integral (Hij,k) of the power spectrum (Fij) of any signal (Aj) in any time window (Wi), within a given neighbourhood of any frequency of interest (k) of a. predefined interval of frequencies of interest generating a map with cells (Cf,g), wherein each signal (Aj) in each time window (Wi) has associated with it a corresponding cell (Cf,g) of the map, and in which, when any frequency (k) of interest is selected, the value of the integral (Hi j,k) of the memorised power spectrum (Fij) of each signal (Aj) in each time window (Wi), within the given neighbourhood of the selected frequency of interest (k), is computed in order to attribute the colour intensity to each of the cells (Cf,g) of the map.

Description

Title: "METHOD AND DEVICE FOR CONVERTING TIME-VARYING SIGNALS ACQUIRED OVER A LARGE NUMBER OF RECORDING CHANNELS".

DESCRIPTION

The present invention relates to a method and a device for converting time- varying signals acquired over a large number of recording channels.

In many fields of application, for the purpose of studying a phenomenon it is necessary to conduct an interpretative analysis of time-varying signals associated with the phenomenon one wishes to study.

The analysis becomes extremely complex when the phenomenon is studied through the acquisition of a wide range of signals that must be comparatively analyzed.

This problem is felt, for example, during the analysis of surface electroencephalographic and intracranial signals acquired with many channels (>100) for pre-surgical evaluation of epileptic patients resistant to pharmacological treatments.

In this case in particular, the neurologist has at his disposal an electroencephalogram tracing as a function of time for each acquisition channel and possibly the power spectrum of the frequencies associated with each tracing. It is obvious that the interpretative analysis of data for the frequencies of interest will be long and complicated, especially due to the fact that the data are not provided to the clinician in the most appropriate form.

The technical task that the present invention has set itself is therefore to provide a method and a device for converting time-varying signals acquired over a large number of recording channels which enable an easy, rapid and effective interpretative analysis of the phenomenon they represent.

The technical task, as well as this and other objects, according to the present invention, are achieved by providing a method for converting time-varying signals acquired over a large number of recording channels, characterised in that it comprises the steps of:

• acquiring all the signals for an acquisition time that is equal for all the signals

• computing, by means of the Fourier transform, the power spectrum of each signal in consecutive time windows

• memorising the power spectrum of each signal in each time window

• attributing and memorising a correlation between an intensity value of a colour and the value of the integral of the power spectrum of any signal in any time window, within a given neighbourhood of any frequency of interest of a predefined interval of frequencies of interest

• generating a map with cells, wherein each signal in each time window has associated with it a corresponding cell of the map, and in which, when any frequency of interest is selected, the value of the integral of the memorised power spectrum of each signal in each time window, within the given neighbourhood of the selected frequency of interest, is computed in order to attribute the colour intensity to each of the cells of the map.

Preferably, said correlation consists of a relationship of direct proportionality.

Preferably, each signal in the consecutive time windows has associated with it a row of consecutive cells of the map.

Preferably, said map has as many stacked rows of cells as there are signals. One of the salient aspects of the invention is that a rapid scroll selection command is provided for automatically updating the map to a frequency of interest.

Preferably, said rapid scroll selection command automatically and continuously scrolls along at least one interval of frequencies of interest for immediate automatic updating of the map to the frequencies of said interval,

in a preferred embodiment of the invention, said signals are electrophysiological and in particular electroencephalographic.

The present invention also relates to a device for converting electrophysiological signals, comprising a plurality of electrodes for multi-channel acquisition of the electrophysiological signals, at least one visual display unit and an electronic processor suitable for converting the electrophysiological signals according to the method described above.

In a preferred but not exclusive application of the invention, the above-described method and/or device are used to study the information acquired from epileptogenic zones of a mammal and/or to verify the reproducibility of electroencephalographic patterns during an epileptic crisis.

This present invention in more detail provides an original method to immediately represent and analyse frequency power changes simultaneously recorded in a large number of channels (more then 100).

The present method is based on producing a data file by converting time-varying signals (Aj) acquired over a large number of recording channels, characterised in that it comprises the steps of:

- memorising (saving in data file) the power spectrum (Fij) of each signal (Aj) in each consecutive time window (Wi); - representation of the desired frequency power by generating a colour intensity map with cells (Cf,g), wherein each signal (Aj) in each time window (Wi) has associated with it a corresponding cell (Cf,g) of the map, and in which, when any frequency (k) of interest is selected, the value of the integral (Hi,j,k) of the memorised power spectrum (Fi,j ) of each signal (Aj) in each time window (Wi), within the given neighbourhood of the selected frequency of interest (k), is computed in order to attribute the colour intensity to each of the cells (Cf,g) of the map.

Further characteristics and advantages of the present invention will become more apparent from the description of a preferred, but not exclusive embodiment of the method and device for converting time-varying signals acquired over a large number of recording channels according to the invention, illustrated by way of non-restrictive example in the appended figures, in which:

figure 1 shows the logical sequence of the steps necessary in order to visually display, in the map representing all the signals acquired in all the acquisition channels, the temporal evolution of the integral of the power spectrum within the given neighbourhood of a specific frequency of interest for a signal acquired in a specific acquisition channel;

figure 2 shows a first, a second and, respectively, a third static image of the map which visually display the temporal evolution of the integral of the power spectrum within the given neighbourhood of a first, a second and, respectively, a third specific frequency of interest. Said static images represent photograms of the dynamic image of the map which is obtained by scrolling the rapid scroll selection command automatically and continuously along at least one interval of frequencies of interest for immediate automatic updating of the map to the frequencies of the interval.

With reference to the aforementioned figures, the conversion method envisages: an acquisition time that is equal for all the signals; computing, by means of the Fourier transform, the power spectrum of each recorded signal in consecutive time windows; memorising the power spectrum of each signal in each time window; attributing and memorising a correlation between an intensity value of a colour and the value of the integral of the power spectrum of any signal in any time window, within a given neighbourhood of any frequency of interest of a predefined interval of frequencies of interest; and generating a map with cells, wherein each signal in each time window has associated with it a corresponding cell of the map, and in which, when any frequency of interest is selected, the value of the integral of the memorised power spectrum of each signal in each time window, within the given neighbourhood of the selected frequency of interest, is computed in order to attribute the colour intensity to each of the cells of the map.

Let us indicate as Aj the j-th signal acquired from the j-th acquisition channel, where the index "j" is an integer ranging from 1 to M, and M in turn also corresponds, therefore, to the total number of recording channels.

Let us indicate as Wi the i-th time window, where the index "i" is an integer ranging from 1 to N, and N in turn corresponds to the total number of time windows.

Let us indicate as Fi,j the power spectrum of the j-th signal in the i-th time window, where the index "j" is the integer as defined above.

Let us indicate as Hij,k the integral within the given neighbourhood of the frequency k of the power spectrum of the j-th signal in the i-th time window. The frequency k is any frequency selectable within an interval of frequencies of interest.

Let us indicate as Hj,k the temporal evolution in the windows Wi of the integral Hi,j,k within the given neighbourhood of the frequency k of the power spectrum of the j-th signal.

By way of example, for an acquisition time of 60 seconds, the time windows Wi all have a duration of 1 second, and each time window Wi will overlap with the preceding time window by 0.2 seconds.

The computation of the integral Hi,j,k of the power spectrum of a signal within the neighbourhood of a frequency of interest is illustrated by way of example in figure 1 in reference to a signal A₅₇, and for a frequency of interest k=l 15±5 Hz.

First the power spectra F_{1 ,57}, F_2,57 F_{N-1 ,57}, F_N,57 are evaluated. Subsequently, the integrals Η 1,57,115±5, H_2,57, _{1 15±5}..... H_N__{1,157, 1 1 5±5 ,} H_{N,57, 1 15±5} are evaluated.

At this point the values of the integrals Η_{1,57, 115±5 ,} Η_{2,57, 115±5}..... H_{N- 1 ,57, 1 15±5},

HN,57,115 ±5 are aggregated, with the same temporal sequence as the time windows Wi , in order to obtain H₅₇, _{115 ±5 .}

H_57,115±5 thus represents the temporal evolution of the integral Hi._{57, 115±5} in the time of acquisition of the signal A₅₇.

It is now possible to visually display H_57115±5 in the map.

The map consists of a matrix of MxN cells C_f,g ordered in a number of rows equal to M (total number of signals) and a number of columns equal to N (total number of time windows).

C_f,g thus indicates the cell obtained by the intersection of the row of cells "f ' with the column of cells "g".

The j-th signal is associated with the j-th row of cells, whereas the i-th time window Wi is associated with the i-th column of cells.

Once the value of k is fixed, the integral Hi,j,k is displayed by the cell C_f,g where f=j and g=i.

In the case considered, therefore, the temporal evolution of the integral H_{i 57 1 15±5} is displayed in the 57-th row of cells, in which, more precisely, the integral H_{1;57il t5±5} is displayed in the cell C_57,1 , the integral H_2,57, 115±5 is displayed in the cell

C_57,2.... the integral ^ΗΝ-1 ,57_, 115±5 is displayed in the cell C_{57, N-1}, and the integral

H_N,57,115±5 is displayed in the cell C_57;N.

There being envisaged a relationship of direct proportionality between the value of the integral H_{i 57 1 15±5} and the colour intensity of the cell C_57,g of the map in which it is represented, the cells C_57,g with a highest colour intensity represent the integrals Hi_{,57, 1 1 5±5} having the greatest value, whereas the cells C_57,g with the lowest colour intensity represent the integrals Η_i,57 , _115±5 having the least value. In figure 1 the peak in the temporal pattern of the integral Hi,_{57, 115±5} is indicated with an asterisk.

By repeating the conversion method for each recorded signal Aj,

the temporal pattern of the integral Η_i,j, _{1 15±5} of the power spectrum associated with the frequency of interest of 1 15±5 Hz can be visually displayed on the map, for all the recording channels.

One of the salient aspects of the invention consists in the fact that it includes a rapid scroll selection command for automatically updating the map to a frequency of interest.

In particular, the rapid scroll selection command, indicated as 1 in figure 2, automatically and continuously scrolls along at least one interval of frequencies of interest and enables the map to be defined and automatically updated to the frequencies of said interval.

In figure 2 the interval of frequencies of interest, each of which is evaluated in a neighbourhood of ±5 Hz, ranges between 0 and 250 Hz.

The command 1 can select a frequency of interest and update the map to the selected frequency of interest, or it can automatically and continuously scroll along the entire interval of interest in order to automatically and rapidly update the map to all the frequencies of the interval.

Figure 2 shows the map updated to the frequencies of interest, 10 ±5 Hz, 30 ±5 Hz and 85 ±5 Hz.

The conversion method is carried out by means of a device based on an electronic processor equipped with means for acquiring the signals Aj, means for memorising the signals Aj and power spectra Fi,j and means for visually displaying the map. Naturally, for the purpose of updating the map, the control unit of the electronic processor acquires the frequency of interest and draws on the memorised power spectra Fij in order to compute the integrals Hi,j,k in the neighbourhood determined by the frequency of interest, and repeats the computation each time the value selected for the frequency of interest changes.

A case of particular interest is the one in which the signals are electrophysiological, and in particular they are electroencephalogram signals acquired in recording channels in parallel during monitoring of a patient.

In such a case the conversion device has signal acquisition electrodes and can be used as a tool to study particular pathologies, typically epilepsy, in particular to study the information acquired from epileptogenic zones of a mammal and/or to verify the reproducibility of patterns associated with an epileptic crisis.

It has been ascertained in practice that the method and the device for converting signals according to the invention are particularly advantageous in that that they enormously simplify the interpretative analysis of a complex phenomenon which requires the acquisition of signals in a large number of recording channels.

In particular, with the conversion of the temporal signals in an image to a gradation of colour intensity which is correlated with the pattern of the integral of the power spectrum of all the signals acquired within a given neighbourhood of any selected frequency of interest and which, thanks to a command, is immediately updated also to all of the frequencies present within an interval of frequencies of interest, the clinician who analyses a complex phenomenon is offered a representation thereof which enables an effective, simple, intuitive and rapid analysis.

The present method was developed in order to rapidly analyze changes in frequency domain of electrophysiological data recorded from large numbers of contacts (more than 100).

The main original features of the present method are: the creation of data files of computed power spectra calculated on more that 100 channels for further rapid extraction of values without recalculation; the simultaneous power spectra representation of a selected frequency range for all analysed channels (in the graph of fig.l X axis is time, Y axis represents all recorded channels and Z axis, i.e. colour intensity, is the power spectrum for a frequency, or frequency band, of interest); and the frequency scanning, continuous manual selection of the frequency of interest by means of a command cursor (fig. 2) sliding across the full spectrum of frequencies. According to the method described above, the method of the present invention is able to rapidly and simultaneously represent for a large number of channels the frequency power for a desired frequency range. Power frequency is represented as a colour intensity map only for one frequency range at once. This approach permits rapidly and simultaneously evaluation of power frequency changes in temporal and spatial domain for all channels.

The method and device for converting signals thus conceived are susceptible of numerous modifications and variants, all falling within the scope of the inventive concept; moreover, all the details may be replaced with other technically equivalent elements.

In practice, the materials used, as well as the dimensions, can be any whatsoever according to need and the state of the art.

Claims

1. A method for converting time- varying signals (Aj) acquired over a large number of recording channels, characterised in that it comprises the steps of:

• acquiring all the signals (Aj) for an acquisition time that is equal for all the signals

• computing, by means of the Fourier transform, the power spectrum (FiJ) of each signal (Aj) in consecutive time windows (Wi)

• memorising the power spectrum (Fij) of each signal (Aj) in each time window (Wi)

• attributing and memorising a correlation between an intensity value of a colour and the value of the integral (Hij,k) of the power spectrum (Fij) of any signal (Aj) in any time window (Wi), within a given neighbourhood of any frequency of interest (k) of a predefined interval of frequencies of interest

• generating a map with cells (Cf,g), wherein each signal (Aj) in each time window (Wi) has associated with it a corresponding cell (Cf,g) of the map, and in which, when any frequency (k) of interest is selected, the value of the integral (Hi j,k) of the memorised power spectrum (Fij) of each signal (Aj) in each time window (Wi), within the given neighbourhood of the selected frequency of interest (k), is computed in order to attribute the colour intensity to each of the cells (Cf,g) of the map.

2. The method for converting signals according to claim 1, characterised in that said correlation consists in a relationship of direct proportionality.

3. The method for converting signals according to any of the preceding claims, characterised in that each signal (Aj) in the consecutive time windows (Wi) has associated with it a row of consecutive cells (Cf,g) of the map.

4. The method for converting signals according to the preceding claim, characterised in that said map has as many stacked rows of cells (Cf,g) as there are signals (Aj).

5. The method for converting signals according to any of the preceding claims, characterised in that it includes a rapid scroll selection command (1 ) for automatic updating of the map to a frequency of interest (k).

6. The method for converting signals according to the preceding claim, characterised in that said rapid scroll selection command ( 1) automatically and continuously scrolls along at least one interval of frequencies of interest (k) for immediate automatic updating of the map to the frequencies (k) of said interval.

7. The method for converting signals according to any of the preceding claims, characterised in that said signals (Aj) are electrophysiological.

8. The method for converting signals according to the preceding claim, characterised in that said electrophysiological signals (Aj) are electroencephalographic.

9. A device for converting electrophysiological signals, comprising a plurality of electrodes for multi-channel acquisition of the electrophysiological signals (Aj), at least one visual display unit and an electronic processor suitable for converting the electrophysiological signals (Aj) according to the method of any of the preceding claims.

10. Use of a method and/or device according to any of the preceding claims to study the information acquired from epileptogenic zones of a mammal and/or verify the reproducibility of electroencephalographic patterns during an epileptic crisis.

1 1. Method for converting signals, characterised by a rapid scroll selection command cursor (1) that manually and continuously scrolls to select one interval of frequencies of interest (k) for immediate automatic updating of the colour map to the frequencies (k) of said interval, selected frequency range used to rapidly and continuously withdraw and represent an information about frequency power from previously saved data file that has as many stacked rows of cells (Cf,g) as there are signals (Aj), to have all computed data of quantified frequency content for all channels for further rapid extraction of values without recalculation.

12. Software implemented device for converting electrophysiological signals, comprising saving quantified data in file structure according to the method of any of the preceding claims, at least one visual display unit and an electronic processor suitable for converting and representing the electrophysiological signals (Aj) according to the method of any of the preceding claims.

13. Use of a method and/or software implemented device according to any of the preceding claims to study the power frequency changes in time-varying signals.