JP2004305704A - Apparatus for classifying action potentials of neurocyte and its program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To realize a constant and long-term observation of action potentials of neurocytes if a physical relationship between a microelectrode and a neurocyte may change. <P>SOLUTION: Action potentials of a plurality of neurocytes 10a, 10b are detected at an electrode recording point 2 provided on a tip of a microelectrode for observing action potentials of neurocytes piercing a cranial nerve or a nerve fascicle tissue. The output data are sequentially classified by each neurocyte per short interval, within which supposedly a position of the microelectrode changes not so much. Based on the classified data obtained from consecutive short intervals, changes of feature extractions over time are followed by sequential calculation of feature extractions corresponding to each neurocyte, by which action potentials are to be unified and the data can be reclassified. All the procedures can be executed by a classifying program with a computer. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

The present invention relates to a device for observing a nerve cell activity of a human or a non-human animal with a microelectrode inserted into a nerve cell tissue, and in particular, a relative positional relationship between a microelectrode and a nerve cell around the microelectrode is changed. The present invention also relates to a method for observing the activity of each nerve cell with high accuracy over a long period of time by tracking changes in feature amounts such as the amplitude and duration of nerve cell action potential.

    Techniques for elucidating brain functions from the activities of neuronal populations are fundamental tools in brain science. Elucidation of the function of individual nerve cells is indispensable for elucidating brain functions, and conventionally, single nerves that classify action potentials for each nerve cell based on features such as amplitude and duration of nerve cell action potential It is performed by a cell activity observation method or a multiple nerve cell activity observation method (for example, Patent Document 1). Here, the feature amount is an amount that changes with movement of the microelectrode even if the action potential of the same nerve cell is, for example, the amplitude and duration of the nerve cell action potential, and a combination thereof. Refers to the amount obtained.

However, changes in the positional relationship between the microelectrodes and the nerve cells associated with changes in the amount of brain tissue strain during respiration, heartbeat, body movement, and electrode insertion change feature quantities such as the amplitude and duration of nerve cell action potentials. Therefore, it was difficult to observe the activity of nerve cells for a long time.

FIG. 5 is a diagram illustrating the fluctuation of the amplitude and the duration of the action potential with respect to the fluctuation of the position of the microelectrode 1. The example in the figure is a case where the activity (firing) of the nerve cell 10a is observed. The amplitude and duration of the nerve cell action potential were calculated using the amplitude and time difference as shown in FIG. The maximum negative amplitude value of the spike-like action potential waveform is defined as the amplitude, and the time difference between the time at that time and the time at which the next positive maximum amplitude value is obtained is divided by the reference time difference of 300 microseconds to obtain the duration. Coefficient. FIG. 5A shows an example in which the microelectrode 1 moves vertically by a distance y, and FIGS. 5B and 5C show recording points 2d at the tip of the electrode 1 inserted into the brain surface of the rat. The amplitude (average value) and the duration coefficient (average value) of the action potential at the recording points 2d and 2c with respect to the depth from the brain surface are shown, respectively.

As a result, the action potential amplitude (ch. 4) of the nerve cell 10b observed at the recording point 2d increases when the electrode insertion depth is shallower than 1480 microns, and gradually decreases when the electrode insertion depth becomes deeper than that. You. That is, at the electrode insertion depth of 1480 microns, the distance between the nerve cell 10b and the recording point 2d is the shortest, and the action potential amplitude is the maximum. On the other hand, the action potential amplitude (ch.3) of the nerve cell 10b observed at the recording point 2c located at a position 80 microns shallower than the recording point 2d has the maximum value at an electrode insertion depth of 1560 microns. Also, it can be seen from the duration coefficient that the duration of the action potential waveform of the nerve cell becomes shorter with the electrode insertion depth.

From these results, the action potential amplitude changes according to the distance between the recording points 2d and 2c and the nerve cell 10b, and the action potential amplitude is large when the distance is short and small when the distance is long. . In addition, the duration of the action potential also changes with the electrode position. Therefore, when the positional relationship between the microelectrode and the nerve cell changes, the feature amount such as the amplitude and duration of the action potential also changes, causing erroneous classification when classifying the nerve action potential.

In order to cope with this, an electrode moving device for nerve cell action potential measurement has been devised which mechanically measures the amount of movement of the microelectrode and corrects the relative position between the nerve cell and the electrode (for example, Patent Document 1). However, 2) requires a special device including a machine control part, and thus is not practical.

Japanese Patent No. 2736326 Japanese Patent No. 3131628

  Technical problems of the present invention include breathing, heartbeat, body movement, amplitude and duration of nerve cell action potential due to fluctuation of positional relationship between microelectrodes and nerve cells due to change in strain of brain nerve tissue at the time of electrode insertion. An object of the present invention is to enable a long-term stable observation of nerve cell activity by tracking a change in the characteristic amount of the target.

  The nerve cell action potential classifying apparatus of the present invention is a human or animal, a nerve electrode for observation of nerve cell activity pierced into cranial nerve or nerve bundle tissue, the action potential of a plurality of nerve cells from an electrode recording point installed at the tip thereof An amplifier for amplifying a waveform, a plurality of nerve cell action potential feature amount detecting means for detecting a feature amount from action potentials of a plurality of nerve cells from an output thereof, and a feature amount of a nerve cell action potential for each nerve based on output data thereof. Short-term neuron action potential classification means for sequentially classifying each cell, and using the results obtained from short-term before and after time among the neuron action potential classification results, the correspondence of feature values for each nerve cell Means for tracking changes in action potential over time by calculating the relationship; means for tracking changes in feature values of nerve cells over time; and using the output result to reclassify action potentials in all sections. Configured as a combination of classification means.

Further, another invention constitutes a classification program for causing a computer to execute a classification process of the above-mentioned device.

  Such a nerve cell action potential classifier tracks changes in features such as the amplitude and duration of nerve cell action potential caused by the relative position change of the nerve cell and the recording point of the microelectrode inserted in the brain nerve tissue. In addition, the classification of the nerve cell action potential can be performed with high accuracy. As a result, the characteristics such as the amplitude and duration of the nerve cell action potential caused by the change in the positional relationship between the microelectrode and the nerve cell due to the change in the amount of strain of the brain tissue during respiration, heartbeat, body movement, and electrode insertion It is possible to obtain a nerve cell action potential classification result in which misclassification hardly occurs due to a change in the amount, and it is possible to observe a stable nerve cell activity for a long time.

As described above, according to the present invention, the effects of changes in the positional relationship between microelectrodes and nerve cells due to changes in the amount of strain in the brain tissue during respiration, heartbeat, body movement, and electrode insertion are suppressed, and nerves are suppressed. Cell activity data can be observed stably for a long time. Thus, the present invention is effective for nerve cells near the surface of the brain, nerve cells near an artery, neuroelectrophysiological experiments during behavior involving body movement, or immediately after microelectrode insertion. In addition, since a special device including a mechanical control unit as in Patent Literature 2 is not required, the present invention can be applied to measurement of nerve cell activity using an implantable electrode in a living body.

An embodiment of an apparatus and an observation program according to the present invention will be described with reference to the drawings. FIG. 1 conceptually shows a nerve cell action potential classifying apparatus of the present invention in a state where microelectrodes of the apparatus are inserted into nerve cell tissue. In the figure, reference numeral 1 denotes a microelectrode for observing nerve cell activity (hereinafter, microelectrode), and 2a to 2g denote recording points at the tips of the microelectrodes (note that recording points 2a, 2b, and 2g are hidden in the figure. 7 are observation output signals of the recording points 2a to 2g, 3 is a multichannel signal amplifier for amplifying the action potential waveform signals of ch.1 to ch.7, and 4 is a plurality of nerve cell action potential feature amount detecting means. Reference numeral 5 denotes a short section neuron action potential classifying unit, 6 denotes a neuron action potential feature change tracking unit, 7 denotes a cell action potential reclassification unit, and 1 to 7 constitute the apparatus of the present invention. 7 can also be configured by a computer, and 10a and 10b represent nerve cells.

The microelectrode 1 is an example in which seven platinum-tungsten alloy wires are sealed in quartz glass and the tips are polished to use the recording points at the tips of a seven-core electrode having a diameter of about 120 microns. The action potentials of the nerve cells 10a, 10b,... Are observed at the recording points 2a to 2g, and the multi-channel signals ch. 1 to ch. The multi-channel signal amplifier 3 amplifies the action potential waveform signal.

FIG. 2 shows a multichannel amplifier 3 observed as a potential difference between each recording point on the microelectrode 1 pierced into the somatosensory area of the cortex of the rat and a platinum wire (reference electrode) implanted under the rat head. (Ch.1 to ch.6) (however, ch.7 was omitted because it was impossible to measure because it was disconnected). The horizontal axis represents time, and the vertical axis represents potential amplitude. n is a spike number obtained by numbering the observed nerve cell action potentials (spikes) in the order of observation, and tn represents a spike firing time of the same waveform. The spike firing time is a time at which the amplitude of the spike waveform becomes maximum in the negative direction. Up to two milliseconds before and after the spike firing time (within the square frame in FIG. 2) were detected as spike waveforms.

  A plurality of nerve cell action potential feature amount detection means 4 performs input processing of a nerve cell action potential waveform sent from the multi-channel signal amplifier 3 and outputs a nerve cell having a correlation value with a template waveform of 0.7 or more. Is detected as a spike waveform generated by one firing of, and a spike number (n) is assigned. Next, an amplitude (negative spike potential) is calculated from the specified waveform as a feature amount of the action potential, and the spike amplitude (vector) and its firing time (tn) are recorded in the memory. The template waveform is obtained by averaging a plurality of spike waveforms observed in a situation where the microelectrodes and the nerve cells are sufficiently close and the signal-to-noise ratio is high.

The nerve cell action potential classifying means 5 analyzes the action potential amplitude data detected by the plurality of nerve cell action potential feature quantity detection means 4 and collects each group (hereinafter referred to as a cluster) found in the amplitude distribution of nerve cell action potentials. Are classified into action potentials. Since the amplitude of the observed nerve cell action potentials varies depending on the relative distance between the recording point of the microelectrode and the nerve cells, the action potentials belonging to the same population were generated from the same nerve cell It can be considered as an action potential.

FIG. 4 (a) shows the ch. 3, ch. 4 shows an example in which the amplitudes of action potentials observed for a predetermined measurement period are plotted in a space with a horizontal axis and a vertical axis. From the figure, the amplitude of the observed action potential is classified into four clusters I, II, III, and IV, and it can be seen that four nerve cells are firing during this measurement period. For example, cluster I corresponds to the action potential amplitude of nerve cell 10a. Since these four nerve cells have different distances from the recording points 2c and 2d, the distribution of the action potential amplitude values is also different. In other words, the position of the firing neuron can be specified from the distribution of the nerve cell action potential amplitude value (however, it is necessary to correct the position fluctuation of the microelectrode).

The action potentials observed at the six recording points of the microelectrode in FIG. 1 can be represented by the points of the six-axis space vector. The short section neuron action potential classifying means 5 classifies action potentials for each cluster found in the amplitude distribution of the action potentials, and gives the same cluster number to action potentials included in each cluster.

Specifically, the distribution of the action potential amplitude vector included in each cluster is approximated by a Gaussian distribution, and the Mahalanobis distance (statistical intercluster distance) between all clusters is the standard deviation. 2. Classification is performed by the hierarchical class cling method in which clusters are connected repeatedly until the number of clusters becomes 5 times or more.

In the initial state of this procedure, each action potential is regarded as one cluster, a unique cluster number is given throughout the entire data section, and the clusters are connected to each other until the number of action potentials included in the cluster becomes larger than the number of recording points. Indicates the variance of the noise amplitude of the waveform observed between each recording point and the reference electrode as the variance of the cluster. The clusters are connected by setting a smaller value among the cluster numbers of the two clusters to be connected as the cluster number of each action potential included in the two clusters.

The nerve cell action potential characteristic amount change tracking means 6 is a means for classifying nerve cells between short-term action potential classification results obtained from two short-term data obtained by the short-term neuron action potential classification means 5 in time. Calculate the correspondence.

With the insertion of the electrode, the positional relationship between the electrode and the nerve cell changes as shown in FIG. 5A, and the action potential amplitude (cluster center) measured at two recording points is shown in FIG. ), The change in the action potential amplitude observed at each recording point is small if the electrode movement amount is sufficiently small, so that the change in the nerve cell action potential amplitude vector obtained by vectorizing the action potential amplitudes is also small. small.

Therefore, if the change in the position of the electrode and the nerve cell is sufficiently small, the amplitude vector of the action potential generated from the same nerve cell is almost the same, and the Mahalanobis distance of the action potential amplitude vector distribution between the clusters before and after the electrode movement is calculated. The smallest one can be regarded as an action potential generated from the same nerve cell, and can be tracked in response to a minute change in the nerve cell action potential amplitude as shown in FIG. FIG. 6A shows a cluster (center) locus of the action potential amplitudes (ch. 3 and ch. 4) calculated from the fluctuation of the action potential amplitude in FIG.

In the embodiment, the action potential amplitude vector of each cluster obtained from the short section action potential classification result (the cluster result in the short section) is approximated by a Gaussian distribution, and those having the minimum Mahalanobis distance between the clusters before and after the electrode movement are compared. Was considered as an action potential originating from the same nerve cell.

FIG. 7 is a diagram for explaining the tracking of the action potential amplitude by the short section action potential classification. As shown in FIG. 7 (a), the clustering sections T1, T2,... Of the short section neuron action potential classifying means 5 overlap each other before and after time (short sections T1 and T2, T2 and T3,...). To In the figure, A, B, C,... Represent action potential amplitude classification data of one cell, which should be tracked so as to be linked to one.

First, by performing short-section neuron action potential classification in the T1 section, the Mahalanobis distance between the clusters A and B is less than 2.5 times the standard deviation, so that they are classified into the same cluster. Next, the clusters B and C are classified into the same cluster in the same manner as when the short-range nerve cell action potential classification of the T2 section is performed. By repeating the above processing up to the section T5, A and B, B and C, C and D,.

By the way, according to only the above description, the section length of the short section nerve cell action potential classifying means 5 may be such that the Mahalanobis distance between clusters A and B is less than 2.5 times the standard deviation. However, it must be distinguished from clusters formed by action potentials of other nerve cells.

That is, assuming a cluster B 'formed by the action potential of another nerve cell in the short section T2, the Mahalanobis distance between A and B must be shorter than the Mahalanobis distance between A and B'. When the distance between B and B 'is 2.5, and the distance between A and B is d, the Mahalanobis distance between A and B' is the closest and is 2.5-d. Therefore, in order to satisfy 2.5-d> d, it is necessary to determine the section length so that d is less than 1.25. In the data used in the embodiment, the section length was set to 65 seconds, so that d <0.25 was satisfied for any cluster.

The neuron action potential feature value change tracking means 6 detects the correspondence of the temporal change of the amplitude value of the nerve cell action potential based on the short-term amplitude data output from the short-term nerve cell action potential classification means 5. That is, the cluster B belongs to the short sections T1 and T2, and is connected to the cluster A in the short section nerve cell action potential classification in the short section T1, and connected to the cluster C in the section T2. , C are determined to be due to action potentials generated from the same nerve cell. When this is repeatedly applied to the preceding and following sections (short sections T2 and T3, T3 and T4,...), A series of clusters A, B, C,. Thus, the cluster chain 1 (A1, B1, C1,...), The cluster chain 2 (A2, B2, B3,...),. . . Get. The nerve cell action potential feature amount change tracking means 6 outputs information about the cluster chain (cluster chain 1, cluster chain 2,...) To the nerve cell action potential re-classification means 7 with a spike number.

The nerve cell action potential reclassification means 7 reassigns a cluster number common to all data sections to each action potential on the basis of the information on the chain of the clusters by the nerve cell action potential feature amount change tracking means 6. Specifically, the chain information of the clusters obtained by the neuron action potential feature amount change tracking means 6, that is, cluster chain 1 (A1, B1, C1,...), Cluster chain 2 (A2, B2, B3, ...),. . . Is used to reassign the cluster number of each action potential included in the chain of each cluster to the smallest of these cluster numbers.

As a result, in the output of the short section neuron action potential classifying means 5, a different cluster number is given for each short section even for the same action potential, but a cluster number unified in all data sections is used for each cluster chain. Given.
Further, the nerve cell action potential reclassification means 7 can capture the firing time (tn) of the action potential of the multiple nerve cell action potential feature amount detection means 4 and record and display the action potential in time series.

In the embodiment, a data number is given as an initial value of a cluster number, and when two clusters are combined by the short-range nerve cell action potential classifying means 5, the smallest of the data numbers of the action potentials included in the cluster to be combined The value is re-assigned as the cluster number after merging. As a result, the classification result of the nerve cell action potential in a state where the cluster number is already unified in all data sections and the change of the nerve cell action potential amplitude is also obtained, is obtained. The result repeatedly used by the classification means 5 over the entire short section can be directly used as the re-classification result of the nerve cell action potential. As described above, even if the relative positional relationship between the recording point and the nerve cell changes, the nerve cell can be specified.

In order to investigate the effect of the nerve cell action potential classifier according to the present invention, the number of nerve cells observed with and without the action potential feature tracking function, which is the core of the present invention, decreases with time. Compare how they do it. The action potential feature tracking function may be enabled by performing the processing described above. To disable the action potential feature tracking function, the short-term neuronal activity may be disabled as shown in FIG. This is realized by setting the length of the clustering processing section in the potential classifying means 5 to the same length as the total time length (observation time) in which the data was obtained.

FIG. 8 shows the processing results of the action potential classifier for action potential waveform data of a plurality of long-term neurons obtained by inserting the microelectrode 1 into the monkey lower temporal cortex under anesthesia. is there. According to this, it can be understood that a decrease in the number of observable nerve cells can be suppressed by activating the action potential feature amount tracking function. Therefore, the use of the nerve cell action potential classifying device according to the present invention makes it possible to observe a plurality of nerve cell action potentials for a long time.

In the embodiment, the amplitude information is used as the feature value of the action potential. However, the duration can be used as the feature value. This is because a change according to the electrode position as shown in FIG. 5C can be observed even during the duration, a cluster as shown in FIG. 4B is formed, and the center of the cluster gradually changes as shown in FIG. 6B. Because they do. Further, any amount that gradually changes in accordance with the electrode position, such as the amplitude, the duration, and the amount obtained by combining these, can be used as the feature amount.

Furthermore, a classification program that causes a computer to execute the classification process of the device of the present invention can be created to realize the device of the present invention. Means for inputting an observed nerve cell action potential waveform signal to a computer for classifying the nerve cell action potential, means for obtaining feature data of nerve cell activity from the input signal of the observed nerve cell action potential waveform, Means for classifying the amount of data for each short term for each neuron and outputting a classification signal, determining the relationship between the classified and output signals calculated in the previous short period and the classified and output signals, A means for outputting a signal for tracking a change in an action potential feature signal belonging to the same classification; and a method for outputting all of the nerve cell action potentials from the plurality of nerve cell action potential classification signals and the signal for tracking the change in the nerve cell action potential feature. Provided is a classification program for classifying a nerve cell action potential for functioning as a means for reclassification over a section.

In addition, the above classification program can be stored in a recording medium so that the computer can read it, and the recording medium can be input to the computer. Furthermore, it is also possible to transmit the classification program to the user's computer on a transmission carrier.

1 shows the overall configuration of the present invention. 1 shows an example of a multi-channel signal of a nerve cell action potential observed by a microelectrode. 4 shows an example of a feature amount of a nerve cell action potential. 4 shows a classification example based on the distribution of the amplitude and duration of a nerve cell action potential. FIG. 9 is a diagram showing changes in the amplitude and duration of the action potential due to changes in the relative distance between the microelectrode and the nerve cell. 4 shows an example of the result of tracking the center of a cluster formed by the distribution of the amplitude and duration of a nerve cell action potential. FIG. 3 is a diagram illustrating an action potential feature amount tracking method based on short section action potential classification according to the present invention. 5 shows an example of a classification result of a nerve cell action potential according to the present invention.

Explanation of reference numerals

DESCRIPTION OF SYMBOLS 1 Microelectrode 2a-2g for neuron action potential observation Microelectrode recording points 10a, 10b ... Nerve cell 3 Multi-channel signal amplifier 4 Plural neuron action potential feature amount detection means 5 Short section neuron action potential classification means 6 Nerve Cell action potential feature amount change tracking means 7 Neuron action potential reclassification means T, T1 to T5 Clustering section

Claims (2)

A microelectrode for observing nerve cell activity,
A plurality of nerve cell action potential feature amount detecting means for inputting an action potential waveform signal of a nerve cell observed by the recording point of the microelectrode and extracting feature amount data of the action potential;
Short-section neuron action potential classification means for classifying and outputting the feature amount data for each neuron for a short period,
Nerve cells that determine the correspondence between the signals output as classified as described above and the signals output as classified in the previous short period, and track changes in feature amounts of action potentials generated from the same neuron Action potential feature change tracking means;
A neural cell action potential reclassifying means for reclassifying a neural cell action potential from an output signal of a plurality of nerve cell action potential feature amount detecting means and a nerve cell action potential feature amount change tracking means; Classifier. A computer to classify nerve cell action potentials,
Means for inputting the observed nerve cell action potential waveform signal,
Means for obtaining feature data of nerve cell activity from the input signal of the observed nerve cell action potential waveform,
Means for classifying the feature data for a short period for each nerve cell and outputting a classification signal;
Classify the signal output as described above in the previous short period to determine the relationship with the output signal, and output a signal that tracks the change in the feature value of the action potential generated from the same nerve cell Means for classifying a nerve cell action potential for functioning as a means for reclassifying a nerve cell action potential from the plurality of nerve cell action potential feature amount detection signals and a signal obtained by tracking a change in the nerve cell action potential feature amount. program.

Images