WO2012040219A2

WO2012040219A2 - Context differential autocorrelation mapping of functional mri

Info

Publication number: WO2012040219A2
Application number: PCT/US2011/052381
Authority: WO
Inventors: Fatta B. Nahab; Prantik Kundu
Original assignee: University Of Miami
Priority date: 2010-09-21
Filing date: 2011-09-20
Publication date: 2012-03-29
Also published as: WO2012040219A3

Abstract

A method and system for measuring brain activity by which the signal to noise ratio is increased and image acquisition time is reduced. The method includes obtaining brain oxygenation information using a functional magnetic resonance imaging apparatus, or other apparatus capable of measuring blood flow in the brain. The oxygenation information may then be encoded into a plurality of voxels, which may each be assigned a reference timecourse. The voxels associated with non-gray matter portions of the brain may be isolated and filtered. Each remaining voxel may then be correlated to every other remaining voxel, and the remaining voxels may be displayed as a function of brain activity.

Description

CONTEXT DIFFERENTIAL AUTOCORRELATION MAPPING OF

FUNCTIONAL MRI

FIELD OF THE INVENTION

The present invention relates to a method of mapping the functional integrity of brain tissue.

BACKGROUND OF THE INVENTION

It is well known that changes in blood flow and blood oxygenation in the brain (collectively known as hemodynamics) are closely linked to neural activity. When neural cells are active they increase their consumption of oxygen for energy. In order to provide neural cells with glucose, the body increases blood flow to regions of increased neural activity. These increases in blood flow lead to local changes in the relative concentration of oxyhemoglobin, the relative concentration of

deoxyhemoglobin, and changes in local cerebral blood volume and perfusion.

Functional Magnetic Resonance Imaging (fMRI) is a type of specialized MRI scan

(neuroimaging) that measures these changes in blood flow related to neural activity in the brain or spinal cord.

The increase in blood flow to the brain during activity is due, in part, to the lack within neurons of internal reserves for glucose and oxygen. As a result, increased neuronal activity requires more glucose and oxygen to be rapidly delivered through blood stream to the activated neurons. Through a process called the hemodynamic response, blood releases glucose to the area of active neurons at a greater rate than in the area of inactive neurons. Hemodynamic response results in a surplus of oxyhemoglobin in the capillaries, venules, and veins of the area of neuronal activity and allows a distinguishable change in the local ratio of oxyhemoglobin to deoxyhemoglobin. Functional magnetic resonance imaging (fMRI) can be used to measure this local ratio of oxyhemoglobin to deoxyhemoglobin.

In particular, hemoglobin is diamagnetic when oxygenated (oxyhemoglobin) but paramagnetic when deoxygenated (deoxyhemoglobin). Therefore, the magnetic resonance signal of blood is different depending on the level of oxygenation. For example, higher blood-oxygen-level dependent (BOLD) signal intensities arise from increases in the concentration of oxygenated hemoglobin, because the blood magnetic susceptibility (the degree of magnetization produced in response to an applied magnetic field) now more closely matches the tissue magnetic susceptibility. By collecting data using an MRI imaging apparatus with parameters sensitive to changes in magnetic susceptibility, BOLD contrast can be detected.

The changes in BOLD can be either positive or negative, depending upon the relative changes in both cerebral blood flow (CBF) and oxygen consumption. For example, increases in CBF that outpace changes in oxygen consumption will lead to increased BOLD signal. Conversely, decreases in CBF that outpace changes in oxygen consumption will cause decreased BOLD signal intensity. The signal difference is very small, but given many repetitions of a thought, action, or experience, statistical methods can be used to determine the areas of the brain that reliably show more of this BOLD signal intensity difference, and therefore the areas of the brain that are active during that thought, action or experience.

Almost all current fMRI research uses BOLD as the method for determining where activity occurs in the brain; however, because the BOLD signals are relative and not individually quantitative, its accuracy is low owing to a significant noise-to- signal ratio. In particular, current fMRI methods include measuring brain activity in the entire brain, including areas of non-gray matter that do not exhibit blood flow and areas of the brain that are not desirable to be measured as part of the data set. This inclusion of undesirable data creates high levels of noise and outlier data points, which affects the correlation of the active areas of the brain to a measure blood flow. Also, this inclusion can significantly increase the overall time of image acquisition. In emergency situations where rapid mapping of brain function is critical, current fMRI techniques may be inadequate.

Other methods to measure neural activity more directly have been attempted. Two examples are the measurement of the Oxygen Extraction Fraction (OEF) in regions of the brain, which measures how much of the oxyhemoglobin in the blood has been converted to deoxyhemoglobin, and directly detecting magnetic fields generated by neuronal currents. However, because the electromagnetic fields created by an active or firing neuron are so weak, the signal-to-noise ratio is extremely low and statistical methods used to extract quantitative data have been largely

unsuccessful as of yet.

Accordingly, to address the drawbacks with using fMRI to detect brain activity, what is needed is a method to evaluate brain activity that reduces noise and increases the overall signal quality such that the entire brain can be functionally mapped with a smaller dataset.

SUMMARY OF THE INVENTION

The present invention advantageously provides an autocorrelation method for measuring brain activity by which the signal to noise ratio is increased. The method may include obtaining brain oxygenation information using a functional magnetic resonance imaging apparatus or other apparatus capable of measuring blood flow in the brain and characterizing the data into a dataset including a plurality of voxels (MCC voxels); characterizing each MCC voxel as being associated with either gray matter or non-gray matter; using the blood flow data from the non-gray matter MCC voxels to calculate a spatial baseline for evaluating functional connectivity; retaining all gray matter MCC voxels and a threshold percentage of non-gray matter MCC voxels for correlation; mass-cross-correlating retained MCC voxels based on blood flow data using a computer; characterizing the existence of functional connectivity between MCC voxels, culling voxels with no functional connectivity; retaining one or more MCC voxels in a subset (as subset voxels); and cross-correlating the subset voxels to determine functional connectivity between voxels associated with gray matter portions of the brain.

Using this correlation data, an automated process may be carried out on a computer or machine in which the overall neurally related BOLD activity expressed in a voxel's fMRI signal timecourse is determined by comparing the correlation to other gray matter voxels and to non-gray matter voxels. The results may be displayed as a function of brain activity (referred to herein as a "map" of brain activity or functional connectivity of voxels).

The method may be used to detect neurodegeneration or loss of neuronal integrity associated with Parkinson's disease, Alzheimer's disease, or other maladies such as seizures, tremors, aging effects, or early detection of brain damage resulting from strokes. Furthermore, this method may be used to rapidly detect hyperacute stroke in an emergency setting.

BRIEF DESCRIPTION OF THE DRAWINGS

A more complete understanding of the present invention, and the attendant advantages and features thereof, will be more readily understood by reference to the following detailed description when considered in conjunction with the accompanying drawings wherein:

FIG. la is a flow chart showing a first portion of an exemplary method of mapping brain activity; and

FIG. lb is a flow chart showing a second portion of an exemplary method of mapping brain activity.

DETAILED DESCRIPTION OF THE INVENTION

The present invention advantageously provides for a method that maps the functional integrity of brain tissue based on its functional connectivity, computed from the interregional similarities in oxygenation over a reference timecourse with the exclusion of noise effects. Using the method described herein, a single

autocorrelation metric may be created for the neuronal activity within a voxel-based BOLD coherence (BC) with other voxels in the brain.

This method uses voxels as the basic unit for comparison. A "voxel" is a volumetric pixel, digitized from functional magnetic resonance or other blood oxygen data (such as BOLD). The relationship of one voxel (the "seed voxel") with regions within other voxels across the brain ("target voxels") may be calculated based on the correlation of the one voxel to another individual voxel, and these relationships may be used to analyze and visualize the functional connectivity of the entire brain.

Secondly, either sequentially or simultaneously, the contribution of uncharacterized non-gray noise and artifacts is removed from the correlation of the voxels over the reference timecourse. Although the functional connectivity of the voxel timecourses may be compared, the comparison may be referred to herein simply as a comparison of voxels. The correlation of a voxel to all other voxels determines whether that voxel has oxygenation similarities or whether it is properly classified as noise and may be discarded. For a given point in time, fMRI signal may be acquired over a field of view and digitized into voxels of an fMRI image. In a series of images acquired over time, each constituent voxel has a history of fMRI signal over time, which is its signal timecourse. Functional MRI signal timecourses from gray matter areas of the brain are likely to reflect neuronal activity, whereas fMRI signal timecourses from non-gray matter has most signal fluctuations from noise because non-gray matter has no neurally-related BOLD signal. If a gray matter voxel has an fMRI timecourse with most correlation to other gray matter voxels, then the timecourse may to be related to neuronal activity. If, on the other hand, a gray matter voxel has an fMRI timecourse with most correlation to non-gray matter, then that voxel represents an area of no interest.

Referring now to FIGS, la and lb, an exemplary method of mapping brain activity is shown. FIGS, la and lb together show the entire method, with FIG. la showing the first step (101), the second step (102), and part of the third step (103); and with FIG. lb showing the second part of the third step (103), and the fourth (104), fifth (105), and sixth (106) steps of the method. In the first step of the method (101), the brain may be mapped using T2 or T2* mapping or anatomical image

segmentation. This may allow for rapid separation of areas of the brain known to contain relevant gray matter and areas that obviously contain non-gray matter (such as cerebral white matter or cerebrospinal fluid).

Additionally, the brain may be imaged using echo-planar imaging fMRI or other method operable to measure blood flow in the brain. For example, for a given point in time, fMRI signal is acquired over a field of view and digitized into voxels. Data such as BOLD oxygenation information for each voxel may be collected and stored on a computer hard drive or other suitable device. Once the brain is mapped and imaged using the appropriate hardware, all data collected may be stored in and/or transferred to a computer, computer board, or other hardware capable of processing the data according to the remaining steps of the method (i.e. the device may have installed therein software, algorithm(s), or other programs useful for performing the method). In the second step of the method (102), the data may be encoded (by a computer or other machine capable of performing the required calculations) into datasets of, for example, three space dimensions and one time dimension. Each spatial location (a voxel) is given a timecourse, for example a few minutes, to express the blood flow data for each location over a reference period of time. In a series of images acquired over time, each constituent voxel may have a history of fMRI signal over the reference timecourse. The As used herein, a "voxel" is defined as the basic unit of MRI signal intensity in three dimensions. Even though other criteria than BOLD may be used, the analysis of a voxel over a reference timecourse is referred to herein as a "BOLD timecourse." Likewise, functional connectivity between voxels may be referred to herein as "neuronal coherence," "functional coherence,"

"coherence," or "BOLD coherence."

Voxels may be labeled as being associated with gray or non-gray matter (as referred to herein, "gray matter voxels" and "non-gray matter voxels") using the results of the mapping step (T2* mapping or anatomical image segmentation). In addition to gray matter voxels, approximately 10% to approximately 20% of the voxels determined by mapping to be in non-gray matter areas of the brain may be retained for later comparison to seed voxels. This threshold of non-gray matter voxels may be taken into consideration because the non-gray matter voxels may be difficult to isolate without excluding a relevant portion of the gray matter voxels; therefore, a percentage of non-gray matter voxels may be used to account for the boundary area between gray matter and non-gray matter. Using approximately 10% to 20% non- gray matter voxels as a boundary was determined to be the most efficient amount for purposes of this calculation; however, any percentage of non-gray matter voxels may be characterized by a correlation coefficient during the reference timecourse, depending on the desired time of image acquisition or the purpose of imaging.

In the third step of the method (103), which may be performed simultaneously with the fourth step (104) of the method, the BOLD timecourse of a seed voxel may be standard-cross-correlated ([S-]CC) with every other voxel timecourse in the form of a linear regression (referred to herein as "mass-cross-correlation" (MCC)). The voxels used in this step may be referred to as "MCC voxels." In this manner, the functional connectivity (or BOLD coherence, BC) between voxels may be characterized. If there is no BOLD coherence between a seed MCC voxel and target voxels, data for that voxel may be culled from the dataset as a structural context of no interest (S-CONI). BOLD coherence between a seed MCC voxel and retained voxels (i.e. gray matter voxels and the approximately 10% to approximately 20% of boundary non-gray matter voxels) may be potentially, but not necessarily, due to neuronal coherence (NC). For example, the BOLD coherence may be the result of NC BOLD coherence (NBC) as between voxels associated with gray matter portions of the brain. Alternatively, the BOLD coherence may be the result of motion artifacts and other noise, or the result of coherence with one or more non-gray matter voxels, and therefore may be characterized as non-NC BOLD coherence (nNBC). Voxel coherence driven by NBC is of particular interest in evaluating brain function;

because there is little or no neural activity in the non-gray matter portions of the brain, the correlation between a seed voxel and one or more non-gray matter voxels may provide very little to no useful neural activity information. Therefore, the coherence may be characterized in order to separate NBC from nNBC effects.

In the fourth step of the method (104), which may be performed

simultaneously with the third step (103), a spatial baseline for the distribution of correlation values may be determined. The spatial baseline may be derived from the BOLD timecourse data from areas of the brain known to be non-gray matter, as by a correlation process as described in the third step (103). Non-gray matter portions of the brain, where no nerve cell bodies are located (for example, white matter and the ventricles), may display limited or no BOLD response. The predominant signal measured in these non-gray areas is therefore noise. The spatial baseline, as a range of values of non-neuronal functional connectivity between non-gray matter voxels, may be used for qualifying and quantifying the correlation of seed voxels to other areas of the brain, as described in the fifth step of the method (105).

In the fifth step of the method (105), a subset cross-correlation may be performed between the voxels to separate NBC from nNBC effects. The subset cross- correlation may include voxels retained from the mass-cross-correlation, that is, seed MCC voxels displaying coherence to target voxels. The voxels considered in the subset of cross-correlations may be referred to as "subset voxels." The maximum NBC value (referred to herein as "pNBC") of a seed subset voxel may be determined through a multivariate contrast with the pre-determined spatial baseline (the nNBC coherence values of areas of non-neuronal origin, as described in the third step (103)).

As described in the fourth step (104), correlation between a gray matter voxel and one or more non-gray matter voxels is driven by effects of no interest or neural interpretability. Therefore, the level of coherence in gray matter regions of the brain may be considered relevant only after outweighing coherence in non-gray regions. A maximum NBC value of a seed subset voxel may be determined by a multivariate contrast with BOLD coherence of predetermined non-neuronal origin to (nNBC) create the structurally differential autocorrelation metric. The proportion of

NBC/nNBC may be referred to as pNBC. If a high level of coherence is found between a seed subset voxel and one or more target voxels (that is, a pNBC value of approximately greater than 1), the seed subset voxel may be deemed a structural context of interest (S-COI). Alternatively, a relatively low level of coherence as compared to the spatial baseline (that is, a pNBC value of approximately less than or equal to 1) between a seed subset voxel and one or more target voxels, the seed subset voxel may be deemed an S-CONI because the coherence may be substantially driven by nNBC. Such S-CONI data may be culled from the subset, thereby removing irrelevant noise from the dataset. In this way, once the spatial baseline is established, a particular portion of the brain may be autocorrelated to irrelevant non-gray matter portions and removed from the dataset, or autocorrelated to relevant gray matter regions and used to determine functional integrity of the brain.

In the sixth step of the method (106), which may be performed simultaneously with one or more of the other steps, the results of the autocorrelation process (i.e. functional connectivity of the brain) may be displayed by a computer or other machine, such as on a monitor, projection display, or the like. The results may be displayed as an overlay to the image of the brain. For example, areas of neuronal activity may be colorized differently than areas of no neuronal activity, such as damaged areas of the brain or non-gray matter. Alternatively, any other display technique is also contemplated, such as a black-and-white image, abstract

representation of the brain, data chart, or other meaningful expression.

It will be appreciated by persons skilled in the art that the present invention is not limited to what has been particularly shown and described herein above. In addition, unless mention was made above to the contrary, it should be noted that all of the accompanying drawings are not to scale. A variety of modifications and variations are possible in light of the above teachings without departing from the scope and spirit of the invention, which is limited only by the following claims.

Claims

What is claimed is:

1. A method for measuring brain activity comprising the steps of:

characterizing functional magnetic resonance imaging data for the brain into a dataset comprising a plurality of MCC voxels, each MCC voxel having a reference timecourse;

characterizing each MCC voxel as being either a gray matter or non-gray matter voxel;

assigning substantially all gray matter MCC voxels to a dataset;

assigning a threshold percentage of non-gray matter MCC voxels to the dataset;

creating a spatial baseline of functional magnetic resonance imaging data for non-gray matter MCC voxel data;

mass-cross-correlating a seed MCC voxel with all other assigned MCC voxels using functional magnetic resonance imaging data;

characterizing the existence of functional connectivity between the seed MCC voxel and other assigned MCC voxels;

assigning one or more seed MCC voxels from the mass-cross-correlation to create a group of subset voxels; and

cross-correlating the subset voxels.

2. The method of claim 1, wherein the functional magnetic resonance imaging data is blood-oxygen-level dependence (BOLD) information.

3. The method of claim 1, wherein each MCC voxel is characterized using T2* mapping, T2 mapping, or anatomical image segmentation.

4. The method of claim 1, wherein the threshold percentage of non-gray matter MCC voxels is between approximately 0.1% and approximately 30%.

5. The method of claim 1, wherein the threshold percentage of non-gray matter MCC voxels is between approximately 10% and approximately 20%.

6. The method of claim 1, wherein characterizing the existence of functional connectivity between the seed MCC voxel and the other assigned MCC voxels includes defining the functional connectivity according to the group consisting of: no coherence between voxels; or coherence between voxels.

7. The method of claim 6, wherein a definition of no coherence between voxels further includes removing the seed MCC voxel from the cross-correlation subset.

8. The method of claim 6, wherein a definition of coherence between compared voxels further includes assigning the seed MCC voxel to the group of subset voxels.

9. The method of claim 8, wherein the cross-correlation of the subset voxels includes determining the pNBC value of one or more seed subset voxels.

10. The method of claim 9, wherein determining the pNBC value further includes comparing the coherence value of the seed subset voxel to one or more gray matter voxels to the coherence value of the spatial baseline.

11. The method of claim 10, wherein a pNBC value of approximately greater than 1 further includes characterizing the coherence of the one or more subset voxels as being neuronal coherence (NBC).

12. The method of claim 10, wherein a pNBC value of approximately less than or equal to 1 further includes characterizing the coherence of the one or more subset voxels as being non-neuronal coherence (nNBC).

13. A method for detecting loss of neuronal integrity comprising:

providing a functional magnetic resonance imaging apparatus; and mapping functional integrity of brain tissue comprising the steps of:

collecting blood-oxygenation-level dependence (BOLD) signal information from a mammalian brain using functional magnetic resonance imaging over a reference timecourse; characterizing the BOLD signal information into a dataset comprising a plurality of voxels;

characterizing each voxel as being either a gray matter or non-gray matter voxel;

assigning substantially all gray matter voxels to a dataset;

assigning one or more non-gray matter voxels to the dataset;

determining a spatial baseline of BOLD signal information from non- gray matter voxels;

assigning data from approximately 5% to approximately 30% of the non-gray matter voxels to the dataset;

filtering noise data from the dataset;

displaying a map of the functional connectivity of the brain.

14. The method of claim 13, wherein data from approximately 10% to approximately 20% of the non-gray matter voxels is assigned to the dataset.

15. The method of claim 13, wherein the step of filtering noise data from the dataset includes:

mass-cross-correlating a seed voxel with one or more other voxels in the dataset using BOLD data;

characterizing the existence of coherence between the seed voxel and the one or more other voxels in the dataset as being no coherence between voxels or coherence between voxels;

assigning one or more seed voxels to a subset group of voxels; and cross-correlating the subset group of voxels.

16. The method of claim 15 wherein a characterization of coherence between voxels further includes assigning the seed voxel to the subset group of voxels.

17. The method of claim 16, wherein the cross-correlation of the subset group of voxels includes determining the ratio of neuronal coherence values between one or more subset voxels in the subset group to non-neuronal coherence values of the spatial baseline.

18. The method of claim 17, wherein

coherence is due to neuronal coherence (NBC) if the ratio is approximately greater than 1 ; and

coherence is due to non-neuronal coherence (nNBC) if the ratio is

approximately less than or equal to 1.

19. The method of claim 13, wherein a voxels is characterized as a gray matter or non-gray matter voxel using T2* mapping, T2 mapping, or anatomical image segmentation.

20. A method for detecting loss of neuronal integrity comprising:

providing a functional magnetic resonance imaging apparatus;

providing a computing device; and

performing a method of mapping functional integrity of mammalian brain tissue comprising the steps of:

mapping the brain tissue using t2 mapping, t2* mapping, or anatomical image segmentation;

collecting BOLD signal information from the brain tissue using functional magnetic resonance imaging over a reference timecourse;

characterizing the BOLD signal information into a dataset comprising a plurality of MCC voxels;

characterizing each MCC voxel as being either a gray matter or non- gray matter MCC voxel;

assigning substantially all gray matter MCC voxels to a dataset;

assigning approximately 10% to approximately 20% of non-gray matter MCC voxels to the dataset;

creating a spatial baseline of non-gray matter BOLD signal

information data; mass-cross-correlating a seed MCC voxel with one or more other assigned MCC voxels using BOLD signal information data;

characterizing the existence of functional connectivity between the seed MCC voxel and one or more other assigned MCC voxels, wherein

if the existence is characterized as coherence between MCC voxels, then the seed MCC voxel is assigned to a group of subset voxels; and

if the existence is characterized as no coherence between MCC voxels, then the seed MCC voxel is removed from the dataset;

cross-correlating the subset voxels, wherein

if a seed subset voxel has a BOLD coherence value equal to or less than the spatial baseline, then the seed subset voxel does not have neuronal connectivity to gray matter portions of the brain;

if a seed subset voxel has a BOLD coherence value greater than the spatial baseline, then the seed subset voxel has neuronal connectivity to gray matter portions of the brain;

using the computing device to generate a display of brain function that includes areas of neuronal connectivity.