Systems and methods for modeling and processing functional magnetic resonance image data using full-brain vector auto-regressive model
Systems and methods for modeling functional magnetic resonance image datasets using a multivariate auto-regressive model which captures temporal dynamics in the data, and creates a reduced representation of the dataset representative of functional connectivity of voxels with respect to brain activity. Raw spatio-temporal data is processed using a multivariate auto-regressive model, wherein coefficients in the model with high weights are retained as indices that best describe the full spatio-temporal data. When there are a relatively small number of temporal samples of the data, sparse regression techniques are used to build the model. The model coefficients are used to perform data processing functions such as indexing, prediction, and classification.
1. A method, comprising:
obtaining a raw spatio-temporal dataset acquired from scanning a brain of a subject performing a given task;
constructing a full spatio-temporal model using the raw spatio-temporal dataset, wherein the full spatio-temporal model represents brain activity that occurs in all regions of the subject's brain in response to the subject performing the given task;
selecting model parameters from the full spatio-temporal model which meet or exceed a threshold parameter that defines a level of causal relation between voxels in the obtained dataset, wherein the threshold parameter is based on a prediction power threshold value that specifies a threshold for which brain activity in a given voxel predicts brain activity in another voxel;
generating a reduced model representation of the full spatio-temporal model using the selected model parameters;
storing the reduced model of the full spatio-temporal model in a data storage system; and
utilizing the reduced model of the full spatio-temporal model to predict future brain activity of the subject by performing a process comprising:
accessing the reduced model of the full spatio-temporal model from the data storage system;
obtaining scan data representing a current estimate of brain activity of the subject performing the given task at an initial time T 0 ;
utilizing the obtained scan data to initialize the reduced model of the spatio-temporal model with the current estimate of the brain activity at the initial time T 0 ; and
utilizing model parameters of the initialized reduced model of the spatio-temporal model to predict future brain activity of the subject at one or more times subsequent to the initial time To without having to utilize scan data of current activity of the subject for the one or more times subsequent to the initial time T 0 ;
wherein the method is performed by a computer executing program code.
2. The method of claim 1 , further comprising providing biofeedback to the subject based on the predicted future brain activity to provide an indication of said predicted future brain activity to said subject.
3. The method of claim 2 , wherein the biofeedback comprises sound or light feedback.
4. The method of claim 1 , wherein the raw spatio-temporal data comprises functional magnetic resonance imaging data.
5. The method of claim 1 , wherein constructing a full spatio-temporal model comprises using a multivariate auto-regressive model to generate a model of the spatio-temporal dataset.
6. The method of claim 1 , wherein selecting model parameters from the full spatio-temporal model which meet or exceed a threshold parameter that defines a level of causal relation between voxels in the acquired dataset, comprises:
obtaining a value of a causal link between a pair of voxels;
comparing the value of the causal link to the threshold parameter;
selecting all causal links between pairs of voxels that meet or exceed the threshold parameter; and
eliminating all causal links between pairs of voxels that do not meet or exceed the threshold parameter.
7. The method of claim 1 , wherein generating a reduced model representation of the full spatio-temporal model using the selected model parameters comprises generating a three-dimensional directed node graph having nodes and directed edges connecting the nodes, wherein nodes in the directed node graph represent voxels and wherein a directed edge in the directed node graph represents a causal relation between two nodes connected by the directed edge.
8. An apparatus comprising:
a memory to store program instructions; and
a processor coupled to the memory and operative to process the stored program instructions to:
obtain a raw spatio-temporal dataset acquired from scanning a brain of a subject performing a given task;
construct a full spatio-temporal model using the raw spatio-temporal dataset, wherein the full spatio-temporal model represents brain activity that occurs in all regions of the subject's brain in response to the subject performing the given task;
select model parameters from the full spatio-temporal model which meet or exceed a threshold parameter that defines a level of causal relation between voxels in the obtained dataset, wherein the threshold parameter is based on a prediction power threshold value that specifies a threshold for which brain activity in a given voxel predicts brain activity in another voxel;
generate a reduced model representation of the full spatio-temporal model using the selected model parameters;
store the reduced model of the full spatio-temporal model in a data storage system; and
utilize the reduced model of the full spatio-temporal model to predict future brain activity of the subject by performing a process comprising:
accessing the reduced model of the full spatio-temporal model from the data storage system;
obtaining scan data representing a current estimate of brain activity of the subject performing the given task at an initial time T 0 ;
utilizing the obtained scan data to initialize the reduced model of the spatio-temporal model with the current estimate of the brain activity at the initial time T 0 ; and
utilizing model parameters of the initialized reduced model of the spatio-temporal model to predict future brain activity of the subject at one or more times subsequent to the initial time T 0 without having to utilize scan data of current activity of the subject for the one or more times subsequent to the initial time T 0 .
9. The apparatus of claim 8 , wherein the processor is further operative to process the stored program instructions to provide biofeedback to the subject based on the predicted future brain activity to provide an indication of said predicted future brain activity to said subject.
10. The apparatus of claim 9 , wherein the biofeedback comprises sound or light feedback.
11. The method of claim 8 , wherein the raw spatio-temporal data comprises functional magnetic resonance imaging data.
12. The method of claim 8 , wherein the full spatio-temporal model is constructed using a multivariate auto-regressive model to generate a model of the spatio-temporal dataset.
13. The method of claim 8 , wherein the processor is operative to process the stored program instructions to select model parameters from the full spatio-temporal model by:
obtaining a value of a causal link between a pair of voxels;
comparing the value of the causal link to the threshold parameter;
selecting all causal links between pairs of voxels that meet or exceed the threshold parameter; and
eliminating all causal links between pairs of voxels that do not meet or exceed the threshold parameter.
14. The method of claim 8 , wherein the processor is operative to process the stored program instructions to generate the reduced model representation of the full spatio-temporal model using the selected model parameters by generating a three-dimensional directed node graph having nodes and directed edges connecting the nodes, wherein nodes in the directed node graph represent voxels and wherein a directed edge in the directed node graph represents a causal relation between two nodes connected by the directed edge.
15. A computer program product comprising a non-transitory computer readable storage medium having computer readable program code stored thereon, wherein the computer readable program code is executable by a computer to perform a method for predicting future brain activity of a subject, the method comprising:
obtaining a raw spatio-temporal dataset acquired from scanning a brain of a subject performing a given task;
constructing a full spatio-temporal model using the raw spatio-temporal dataset, wherein the full spatio-temporal model represents brain activity that occurs in all regions of the subject's brain in response to the subject performing the given task;
selecting model parameters from the full spatio-temporal model which meet or exceed a threshold parameter that defines a level of causal relation between voxels in the obtained dataset, wherein the threshold parameter is based on a prediction power threshold value that specifies a threshold for which brain activity in a given voxel predicts brain activity in another voxel;
generating a reduced model representation of the full spatio-temporal model using the selected model parameters;
storing the reduced model of the full spatio-temporal model in a data storage system; and
utilizing the reduced model of the full spatio-temporal model to predict future brain activity of the subject by performing a process comprising:
accessing the reduced model of the full spatio-temporal model from the data storage system;
obtaining scan data representing a current estimate of brain activity of the subject performing the given task at an initial time T 0 ;
utilizing the obtained scan data to initialize the reduced model of the spatio-temporal model with the current estimate of the brain activity at the initial time T 0 ; and
utilizing model parameters of the initialized reduced model of the spatio-temporal model to predict future brain activity of the subject at one or more times subsequent to the initial time To without having to utilize scan data of current activity of the subject for the one or more times subsequent to the initial time T 0 .
16. The computer program product of claim 15 , further comprising computer readable program code that is executable by a computer, to perform a method of providing biofeedback to the subject based on the predicted future brain activity to provide an indication of said predicted future brain activity to said subject.
17. The computer program product of claim 16 , wherein the biofeedback comprises sound or light feedback.
18. The computer program product of claim 15 , wherein the raw spatio-temporal data comprises functional magnetic resonance imaging data.
19. The computer program product of claim 15 wherein constructing a full spatio temporal model comprises using a multivariate auto-regressive model to generate a model of the spatio-temporal dataset.
20. The computer program product of claim 15 , wherein selecting model parameters from the full spatio-temporal model which meet or exceed a threshold parameter that defines a level of causal relation, between voxels in the acquired dataset, comprises:
obtaining a value of a causal link between a pair of voxels;
comparing the value of the causal link to the threshold parameter;
selecting all causal links between pairs of voxels that meet or exceed the threshold parameter; and
eliminating all causal links between pairs of voxels that do not meet or exceed the threshold parameter.
21. The computer program product of claim 15 , wherein generating a reduced model representation of the full, spatio-temporal model using the selected, model parameters comprises generating a three-dimensional directed node graph having nodes and directed edges connecting the nodes, wherein nodes in the directed node graph represent voxels and wherein a directed edge in the directed node graph represents a causal relation between two nodes connected by the directed edge.