Methods and devices for analysis of clustered data, in particular action potentials (i.e. neuron firing signals in the brain)

• US 8,150,795 B2
• Filed: 12/22/2005
• Issued: 04/03/2012
• Est. Priority Date: 12/22/2004
• Status: Active Grant

• Current Assignee: Wisconsin Alumni Research Foundation (University of Wisconsin System)
• Original Assignee: Wisconsin Alumni Research Foundation (University of Wisconsin System)
• Inventors: Montgomery, Erwin B. Jr., Huang, He, Assadi, Amir H.
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): Gonzales, Vincent
• Application Number:
  US11/722,229
• Publication Number:
  US 20090043220A1
• Time in Patent Office: 2,294 Days
• Field of Search: 706/62
• US Class Current: 706/62
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 2218/22 Source localisation; Inverse modelling

Abstract

Methods for clustering of multi-dimensional data allow unsupervised grouping of multi-dimensional data points into clusters having like characteristics. The methods may be usefully applied to extracellular action potentials (neuronal spikes) measured from the brain, whereby spike data may be grouped in accordance with dimensions such as spike period, spike shape, etc., to assist in identification and location of individual neurons and/or regions of the brain.

Claims

1. A method of identifying regions within the brain, the method comprising the steps of:

a. providing a collection of N-dimensional data points, each data point being representative of a location in the brain and containing N>

1 different types of data therefrom;

b. sampling N different types of data from a location in the brain, thereby defining a sampled N-dimensional data point;

c. automatically defining within a processor M discrete data point clusters (M>

1) from the collected data points and from the sampled data point, wherein;

i. each data point cluster contains data points which are proximate in N-dimensional space, and

ii. the M data point clusters correspond to M discrete regions within the grey matter of the brain,

and further wherein the M discrete data point clusters are defined by the following steps;

(1) within a data set defined by the collected data points and the sampled data point, identifying the data point which has the greatest proportion of closely proximate data points in N-space, thereby identifying an index data point;

(2) defining a data cluster nucleus which contains the index data point and closely proximate data points;

(3) defining in N-space the centroid of the data points of the data cluster nucleus, thereby defining a nucleus centroid; and

(4) expanding the data cluster nucleus to include data points which are outside of, but closely proximate in N-space to, the data cluster nucleus, wherein the expanded data cluster nucleus defines one of the M data point clusters;

d. removing from the data set the data points corresponding to the defined data point cluster;

e. defining a subsequent data point cluster by use of the foregoing step c.; and

f. indicating whether the sampled data point is within a particular data point cluster, and therefore within a particular one of the M regions of the brain.

2. The method of claim 1 wherein N>

4.

3. The method of claim 1 wherein M is neither:

a. predefined within the processor, nor

b. entered by a user.

4. The method of claim 1 wherein the N different types of data sampled from a location in the brain are sampled by a probe, the probe including two or more electrodes thereon, the electrodes being capable of capturing separate data points.

5. The method of claim 1 wherein the data sampled at the different locations of the brain is electrical signal data.

6. The method of claim 5 wherein the sampled electrical signal data includes at least two of:

a. spike time;

b. time between spikes;

c. spike frequency;

d. spike amplitude;

e. slope of spike rise; and

f. slope of spike fall.

7. A method of identifying regions within the brain, the method comprising the steps of:

a. providing a collection of N-dimensional data points, each data point being representative of a location in the brain and containing N>

1 different types of data therefrom;

b. sampling N different types of data from a location in the brain, thereby defining a sampled N-dimensional data point;

c. automatically defining within a processor M discrete data point clusters (M>

1) from the collected data points and from the sampled data point, wherein;

I. each data point cluster contains data points which are proximate in N-dimensional space, and

II. the M data point clusters correspond to M discrete regions within the grey matter of the brain,

and further wherein the M discrete data point clusters are defined by the following steps;

i. within a data set defined by the collected data points and the sampled data point, identifying the data point which has the greatest proportion of closely proximate data points in N-space, thereby identifying an index data point;

ii. defining a data cluster nucleus which contains the index data point and closely proximate data points;

iii. defining in N-space the centroid of the data points of the data cluster nucleus, thereby defining a nucleus centroid; and

iv. expanding the data cluster nucleus to include data points which are outside of, but closely proximate in N-space to, the data cluster nucleus, wherein the step of expanding the data cluster nucleus includes;

(1) identifying the data point in the data set which is;

(a) closest in N-space to the data cluster nucleus, but

(b) outside of the data cluster nucleus,

thereby defining a candidate data point;

(2) if the distance in N-space between the nucleus centroid and the candidate data point falls within a predetermined distance;

(a) expanding the data cluster nucleus to include the candidate data point, and

(b) returning to the foregoing step iii. to redefine the nucleus centroid,

wherein the expanded data cluster nucleus defines one of the M data point clusters; and

d. indicating whether the sampled data point is within a particular data point cluster, and therefore within a particular one of the M regions of the brain.

8. The method of claim 7 wherein N>

4.

9. The method of claim 7 wherein M is neither:

a. predefined within the processor, nor

b. entered by a user.

10. The method of claim 7 wherein the N different types of data sampled from a location in the brain are sampled by a probe, the probe including two or more electrodes thereon, the electrodes being capable of capturing separate data points.

11. The method of claim 7 wherein the data sampled at the different locations of the brain is electrical signal data.

12. The method of claim 11 wherein the sampled electrical signal data includes at least two of:

a. spike time;

b. time between spikes;

c. spike frequency;

d. spike amplitude;

e. slope of spike rise; and

f. slope of spike fall.

13. A method of identifying regions within the grey matter of the brain, the method comprising the steps of:

a. sampling data at different locations within the grey matter, wherein N different types of data are sampled at each location (N>

1), thereby generating a data set containing N-dimensional data points sampled at the locations;

b. resolving within a processor M data point clusters (M>

1) from the data points, wherein each data point cluster includes data points which are proximate in N-dimensional space, and wherein the step of resolving M data point clusters includes;

(1) identifying within the data set the data point which has the greatest proportion of closely proximate data points in N-space, thereby identifying an index data point;

(2) defining a data cluster nucleus which contains the index data point and closely proximate data points;

(3) defining in N-space the centroid of the data points of the data cluster nucleus, thereby defining a nucleus centroid; and

(4) expanding the data cluster nucleus to include data points which are outside of, but closely proximate in N-space to, the data cluster nucleus, wherein the step of expanding the data cluster nucleus includes;

(a) identifying the data point in the data set which is;

i. closest in N-space to the data cluster nucleus, but

ii. outside of the data cluster nucleus,

thereby defining a candidate data point;

(b) if the distance in N-space between the nucleus centroid and the candidate data point falls within a predetermined distance;

i. expanding the data cluster nucleus to include the candidate data point, and

ii. returning to the foregoing step (3) to redefine the nucleus centroid,

wherein the expanded data cluster nucleus defines one of the M data point clusters;

c. defining M regions of the brain, each region including the locations corresponding to the data points of the data cluster.

14. The method of claim 13 wherein N>

4.

15. The method of claim 13 wherein M is unknown to the processor prior to the step of resolving the M data point clusters.

16. The method of claim 13 wherein the M data point clusters resolved from the data points have boundaries which:

a. are defined by their outermost data points in N-dimensional space; and

b. do not overlap.

17. The method of claim 13 wherein the data sampled at the different locations of the brain is electrical signal data.

18. The method of claim 17 wherein the sampled electrical signal data includes at least two of:

a. spike time;

b. time between spikes;

c. spike frequency;

d. spike amplitude;

e. slope of spike rise; and

f. slope of spike fall.

19. A method of identifying regions within the grey matter of the brain, the method comprising the steps of:

a. sampling data at different locations within the grey matter, wherein N different types of data are sampled at each location (N>

1), thereby generating a data set containing N-dimensional data points sampled at the locations;

b. defining within a processor a data point cluster from the data points, wherein the data point cluster includes data points which are proximate in N-dimensional space, the step of defining the data point cluster including;

(1) identifying within the data set the data point which has the greatest proportion of closely proximate data points in N-space, thereby identifying an index data point;

(2) defining a data cluster nucleus which contains the index data point and closely proximate data points;

(3) defining in N-space the centroid of the data points of the data cluster nucleus, thereby defining a nucleus centroid; and

(4) expanding the data cluster nucleus to include data points which are outside of, but closely proximate in N-space to, the data cluster nucleus, wherein the expanded data cluster nucleus defines one of the M data point clusters,

c. removing from the data set the data points corresponding to the defined data point cluster; and

d. defining a subsequent data point cluster by use of the foregoing step b.,

each data point cluster defining a region of the brain, with each region including the locations corresponding to the data points of the data cluster.

20. The method of claim 19 wherein N>

4.

21. The method of claim 19 wherein M is unknown to the processor prior to the step of resolving the M data point clusters.

22. The method of claim 19 wherein the M data point clusters defined from the data points have boundaries which:

a. are defined by their outermost data points in N-dimensional space; and

b. do not overlap.

23. The method of claim 19 wherein the data sampled at the different locations of the brain is electrical signal data.

24. The method of claim 23 wherein the sampled electrical signal data includes at least two of:

a. spike time;

b. time between spikes;

c. spike frequency;

d. spike amplitude;

e. slope of spike rise; and

f. slope of spike fall.

25. A device for identifying regions within the grey matter of the brain, the device comprising:

A. a probe insertable within a brain, the probe having an electrode thereon which captures data points from the brain, each data point having N dimensions (N>

1);

B. a processor in communication with the electrode, wherein the processor;

I. receives the measured N-dimensional data points, and

II. resolves M data point clusters (M>

1) from the data points, each data cluster including data points which are proximate in N-dimensional space,

thereby identifying M regions of the brain, each region corresponding to one of the data clusters,

wherein the processor is configured to determine the M data point clusters by use of the following steps;

a. within a data set defined by the data points, identifying the data point which has the greatest proportion of closely proximate data points in N-space, thereby identifying an index data point;

b. defining a data cluster nucleus which contains the index data point and closely proximate data points;

c. defining in N-space the centroid of the data points of the data cluster nucleus, thereby defining a nucleus centroid; and

d. expanding the data cluster nucleus to include data points which are outside of, but closely proximate in N-space to, the data cluster nucleus, by;

(1) identifying the data point in the data set which is;

(a) closest in N-space to the data cluster nucleus, but

(b) outside of the data cluster nucleus, thereby defining a candidate data point;

(2) if the distance in N-space between the nucleus centroid and the candidate data point falls within a predetermined distance;

(a) expanding the data cluster nucleus to include the candidate data point, and

(b) returning to the foregoing step c. to redefine the nucleus centroid,

wherein the expanded data cluster nucleus defines one of the M data point clusters.

26. The device of claim 25 wherein N>

4.

27. The device of claim 25 wherein M is neither:

a. predefined within the processor, nor

b. entered by a user.