Method, system, and computer-accessible medium for classification of at least one ICTAL state

• US 9,443,141 B2
• Filed: 06/02/2009
• Issued: 09/13/2016
• Est. Priority Date: 06/02/2008
• Status: Active Grant

• Current Assignee: New York University
• Original Assignee: New York University
• Inventors: Mirowski, Piotr W., Madhavan, Deepak, Lecun, Yann, Kuzniecky, Ruben
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Wu, Fuming
• Application Number:
  US12/995,925
• Publication Number:
  US 20110218950A1
• Time in Patent Office: 2,660 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: A61B 5/245 specially adapted for magnetoencephalographic [MEG] signals
• CPC Class Codes: A61B 5/318 Heart-related electrical modalities, e.g. electrocardiography [ECG]
• CPC Class Codes: A61B 5/369 Electroencephalography [EEG] devices for psychotechnics A61B5/16
• CPC Class Codes: A61B 5/372 Analysis of electroencephalograms
• CPC Class Codes: A61B 5/4094 Diagnosing or monitoring seizure diseases, e.g. epilepsy
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: A61B 5/7275 Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• CPC Class Codes: G06F 2218/12 Classification; Matching
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 3/126 Evolutionary algorithms, e.g. genetic algorithms or genetic programming
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

An exemplary methodology, procedure, system, method and computer-accessible medium can be provided for receiving physiological data for the subject, extracting one or more patterns of features from the physiological data, and classifying the at least one state of the subject using a spatial structure and a temporal structure of the one or more patterns of features, wherein at least one of the at least one state is an ictal state.

Claims

1. A method for classifying at least one state of a subject, comprising:

receiving physiological data for the subject, the physiological data including a plurality of features having at least a first feature and a second feature that are based on a respective measure of a synchronization for at least one pair of channels, wherein the synchronization associated with (i) the first feature is over a first time period; and

(ii) the second feature is over a second time period;

generating an array of the features having the first and second features as entries in the array;

generating a multilayer convolutional neural network, wherein a first layer and a second layer of the multilayer convolutional neural network is based on a convolution across time, and a third layer of the multilayer convolutional neural network is based on a convolution across time, space and frequency; and

using a computer hardware arrangement, classifying the at least one state based on information provided in the array using the at least one multilayer convolutional neural network.

2. The method of claim 1, wherein the features are spatially-varying and time-varying.

3. The method of claim 1, wherein a temporal structure of the features comprises a local time structure.

4. The method of claim 1, wherein the features comprise bivariate features.

5. The method of claim 1, wherein the classifying procedure is performed using at least one of a logistic regression or a support vector machine.

6. The method of claim 1, further comprising:

electronically storing the received physiological data;

receiving further physiological data for the subject; and

using the computing arrangement, classifying the at least one state of the subject based on the further physiological data.

7. The method of claim 4, wherein the bivariate features are based on the measure of the synchronization which is based on wavelet analysis-based synchrony values grouped in one or more frequency bands.

8. The method of claim 1, wherein the first feature is based on a first frequency and the second feature is based on at least one second frequency.

9. The method of claim 1, wherein at least one further layer of the multilayer convolutional neural network is based on subsampling.

10. The method of claim 1, wherein the channels are EEG channels.

11. The method of claim 1, wherein the plurality of features includes a third feature and a fourth feature that are based on a respective measure of the synchronization for at least one second pair of channels, wherein the synchronization associated with (i) the third feature is over the first time period;

and (ii) the fourth feature is over the second time period.

12. The method of claim 1, wherein the plurality of features includes a third feature and a fourth feature that are based on a respective measure of the synchronization for at least one second pair of channels, wherein the synchronization associated with (i) the third feature is over the first time period;

and (ii) the fourth feature is over the second time period.

13. A non-transitory computer-accessible medium for classifying at least one state of a subject, the computer-accessible medium including instructions thereon, wherein, when a computing arrangement executes the instructions, the computing arrangement is configured to perform procedures comprising:

receiving physiological data for the, the physiological data including a plurality of features having at least a first feature and a second feature that are based on a respective measure of a synchronization for at least one pair of channels, wherein the synchronization associated with (i) the first feature is over a first time period; and

(ii) the second feature is over a second time period;

generating an array of the features having the first and second features as entries in the array;

generating a multilayer convolutional neural network, wherein a first layer and a second layer of the multilayer convolutional neural network is based on a convolution across time, and a third layer of the multilayer convolutional neural network is based on a convolution across time, space and frequency; and

classifying the at least one state based on the array using the at least one multilayer convolutional neural network.

14. The computer-accessible medium of claim 13, wherein the features are represented by two-dimensional data.

15. The computer-accessible medium of claim 13, wherein the features are represented by three-dimensional data.

16. The computer-accessible medium of claim 13, wherein the at least one state comprises at least one of an ictal classification, a peri-ictal classification, a pre-ictal classification or an interictal classification.

17. The computer-accessible medium of claim 13, wherein the features comprise bivariate features.

18. The computer-accessible medium of claim 13, wherein the computer arrangement is further configured to maintain a memory of previously received physiological data after classifying the ictal state.

19. The computer-accessible medium of claim 13, wherein the first feature is based on a first frequency, and the second further feature is based on at least one second frequency.

20. The computer-accessible medium of claim 13, wherein at least one further layer of the multilayer convolutional neural network is based on subsampling.

21. The computer-accessible medium of claim 13, wherein the channels are EEG channels.

22. The computer-accessible medium of claim 13, wherein the computing arrangement is further configured to generate (i) a fourth layer of the multilayer convolutional neural network is generated based on subsampling, and (ii) a fifth layer of the multilayer convolutional neural network based on the convolution across time.

23. A system for classifying at least one state of a subject, which comprises:

a computer hardware arrangement configured to;

receive physiological data for the, the physiological data including a plurality of features having at least a first feature and a second feature that are based on a respective measure of a synchronization for at least one pair of channels, wherein the synchronization associated with (i) the first feature is over a first time period; and

(ii) the second feature is over a second time period;

generate an array of the features having the first and second features as entries in the array;

generate a multilayer convolutional neural network, wherein a first layer and a second layer of the multilayer convolutional neural network is based on a convolution across time, and a third layer of the multilayer convolutional neural network is based on a convolution across time, space and frequency; and

classify the at least one state based on the array using a trained classifier module that uses the at least one multilayer convolutional neural network.

24. The system of claim 23, wherein the features are represented by three-dimensional data.

25. The system of claim 24, wherein the three-dimensional data is data of time versus a particular pair of channels versus frequency.

26. The system of claim 23, wherein the at least one state comprises at least one of an ictal classification, a peri-ictal classification, a pre-ictal classification or an interictal classification.

27. The system of claim 23, wherein the physiological data is at least one of EEG data, multi-channel EEG data, fMR1 data, MEG data, EKG data, pulse data, respiration data, temperature data, eye movement data or blood chemistry data.

28. The system of claim 27, wherein the features are spatially-varying and time-varying.

29. The system of claim 28, wherein the features are frequency-varying.

30. The system of claim 23, wherein the classification procedure discriminates between pre-ictal and inter-ictal states.

31. The system of claim 23, wherein a temporal structure of the features comprises a local time structure.

32. The system of claim 23, wherein the features comprise bivariate features.

33. The system of claim 32, wherein the bivariate features are based on the measure of the synchronization for each of the pairs of channels.

34. The system of claim 33, wherein the measure of the synchronization is based on at least one of a cross-correlation, a nonlinear interdependence, a difference of Lyapunov exponents or a phase-locking synchrony.

35. The system of claim 34, wherein the synchrony comprises one of a phase-locking statistic, an entropy of a phase difference and distribution, or a coherence.

36. The system of claim 23, wherein the classification procedure uses at least one of a logistic regression or a support vector machine.

37. The system of claim 23, wherein the classification procedure is causes a regularization of the array using a lasso regularization or an L1-norm regularization.

38. The system of claim 23, wherein the classification procedure is causes a regularization of the array using a ridge regularization or an L2-norm regularization.

39. The system of claim 23, wherein the classification procedure is trained using a stochastic gradient descent.

40. The system of claim 23, wherein the classification procedure is trained by determining finite impulse response filters of at least a portion of the features.

41. The system of claim 40, wherein the features are consecutive features.

42. The system of claim 23, wherein the computer hardware arrangement is further configured to perform a sensitivity analysis on the features before classifying the at least one state.

43. The system of claim 42, wherein the sensitivity analysis is used to evaluate an importance of a particular feature according to a spatial location, time or a frequency location in the features thereof.

44. The system of claim 38, wherein the classification procedure uses a support vector machine and the features that are in a training dataset and that serve as support vectors for the classification procedure are used to evaluate an electrophysiological recording or the development of ictal states over time.

45. The system of claim 33, wherein the measure of the synchronization is based on wavelet analysis-based synchrony values grouped in one or more frequency bands.

46. The system of claim 23, wherein the first feature is based on a first frequency, and the second feature is based on at least one further frequency.

47. The system of claim 23, wherein at least one further layer of the multilayer convolutional neural network is based on subsampling.

48. The system of claim 23, wherein the channels are EEG channels.

49. The system of claim 23, wherein the plurality of features includes a third feature and a fourth feature that are based on a respective measure of the synchronization for at least one second pair of channels, wherein the synchronization associated with (i) the third feature is over the first time period;

and (ii) the fourth feature is over the second time period.