Systems and methods for modeling neural architecture
Systems and methods are described herein for modeling neural architecture. Regions of interest of a brain of a subject can be identified based on image data characterizing the brain of the subject. the identified regions of interest can be mapped to a connectivity matrix. The connectivity matrix can be a weighted and undirected network. A multivariate transformation can be applied to the connectivity matrix to transform the connectivity matrix into a partial correlation matrix. The multivariate transformation can maintain a positive definite constraint for the connectivity matrix. The partial correlation matrix can be transformed into a neural model indicative of the connectivity matrix.
1. A computer-implemented method for modeling neural architecture comprising:
identifying, by one or more processors, regions of interest of a brain of a subject based on image data characterizing the brain of the subject;
mapping, by the one or more processors, the regions of interest to a connectivity matrix, wherein the connectivity matrix is a weighted and directed or undirected network;
applying, by the one or more processors, a multivariate transformation to the connectivity matrix to transform the connectivity matrix into a partial correlation matrix, wherein the multivariate transformation maintains a positive definite constraint for the connectivity matrix; and
transforming, by the one or more processors, the partial correlation matrix into a neural model indicative of the connectivity matrix.
2. The computer-implemented method of claim 1 ,
wherein the transforming the partial correlation matrix into the neural model indicative of the connectivity matrix is based on statistics determined according to the connectivity matrix;
further comprising, transforming, by the one or more processors, the statistics into biomarker parameters.
3. The computer-implemented method of claim 2 , wherein the statistics comprise endogenous statistics.
4. The computer-implemented method of claim 3 , wherein the endogenous statistics comprise one of two-stars, transitive triads, intercept term, and any of a class of statistics computed as sums of sub-network products or transformations thereof.
5. The computer-implemented method of claim 3 , wherein the statistics further comprise exogenous statistics.
6. The computer-implemented method of claim 1 , further comprising scaling, by the one or more processors, the connectivity matrix prior to the multivariate transformation.
7. The computer-implemented method of claim 1 , further comprising determining and mapping, by the one or more processors, the regions of interest according to an activity of the subject.
8. The computer-implemented method of claim 1 , wherein the regions of interest are determined and mapped according to an atlas associated with the subject.
9. The computer-implemented method of claim 8 , wherein the atlas delineates one of a whole brain network, Default Mode Network, a Dorsal Attention Network, a Frontoparietal Control Network, a Ventral Attention Network, a Visual Network, an Auditory Network, a Reward Network, a Subcortical Network, a Salience Network, and any other network or subnetwork that captures functional or structural connectivity, and a combination thereof.
10. The computer-implemented method of claim 1 , further comprising:
identifying, by the one or more processors, distinct regions of interest in the image data; and
mapping, by the one or more processors, the distinct regions of interest of the image data to the connectivity matrix, wherein the regions of interest and the distinct regions of interest provide a joint modeling framework.
11. The computer-implemented method of claim 1 ,
wherein transforming, by the one or more processors, the partial correlation matrix into the neural model indicative of the connectivity matrix is according to a descriptive neural modeling engine, wherein values of the descriptive statistics provide an indication of a neurological condition and/or a related severity of the neurological condition when compared against a reference distribution of the descriptive statistics, wherein descriptive statistic corresponds to a biomarker for the neurological condition.
12. The computer-implemented method of claim 1 , wherein transforming the partial correlation matrix into the neural model indicative of the connectivity matrix is according to a neurological generalized exponential random graph model (neuro-GERGM) algorithm.
13. The computer-implemented method of claim 1 , wherein the image data comprises one of magnetic resonance imaging (MRI) data, functional magnetic resonance imaging (fMRI) image data, diffusion weighted imaging (DWI) data, diffusion tensor imaging (DTI) data, electromyogram (EMG) data, electroencephalogram (EEG) data, positron emission tomography (PET) data, magnetoencephalography (MEG) data, Electrocorticography (ECoG) data, ultrasound imaging data, and a combination thereof.