Decoding brain states using backward edge elimination and graph kernels in fMRI connectivity networks

In this paper we propose a new method to discriminate cognitive brain states directly from functional Magnetic Resonance Images (fMRI). First, we apply Robust Principal Component Analysis (RPCA) to construct low dimensional linear-subspace representations from the noisy fMRI images for each subject and then perform a Gaussian Naive Bayes (GNB) classification. In previous studies the discrimination of cognitive brain states from fMRI is done by transforming the fMRI into a time sequence of voxels from which the brain states are inferred. RPCA improved the classification rate of a real benchmark fMRI data.

Background The temporoparietal junction (TPJ) has been linked to lower-level attentional and higher-level social processing, both of which are affected in schizophrenia (SZ) and major depressive disorder (MDD). We examined resting functional connectivity of bilateral anterior and posterior TPJ in SZ and MDD to evaluate potential anomalies in each disorder and differences between disorders. Methods Resting-state functional magnetic resonance imaging data were acquired from 24 patients with SZ, 24 patients with MDD, and 24 age-matched healthy controls. We performed seed-based functional connectivity analyses with seed regions in bilateral anterior and posterior TPJ, covarying for gender and smoking. Results SZ had reduced connectivity versus controls between left anterior TPJ and dorsolateral prefrontal cortex (dlPFC) and posterior cingulate cortex (PCC); between left posterior TPJ and middle cingulate cortex, left dorsal PFC, and right lateral PFC; between right anterior TPJ and bilateral PCC; and between right posterior TPJ and middle cingulate cortex, left posterior insula, and right insula. MDD had reduced connectivity versus controls between left posterior TPJ and right dlPFC and between right posterior TPJ and PCC and dlPFC. SZ had reduced connectivity versus MDD between right posterior TPJ and left fusiform gyrus and right superior-posterior temporal cortex. Conclusion Functional connectivity to the TPJ was demonstrated to be disrupted in both SZ and MDD. However, TPJ connectivity may differ in these disorders with reduced connectivity in SZ versus MDD between TPJ and posterior brain regions.

Social cooperation often requires taking different roles in order to reach a shared goal. By defining individual tasks, these roles dictate processing demands of the collaborators. The main aim of the present study was to examine the hypothesis that induced alpha and lower beta oscillations provide insights into affective and cognitive brain states during social cooperation. Toward this end, an experimental game was used in which participants had to navigate a Pacman figure through a maze by sending and receiving information about the correct moving direction. Supporting our hypotheses, individual roles taken by the collaborators during gameplay were associated with significant changes in alpha and lower beta power. Furthermore, effects were similar when participants played the Pacman Game with human or computer partners. Findings are discussed from the perspective of the information-via-desynchronization hypothesis proposing that alpha and lower beta power decreases reflect states of enhanced cortical information representation. Overall, experimental games are a useful tool for extending basic research on brain oscillations to the domain of naturalistic social interaction as emphasized by the second-person neuroscience perspective.

To evaluate the feasibility of recruitment, preliminary efficacy, and acceptability of auricular percutaneous electrical nerve field stimulation (PENFS) for the treatment of fibromyalgia in veterans, using neuroimaging as an outcome measure and a biomarker of treatment response. Randomized, controlled, single-blind. Government hospital. Twenty-one veterans with fibromyalgia were randomized to standard therapy (ST) control or ST with auricular PENFS treatment. Participants received weekly visits with a pain practitioner over 4weeks. The PENFS group received reapplication of PENFS at each weekly visit. Resting-state functional connectivity magnetic resonance imaging (rs-fcMRI) data were collected within 2weeks prior to initiating treatment and 2weeks following the final treatment. Analysis of rs-fcMRI used a right posterior insula seed. Pain and function were assessed at baseline and at 2, 6, and 12weeks post-treatment. At 12weeks post-treatment, there was a nonsignificant trend toward improved pain scores and significant improvements in pain interference with sleep among the PENFS treatment group as compared with the ST controls. Neuroimaging data displayed increased connectivity to areas of the cerebellum and executive control networks in the PENFS group as compared with the ST control group following treatment. There was a trend toward improved pain and function among veterans with fibromyalgia in the ST + PENFS group as compared with the ST control group. Pain and functional outcomes correlated with altered rs-fcMRI network connectivity. Neuroimaging results differed between groups, suggesting an alternative underlying mechanism for PENFS analgesia.

The aim of this study is to investigate chemotherapy-induced alterations in the functional framework of the brain, and probe the relationship between these changes and executive function impairments in breast cancer patients. Thirty-three breast cancer patients (BC) after receiving chemotherapy and 31 matched healthy controls (HC) were enrolled in this study. All participants received resting-state functional connectivity magnetic resonance imaging (rs-fcMRI) and neuropsychological background tests. The lower functional connectivity of the posterior cingulate cortex (PCC) was found in the left postcentral gyrus, left precentral gyrus, right superior temporal gyrus, right cingulate gyrus and right middle frontal gyrus. A significant negative correlation was found between the response time on the Trail Making Tests and the functional connectivity strength between the PCC and right middle frontal and right cingulate gyri in breast cancer patients. In addition, the strength of the functional connectivity between the PCC and right middle frontal gyrus had a negative correlation with the response times on the Stroop Interference Test in breast cancer patients. This study demonstrated that BC patients after receiving chemotherapy have abnormal functional connectivity. These findings suggest that functional connectivity changes might play an important role in chemotherapy-induced executive function impairments in breast cancer patients.

Unprecedented 4D spatiotemporal infant regional cerebral blood flow framework and region-specific physiology–function coupling across infancy were elucidated, highlighting strong physiology–function coupling specifically at the default-mode network to meet extraneuronal metabolic demand for network emergence.

Different graph kernels may correspond to using different notions of similarity or may be using information coming from multiple sources. In this paper, we develop a common method to construct combined graph kernel (CGK) which is based on a family of graph kernels. We define three kinds of CGK. The first one is called the weighted combined graph kernel and is a parametric CGK. The second one is called the accuracy ratio weighted combined graph kernel and is a non-parametric CGK. The third one is called the product combined graph kernel and also belongs to non-parametric CGK. The three kinds of definition of CGK can be applied for constructing CGK based on a family of graph kernels. This family of kernels is demonstrated based on the Weisfeiler–Lehman (WL) sequence of graphs in this paper, including a highly efficient subtree kernel, edge kernel, and shortest path kernel. Experiments demonstrate that our CGK based on WL graph kernels outperforms the corresponding single WL graph kernel on several classification benchmark data sets.

Individual differences in how the brain responds to novelty are present from infancy. A common method of studying novelty processing is through event-related potentials (ERPs). While ERPs possess millisecond precision, spatial resolution remains poor, especially in infancy. This study aimed to balance spatial and temporal precision by combining ERP data with functional magnetic resonance imaging (fMRI) data. Twenty-nine infants (15 female) underwent resting-state fMRI (Mage = 4.73 months) and electroencephalography (EEG) during a three-stimulus auditory oddball task (Mage = 5.19 months). The mismatch response (MMR) and P3 were computed from ERP data, and resting-state functional connectivity (rs-FC) was computed from fMRI data. We first source localized the MMR and P3 responses to five regions-of-interest (ROIs), based on prior literature. We then performed network-level enrichment analyses to identify associations between rs-FC and MMR and P3, at each of the five ROIs. In line with prior work, source-localized EEG analyses implicated the bilateral auditory cortices, posterior cingulate cortex, and superior parietal cortex in the generation of MMR and P3 responses. The MMR and P3 related to functional connectivity within the somatomotor network as well as between the somatomotor and the dorsal and ventral attention networks (DAN/VAN). This was especially true for novelty response ERPs recorded at superior parietal lobule, known for its implications in initial reorienting to novel stimuli. The DAN, known for its implication in initial reorienting to support novelty detection, was implicated for the MMR. In contrast, the VAN, known for its support of later-stage, complex adjustments in attention, related to the later P3. This work further solidifies our understanding of the underlying networks implicated in the development of immediate responses to stimuli. Altered configurations of such networks may increase the risk for heightened sensitivity to novelty in certain individuals, which could have behavioral and clinical significance. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Multivariate classification techniques have been widely applied to decode brain states using functional magnetic resonance imaging (fMRI). Due to variabilities in fMRI data and the limitation of the collection of human fMRI data, it is not easy to train an efficient and robust supervised-learning classifier for fMRI data. Among various classification techniques, sparse representation classifier (SRC) exhibits a state-of-the-art classification performance in image classification. However, SRC has rarely been applied to fMRI-based decoding. This study aimed to improve SRC using unlabeled testing samples to allow it to be effectively applied to fMRI-based decoding. We proposed a semisupervised-learning SRC with an average coefficient (semiSRC-AVE) method that performed the classification using the average coefficient of each class instead of the reconstruction error and selectively updated the training dataset using new testing data with high confidence to improve the performance of SRC. Simulated and real fMRI experiments were performed to investigate the feasibility and robustness of semiSRC-AVE. The results of the simulated and real fMRI experiments showed that semiSRC-AVE significantly outperformed supervised learning SRC with an average coefficient (SRC-AVE) method and showed better performance than the other three semisupervised learning methods.

Diagnostic delay in functional seizures is associated with abnormal processing of facial emotions

To evaluate brain changes after cognitive rehabilitation in patients with clinically stable relapsing-remitting (RR) multiple sclerosis (MS) by using neuropsychologic assessment and structural and functional magnetic resonance (MR) imaging techniques. The study was conducted with approval of the involved institutional review boards. Written informed consent was obtained from each participant. Twenty patients with RR MS and cognitive deficits at baseline were randomly assigned to undergo treatment (n = 10), which entailed computer-assisted cognitive rehabilitation of attention and information processing and executive functions, or to serve as a control subjects (n = 10) without cognitive rehabilitation. All patients underwent a standardized neuropsychologic assessment and MR imaging at baseline and after 12 weeks. Changes in gray matter (GM) volumes on three-dimensional T1-weighted images and changes in normal-appearing white matter (NAWM) architecture on diffusion-weighted images were assessed. Changes in functional activity at functional MR imaging during the Stroop task and at rest were also investigated by using linear models. As compared with their performance at baseline, the patients in the treatment group improved at tests of attention and information processing and executive functions. Neither structural modifications to GM volume nor modifications to NAWM architecture were detected at follow-up in both groups. Functional MR imaging demonstrated modifications of the activity of the posterior cingulate cortex (PCC)/precuneus and dorsolateral prefrontal cortex (PFC) during the Stroop task, as well as modifications of the activity of the anterior cingulum, PCC and/or precuneus, left dorsolateral PFC, and right inferior parietal lobule at rest in the treatment group compared with the control group. In the treatment group, functional MR imaging changes were correlated with cognitive improvement (P < .0001 to .01). Rehabilitation of attention and information processing and executive functions in RR MS may be effected through enhanced recruitment of brain networks subserving the trained functions.

Lateralized brain regions subserve functions such as language and visuospatial processing. It has been conjectured that individuals may be left-brain dominant or right-brain dominant based on personality and cognitive style, but neuroimaging data has not provided clear evidence whether such phenotypic differences in the strength of left-dominant or right-dominant networks exist. We evaluated whether strongly lateralized connections covaried within the same individuals. Data were analyzed from publicly available resting state scans for 1011 individuals between the ages of 7 and 29. For each subject, functional lateralization was measured for each pair of 7266 regions covering the gray matter at 5-mm resolution as a difference in correlation before and after inverting images across the midsagittal plane. The difference in gray matter density between homotopic coordinates was used as a regressor to reduce the effect of structural asymmetries on functional lateralization. Nine left- and 11 right-lateralized hubs were identified as peaks in the degree map from the graph of significantly lateralized connections. The left-lateralized hubs included regions from the default mode network (medial prefrontal cortex, posterior cingulate cortex, and temporoparietal junction) and language regions (e.g., Broca Area and Wernicke Area), whereas the right-lateralized hubs included regions from the attention control network (e.g., lateral intraparietal sulcus, anterior insula, area MT, and frontal eye fields). Left- and right-lateralized hubs formed two separable networks of mutually lateralized regions. Connections involving only left- or only right-lateralized hubs showed positive correlation across subjects, but only for connections sharing a node. Lateralization of brain connections appears to be a local rather than global property of brain networks, and our data are not consistent with a whole-brain phenotype of greater “left-brained” or greater “right-brained” network strength across individuals. Small increases in lateralization with age were seen, but no differences in gender were observed.

Romantic love is a motivational state associated with a desire to enter or maintain a close relationship with a specific other person. Functional magnetic resonance imaging (fMRI) studies have found activation increases in brain regions involved in the processing of reward, motivation and emotion regulation, when romantic lovers view photographs of their partners. However, not much is known about whether romantic love affects the brain’s functional architecture during rest. In the present study, resting state functional magnetic resonance imaging (rsfMRI) data was collected to compare the regional homogeneity (ReHo) and functional connectivity (FC) across an “in-love” group (LG, N = 34, currently intensely in love), an “ended-love” group (ELG, N = 34, ended romantic relationship recently), and a “single” group (SG, N = 32, never fallen in love). Results show that: (1) ReHo of the left dorsal anterior cingulate cortex (dACC) was significantly increased in the LG (in comparison to the ELG and the SG); (2) ReHo of the left dACC was positively correlated with length of time in love in the LG, and negatively correlated with the lovelorn duration since breakup in the ELG; (3) FC within the reward, motivation, and emotion regulation network (dACC, insula, caudate, amygdala, and nucleus accumbens) as well as FC in the social cognition network [temporo-parietal junction (TPJ), posterior cingulate cortex (PCC), medial prefrontal cortex (MPFC), inferior parietal, precuneus, and temporal lobe] was significantly increased in the LG (in comparison to the ELG and SG); (4) in most regions within both networks FC was positively correlated with the duration of love in the LG but negatively correlated with the lovelorn duration of time since breakup in the ELG. This study provides first empirical evidence of love-related alterations in brain functional architecture. Furthermore, the results shed light on the underlying neural mechanisms of romantic love, and demonstrate the possibility of applying a resting-state fMRI approach for investigating romantic love.

To prospectively identify brain regions in which task-related changes in activation during a memory encoding task, measured with functional magnetic resonance (MR) imaging, correlate with degree of memory impairment across Alzheimer disease (AD), mild cognitive impairment (MCI), and elderly control subjects. The institutional review board approved this HIPAA-compliant study, and each patient gave written informed consent. Seventy-five subjects (mean age, 72.9 years+/-7.2 [standard deviation]; 37 men, 38 women)-13 patients with mild AD, 34 individuals with amnestic MCI, and 28 healthy elderly control subjects-were imaged at 4.0 T during novel encoding (NE) and familiar encoding (FE) of face-name pairs presented within a block design for later retrieval. Blood oxygen level-dependent (BOLD) changes were assessed across the entire brain for each group. Between-subject analysis identified brain regions demonstrating a monotonic increase or decrease in activation magnitude, from control subjects to patients with MCI to patients with mild AD. BOLD response was also correlated with score on the delayed portion of the California Verbal Learning Test (CVLT). In controls, the task elicited positive activation (NE>FE) in the dorsolateral prefrontal, lateral parietal, and medial temporal regions, and negative activation (FE>NE) in the midline frontal and parietal regions. Along the spectrum from control subjects to patients with AD, there was decreasing activation in the medial temporal lobe (MTL), including the hippocampus and parahippocampal and fusiform gyri, and increasing activation in the posteromedial cortices (PMCs), primarily in the precuneus and posterior cingulate gyrus. Activation magnitude in the PMCs significantly (P<.001, r=-0.502) correlated with CVLT score. Compared with activation in the MTL, deactivation in the PMCs could be a more sensitive marker of early AD at functional MR imaging.

Despite much research on the function of the insular cortex, few studies have investigated functional subdivisions of the insula in humans. The present study used resting-state functional connectivity magnetic resonance imaging (MRI) to parcellate the human insular lobe based on clustering of functional connectivity patterns. Connectivity maps were computed for each voxel in the insula based on resting-state functional MRI (fMRI) data and segregated using cluster analysis. We identified 3 insular subregions with distinct patterns of connectivity: a posterior region, functionally connected with primary and secondary somatomotor cortices; a dorsal anterior to middle region, connected with dorsal anterior cingulate cortex, along with other regions of a previously described control network; and a ventral anterior region, primarily connected with pregenual anterior cingulate cortex. Applying these regions to a separate task data set, we found that dorsal and ventral anterior insula responded selectively to disgusting images, while posterior insula did not. These results demonstrate that clustering of connectivity patterns can be used to subdivide cerebral cortex into anatomically and functionally meaningful subregions; the insular regions identified here should be useful in future investigations on the function of the insula.