Resting‐state functional MRI: Functional connectivity analysis of the visual cortex in primary open‐angle glaucoma patients

Abstract

Purpose: To analyze functional connectivity (FC) of the visual cortex using resting‐state functional MRI in human primary open‐angle glaucoma (POAG) patients. Materials and Methods: Twenty‐two patients with known POAG and 22 age‐matched controls were included in this IRB‐approved study. Subjects were evaluated by 3 T MR using resting‐state blood oxygenation level dependent and three‐dimensional brain volume imaging (3D‐BRAVO) MRI. Data processing was performed with standard software. FC maps were generated from Brodmann areas (BA) 17/18/19/7 in a voxel‐wise fashion. Region of interest analysis was used to specifically examine FC among each pair of BA17/18/19/7. Results: Voxel‐wise analyses demonstrated decreased FC in the POAG group between the primary visual cortex (BA17) and the right inferior temporal, left fusiform, left middle occipital, right superior occipital, left postcentral, right precentral gyri, and anterior lobe of the left cerebellum. Increased FC was found between BA17 and the left cerebellum, right middle cerebellar peduncle, right middle frontal gyrus, and extra‐nuclear gyrus (P < 0.05). In terms of the higher visual cortices (BA18/19), positive FC was disappeared with the cerebellar vermis, right middle temporal, and right superior temporal gyri (P < 0.05). Negative FC was disappeared between BA18/19 and the right insular gyrus (P < 0.05). Region of interest analysis demonstrated no statistically significant differences in FC between the POAG patients relative to the controls (P > 0.05). Conclusion: Changes in FC of the visual cortex are found in patients with POAG. These include alterations in connectivity between the visual cortex and associative visual areas along with disrupted connectivity between the primary and higher visual areas. Hum Brain Mapp 34:2455–2463, 2013. © 2012 Wiley Periodicals, Inc.

INTRODUCTION

Resting‐state functional magnetic resonance imaging (fMRI) has recently become a focus of functional brain imaging. This technique is promising for the evaluation of spontaneous resting human brain activity as well as in assessment of the functional relationship between brain regions. Functional MRI has been used to study Alzheimer disease [Rombouts and Scheltens, 2005; Rombouts et al., 2005; Zhang et al., 2009], schizophrenia [Whitfield‐Gabrieli et al., 2009], depression [Yao et al., 2009], epilepsy [Zhang et al., 2010], addiction disorders [Liu et al., 2010], and attention deficit hyperactivity disorder (ADHD) [Tian et al., 2008], among other conditions. Glaucoma is the leading cause of blindness worldwide. Not only does glaucoma injure the optic nerve, cross‐synaptic degeneration of the lateral geniculate body (the secondary neuron) as well as the visual cortex (the tertiary neuron) has been previously demonstrated. The primary visual area, which receives primary visual stimulation, also forms numerous connections with the secondary visual cortex. The role of these connections may relate to the integration of visual information and generation of conscious perception. Associated pathological changes of primary open‐angle glaucoma (POAG) include dendritic neuronal degeneration, pyknosis, and death of neuronal soma, as well progressive loss of cells throughout the visual pathway, and atrophy of the lateral geniculate body and visual cortex [Gupta and Yucel, 2007; Ito et al., 2009; Yucel et al., 2003]. These structural changes in the visual cortices of glaucoma patients are likely to be caused by anterograde cross‐synaptic degeneration, thus raising the question of what, if any, functional changes also occur.

This study uses resting‐state fMRI to evaluate functional connectivity (FC) changes in POAG patients.

MATERIALS AND METHODS

Subjects

After obtaining approval of the Medical Ethics Review Committee of Tongji Hospital and obtaining informed consent in accordance with the Declaration of Helsinki, 22 POAG patients and 22 gender and age‐matched healthy volunteers were enrolled in the study from April to August 2010. Seventeen males and 5 females aged 21–54 years old (mean age 25) were included in the POAG group and were matched with 17 male and 5 female healthy volunteers aged 21–55 years old (mean age 36), included in the control group. There were no statistically significant differences in age and gender between the two groups (P > 0.05). Informed written consent was obtained for all participants.

Subjects were recruited into the study based on the clinical diagnostic criteria of POAG: a history of open anterior chamber angle, visual field defects, abnormal optic disk, and increased intraocular pressure [Fleming et al., 2005]. Subjects underwent a thorough history and physical examination including an ophthalmologic examination. Inclusion criteria for the POAG group were: (1) a clinical examination confirming POAG and (2) the presence of a visual field defect. Exclusion criteria for the POAG group included (1) clinical evidence or history of other oculopathy; (2) abnormality detected in the optic pathway or brain on routine noncontrast MRI examination; (3) evidence of systemic disease including hypertension and diabetes; (4) use of alcohol, caffeine, or nicotine within the last 3 months. Inclusion criteria of the control group were age and gender matched healthy volunteers to patient group without clinical evidence or history of glaucoma. Exclusion criteria were (1) evidence of hypertension, diabetes, or other systemic disease by routine clinical examination or history; (2) the presence of ocular disease by routine clinical ophthalmic test; and (3) neurological or neurosurgical disease identified by history or examination.

MRI Examination

All subjects underwent MRI examinations. No subjects had taken any medications on the day of the experiment. A 3 T MR scanner (Signa HDxt, GE Healthcare, Milwaukee, WI) with 8‐channel head array coil was used in this study. Subjects were scanned in a supine, head‐first position with symmetrically placed cushions on both sides of head to decrease motion. Resting‐state blood oxygenation level dependent (BOLD) and three‐dimensional brain volume imaging (3D‐BRAVO) sequences were obtained. Conditions of eyes‐closed wakefulness without head motion were ensured throughout acquisition of the resting‐state BOLD examination [Liu et al., 2006; Zang et al., 2004]. Additionally, the subjects were instructed to not engage in any specific thinking activity. For the purposes of real‐time imaging processing, gradient echo echo planar imaging sequences with following parameters were acquired: repetition time (TR)/echo time (TE) 2000/30 ms, field of view (FOV) 240 mm × 240 mm, matrix 64 × 64, slices 33, slice thickness 3 mm, slice gap 1 mm, scan time 8 min. A 3D‐BRAVO sequence was used for structural data acquisition with: TR/TE/inversion time (TI) 6/2/380 ms, FOV 240 mm × 240 mm, slice 1.2 mm, slice gap 0 mm, matrix 192 × 192, number of signal averages (NEX) 1, flip angle 15°, bandwidth 41.67 Hz, scan time 1 min 40 s.

Data Processing

Standard professional data processing software, Data Processing Assistant for Resting‐state fMRI (DPARSF 1.0, http://restfmri.net/forum/DPARSF), was used for data analysis. DPARSF is plug‐in software run on a matrix laboratory platform (MATLAB R2008a) and is based on statistical parametric mapping (SPM 5, http://www.fil.ion.ucl.ac.uk/spm) and a resting‐state fMRI data analysis toolkit (REST 1.5, by Song et al., http://www.restfmri.net).

Preprocessing

The preprocessing steps are illustrated as follows. After converting DICOM files to NIFTI images, the first 10 time points were discarded. Slice timing and head motion correction were then performed. One patient was rejected after head motion correction, and the remaining data was then normalized to Montreal Neurological Institute (MNI) space by using echo planer imaging (EPI) templates and resampling to 3‐mm isotropic voxels. After smoothing with a 4‐mm full width half maximum (FWHM) Gaussian kernel, the linear trend of time courses was removed and the band‐pass temporal filtering (0.01–0.08 Hz) was then performed. A default mask made from SPM5′s a priori mask named brainmask.nii, which covers whole brain contained gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF), was used with a threshold of 50% [Chao‐Gan and Yu‐Feng, 2010].

Removal of nuisance covariate effects

The region of interest (ROIs) were defined and the built‐in mask file of REST software was selected to extract covariates, including head motion parameters, as well as whole brain, WM, and CSF signal. These covariates were subsequently integrated into one file (i.e., covlist.txt). The covariates were then removed [Chao‐Gan and Yu‐Feng, 2010].

FC analysis

Assessments of FC can be performed in many ways, among which the ROI methods are most frequently used. These include voxel‐wise and ROI‐wise analyses. The procedure of the ROI‐wise method is outlined as follows: time sequence data is extracted from a given ROI and that data is used as a regressor for linear correlative analysis. The correlation coefficients between the ROI and voxels of whole brain cortices are calculated, and subsequently, the regions of brain which relate to the ROI in their time courses are highlighted. In this way, FC between the ROIs and target brain regions are illustrated. If the mean time courses of a brain area and the ROI are significantly related, it is concluded that FC exists between these two regions.

ROI definition

ROIs were defined within Brodmann areas (BA) of the brain related to vision (including BA17, BA18, BA19, and BA7) based on anatomic definitions. Bilateral circular ROIs, 12 mm in diameter, was placed in BA17 (i.e., striate cortex), centered on the calcarine sulcus in the medial occipital lobe. BA17 is defined as being part of the primary visual area and is the primary area receiving the direct visual stimulation. ROIs in BA18 and BA19 brain areas, the same size and shape as those in BA17, were similarly placed in the center of BA18 and BA19. BA18/19 are extra‐striatal cortices which are adjacent to the BA17 area in the occipital lobe and are defined as part of the higher visual cortex—the major function of which is thought to be the integration of visual information and generation of conscious perceptions. The BA7 brain area, part of the dorsal visual network, is located in the superior parietal lobe, and in this area, circular bilateral ROIs, 12 mm in diameter, were also placed [Jann et al., 2010]. Anatomic distributions of aforementioned brain regions and their ROIs are shown in Figures 1 and 2. The centers of the ROIs in MNI space were (−8, −76, 10) and (7, −76, 10) for BA17, (−26, −94, 0) and (24, −94, 0) for BA18, (−41, −80, 0) and (40, −80, 0) for BA19, as well as (−11, −76, 53) and (13, −76, 53) for BA7.

Figure 1.

Distribution of Brodmann areas related to visual function, including BA17 (Khaki area), BA18 (yellow area), BA19 (tan area), and BA7 (maroon area).

Figure 2.

The ROIs were placed in BA17 (top), BA18/19 (middle), and BA 7 (bottom). The MNI coordinates of the ROI centers were (−8, −76, 10) and (7, −76, 10) for BA17, (−26, −94, 0) and (24, −94, 0) for BA18, (−41, −80, 0) and (40, −80, 0) for BA19, and (−11, −76, 53) and (13, −76, 53) for BA7, respectively.

Voxel‐wise analysis

Using the BA17 ROI, FC analysis of voxels in this region was performed relative to those of the remaining brain on a voxel by voxel basis. A FC map of the whole brain demonstrating FC to BA17 was constructed. The BA18/19 ROI was analogously used to create a FC map illustrating the FC between BA18/19 and the remaining brain.

The statistical analysis module of the REST software was used to analyze the FC maps. A one‐sample t‐test was used to analyze the FC maps of POAG and control group, so as to obtain the maps of the t statistics (t‐maps) for each group. Two‐sample t‐test analysis was used to detect whether there were significant statistical differences between the FC of POAG and control groups. Statistical t‐maps were also created from two‐sample t‐tests. This was performed on original FC maps of both BA17 and BA18/19. All FC t‐maps were corrected by the AlphaSim program (http://afni.nimh.nih.gov/pub/dist/doc/manual/AlphaSim.pdf), with connectivity standards set as: rim connection 5 mm, Gaussian smoothing core 4 mm, P value before calibration < 0.05, minimal cluster of voxels 85. Regions consisting of fewer than 85 voxels were thus removed from the final t‐map.

Then, the positive FC (>0), negative FC (<0), or no FC (= 0) areas were, respectively, abstracted from one sample t‐maps of patients (P) and controls (C) group to make different masks in the form of intersection of sets (∩). The “decreased FC” (P‐C < 0) was refined into “decreased positive FC” (using the mask P > 0 ∩ C > 0), “disappeared positive FC” (P = 0 ∩ C > 0), “appeared negative FC” (P < 0 ∩ C = 0), or “increased negative FC” (P < 0 ∩ C < 0) by applying masks of significant within‐group one‐sample t‐test on the two‐sample t‐test results. The same method was used in refining the “increased FC” (P‐C > 0) into “increased positive FC” (P > 0 ∩ C > 0), “appeared positive FC” (P > 0 ∩ C = 0), “disappeared negative FC” (P = 0 ∩ C < 0) or “decreased negative FC” (P < 0 ∩ C < 0).

ROI‐wise analysis

ROI were placed in BA17, BA18, BA19, and BA7 as above and FC correlation coefficients (r‐values) analyzed among them on a pair‐wise basis. This yielded six total r‐values including BA17‐BA18, BA17‐BA19, BA17‐BA7, BA18‐BA19, BA18‐BA7, and BA19‐BA7. As the r‐values obtained (ranging from −1 to 1) did not obey a normal distribution, Fisher's Z‐transformation was used for conversion from an r to Z value. Six Z‐values were, thus, obtained for each subject. For subsequent statistical analysis, Statistical Package for the Social Sciences (SPSS 13.0) software was used. Mean Z‐values of the POAG group and of the control group were obtained. The Z‐values between the POAG and control group were then compared for each region using a two‐sample t‐test. P‐values less than 0.05 were considered statistically significantly different. Adjustments for the multiple comparisons performed were made via a Bonferroni correction.

Controlling for the effects of atrophy within ROIs

The presence of brain atrophy may cause partial volume effects in functional imaging techniques. To account for this, a voxel‐based‐morphometry (VBM) analysis of structural images was performed. Each structural image was co‐registered to the mean functional image via linear transformation following motion correction. The transformed structural images were then segmented into GM, WM, and CSF by using a unified segmentation algorithm, transforming the structural images into MNI space. The motion corrected functional images were spatially normalized to MNI space using the normalization parameters estimated during segmentation. Thus, the structural images corresponded to the functional images in terms of MNI coordinates. Brain volumes of the ROIs described above were then obtained by VBM analysis with these values taken as covariates in both the voxel‐wise and ROI‐wise analyses. The statistical threshold was set at P < 0.05.

RESULTS

Voxel‐Wise Analysis

One‐sample FC t‐maps

The t‐maps illustrating the FC between the BA17 ROI and the remaining brain voxels of the control and patient group are shown in Figure 3 (top and middle images). T‐maps for the BA18/19 ROIs are provided in Figure 4 (top and middle images).

Figure 3.

A sample FC t‐map between BA17 and the voxels of the remaining brain of the control (top) and patient groups (middle). A map of FC differences between POAG and control groups is also shown (bottom). Brain regions demonstrating statistically significant differences between the groups in terms of FC with BA17 are listed in Table 1.

Figure 4.

One sample FC t‐map between BA18/19 and the voxels of the remaining brain of the control (top) and patient groups (middle). A map of FC differences between POAG and control groups is also shown (bottom). Brain regions demonstrating statistically significant differences between the groups in terms of FC with BA18/19 are listed in Table 2.

Map of FC differences between POAG and control groups (BA17‐ROI)

The FC between BA17 ROI and other brain regions of the POAG group were compared with that of control group. A map of FC differences was created and is shown in Figure 3 (bottom). The brain regions with significant FC differences are identified in Table 1 by the Brodmann Area (BA), the volume of brain region with FC difference, MNI coordinates, the difference between two groups, and the type of FC changes seen in POAG patients. The difference is labeled as “T‐value,” representing the calculated t statistic. The higher T‐values represent a greater extent of differences between the two groups. When compared with the control group, POAG patients demonstrated a disappeared positive FC between BA17 and the right inferior temporal gyrus, left middle occipital gyrus, left postcentral gyrus, and the right precentral gyrus. Appeared negative FC was showed between BA17 and the anterior lobe of the left cerebellum in the POAG patients. The positive FC between the right superior occipital gyrus and the BA17 ROI persisted but was diminished in the POAG group (P < 0.05). Disappeared negative FC were noted between the extra‐nuclear, right middle frontal gyrus, right middle cerebellar peduncle, left cerebellum, and the BA17 ROI in the POAG group relative to the controls (P < 0.05).

Table 1.

Brain regions with BA17 FC differences between POAG versus control groups (P < 0.05)

Map of FC differences between POAG and control groups (BA18/19‐ROI)

The FC between the BA18/19‐ROIs and the remaining brain regions were also compared between the control and POAG groups. The map of those FC differences is illustrated in Figure 4 (bottom). The brain regions with significant FC differences are identified in Table 2 by the BA region, the volume of active brain region, MNI axis, the difference between two groups, and the types of FC change. When compared with the control group, disappeared positive FC was showed in the vermis in POAG patients, right middle temporal gyrus, and right superior temporal gyrus. Disappeared negative FC was demonstrated between the right insular gyrus and BA18/19 in the POAG group (P < 0.05).

Table 2.

Brain regions with BA18/19 FC differences between POAG versus control groups (P < 0.05)

ROI‐Wise Analysis

Using ROI‐wise analysis, Z‐values reflecting the FC among visual areas of the brain (BA17, BA18, BA19, and BA7) were calculated for the comparisons between the POAG and control groups. No statistically significant differences were found among these ROIs (P > 0.05) (Table 3).

Table 3.

FC differences between brain regions based on ROI‐wise analysis (POAG vs. control group)

Brain regions	Mean Z‐value (control group)	Mean Z‐value (POAG group)	P value
BA17‐BA18	0.29 ± 0.08	0.24 ± 0.06	0.62
BA17‐BA19	0.38 ± 0.07	0.24 ± 0.05	0.08
BA17‐BA7	−0.10 ± 0.06	−0.19 ± 0.05	0.29
BA18‐BA19	0.69 ± 0.05	0.58 ± 0.08	0.24
BA18‐BA7	−0.17 ± 0.05	−0.13 ± 0.05	0.56
BA19‐BA7	−0.20 ± 0.05	−0.08 ± 0.08	0.21

DISCUSSION

Major Findings and Relation to Previous Studies

In this study, it was found that the positive FC between BA17 and the right inferior temporal gyrus (BA37), left middle occipital gyrus (BA19), left postcentral gyrus (BA4), right precentral gyrus (BA6), and the right superior occipital gyrus (BA19) was decreased in POAG patients. The BA17 area is part of the primary visual cortex—the predominant brain area receiving direct input relating to visual stimuli. A recent study of fMRI in task mode with visual stimuli demonstrated diminished BOLD response within the primary visual cortex in POAG patients. This decreased activity may be reflected in the fact that early POAG patients demonstrate greater difficulties in dealing with their daily visual tasks than would be otherwise be predicted by the extent of visual field losses [Qing et al., 2010]. Our study was conducted in a task‐free mode, and the results regarding the specific regions discussed above most likely relate to spontaneous fluctuations that are known to occur in the primary visual area in the task‐free state [Kenet et al., 2003, Shapley et al., 2003, Nir et al., 2006]. Research by Wang et al. demonstrated areas of nonrandom activity within the primary visual area, and also in associated neural networks including the bilateral visual areas (precuneus, cuneus, and lingual gyri), the left precentral/middle frontal gyrus, and the right precentral/postcentral gyrus, as well as the bilateral temporal lobes (bilateral middle and right inferior temporal gyri along with bilateral fusiform and left parahippocampal gyri) [Wang et al., 2008]. Wang et al. suggested that such task‐free state activity may be associated with memory‐related imagery or visual memory consolidation. In this work, diminished FC between BA17 and these regions in POAG patients may thus reflect impairment in either of these processes, possibly secondary to diminished function in the primary visual cortex. Disappeared positive FC between BA17 and the inferior temporal gyrus, which plays a significant role in integrating visual, auditory, and tactile stimulation, may also relate to impairment in the incorporation of visual with other stimuli in PAOG patients. The voxel‐wise analysis in this study also demonstrates disappeared and decreased positive FC between the primary visual area (BA17) and higher visual cortices (BA19), which suggests decreased transmission of visual information from primary to higher visual cortices. This difference was, however, only borderline statistically significant in the ROI‐wise analysis (P = 0.075).

Some FC changes of specific resting state networks (RSN) were found in this study, coinciding with those studied by Jann et al. [2003] in a combined fMRI and electroencephalogram (EEG) analysis. Specifically, Jann et al. described five RSN related to higher cognitive functions such as self‐reflection, working memory, and language, and five other RSNs correlated with somatosensory functions. Two RSN components exhibiting diminished FC with BA 17 in POAG patients are the left postcentral gyrus (BA4; primary motor cortex; somato‐motor network/RSN6) and right precentral gyrus (BA6; right working memory or language network/RSN5). These findings may reflect diminished integration of visual information into working/language memory or somatomotor information processing owing to dysfunction of the primary visual cortex or possibly secondary to the decreased FC between the primary and higher visual cortices. The left fusiform gyrus (BA19; ventral visual network/RSN9) is part of the associative visual cortex—a network important for object recognition. Decreased positive FC between BA17 and this area may thus reflect impairment in object identification in POAG patients.

In our study, there was disappeared positive FC between the BA18/19 area and the vermis, right middle temporal gyrus (BA37), and right superior temporal gyrus (BA22) in the POAG group. One potential interpretation for the findings in the temporal gyrus is analogous to that discussed above for the primary visual cortex. With respect to the cerebellum, looped structural connectivity between cerebral and cerebellar cortices has been previously confirmed anatomically [Middleton and Strick, 1994]. The positive FC between the vermis and primary/higher visual cortex may be reduced simply as a result of diminished input from the hypoactive primary/higher visual area. These findings may also reflect diminished incorporation of visual information into functions of the vermis such as posture and locomotion.

In addition to the areas of decreased FC above, areas of increased FC were noted in some brain regions of POAG patients. Disappeared negative FC was noted between BA17 and brain regions of extra‐nuclear (BA25), right middle frontal gyrus (BA10), right middle cerebellar peduncle, and left cerebellum. Disappeared negative FC was also noted in POAG patients between higher visual cortices (BA18/19) and the right insular gyrus (BA48). Viewed alternatively, a decrease in negative FC in POAG is tantamount to an increase in positive FC in POAG patients relative to controls. Analogous results have been seen in Alzheimer disease with resting state fMRI, whereby areas of relatively increased positive FC are exhibited in what is otherwise a neurodegenerative condition [Zhang et al., 2009]. As has been postulated in Alzheimer disease, the changes in FC demonstrated in the above regions may relate to compensatory recruitment. The compensatory recruitment hypothesis applied to POAG would imply that certain brain areas exhibit diminished positive FC with the primary or higher visual cortices preferentially in POAG, whereas other regions affected later in the course of the disease (or possibly not at all) increase in positive FC with the visual cortices to compensate. For example, increased FC between BA17 and right middle cerebellar peduncle and left cerebellum, may relate to functional compensation secondary to the diminished FC of cerebellar area in POAG. An alternative explanation for the increased FC in the regions above may simply relate to a loss of inhibitory input to right cerebellar peduncle, the left cerebellum, extra‐nuclear gyrus, or right middle frontal gyrus, leading to an apparent loss in negative FC. With postretinal neural atrophy, decreased activity in the primary visual area may result in a loss of inhibition of these brain regions. What is more, FC between the medial frontal cortex—an important subsystem of the default mode network—and BA17 are increased in POAG. The medial temporal lobe provides information from prior experiences (in the forms of memories) to the medial frontal cortex, in which the information was used in mental stimulations involving the self [Buckner et al., 2008]. Increased FC between the primary visual area and the medial frontal cortices may thus relate to increased incorporation of visual information into such activities in POAG patients.

Study Strengths

FC analysis can effectively describe the pattern of interaction between brain regions and detect functional correlations between spatially separated brain regions. The principle strength of this study is in its controlled, age and gender‐matched design. The study was also prospective, and to the best knowledge of the authors, is the first such report in the English literature to illustrate the FC alterations in POAG patients, in distinction to prior studies which have focused only on structural changes. Although structural studies have demonstrated glaucoma‐related neuropathy throughout the visual network in POAG, including the visual cortex [Lam et al., 2003], brain volume measurements performed in the present work account for the effects of volume changes on FC, and thus, the changes described herein are more likely to be purely functional.

Study Weaknesses

In ROI analysis, the ROI region, which may be part of a RSN, is ignored. Theoretically, RSN is an integrated system, which contains multiple sub networks. Thus, errors could occur due to ignoring the entirety of a RSN when a specific seed point is defined to investigate or make assumptions about the spatial connectivity structure of a complete RSN. Furthermore, potential deviation of the ROI definition could result in the FC analysis detecting a smaller, more specific and overlapping subnetwork systems rather than the more significant larger one.

CONCLUSIONS

This study demonstrates, for the first time, using resting‐state fMRI, the FC changes of the visual cortices and visual pathway caused by loss of retinal ganglion cells in POAG [Gupta et al., 2006]. The communication between the primary visual area and higher visual cortices is decreased in POAG, as is positive FC between both of these areas and more remote regions of the brain. Areas of increased positive FC with visual cortices may relate to compensatory recruitment or diminished inhibitory input.