The neural pathways and genetic substrates of non-suicidal self-injury as a "sensation of pain" addiction in drug-naïve depressed adolescents.

Suanzaoren decoction improving chronic insomnia with little effects on functional connectivity within the sensorimotor network.

3D masked autoencoder with spatiotemporal transformer for modeling of 4D fMRI data.

Social anxiety refers to excessive fear of social situations and is then accompanied by social avoidance behaviours. While the neural mechanisms of social anxiety disorder in clinical populations have been widely investigated, the functional connectivity underlying social anxiety in the nonclinical population remains poorly understood. The present study addressed this gap by employing connectome-based predictive modelling (CPM) to identify resting-state functional connectivity associated with social anxiety in healthy college students. Our findings revealed a social anxiety connectome that contributed to predicting individuals' social anxiety, which mainly includes the connections within the default mode network (DMN) (i.e., positive network) and those between the frontal parietal network (FPN) and visual network (i.e., negative network). Importantly, the robustness and specificity of this connectome were validated by using different brain atlases and cross-validation schemes and controlling for the influence of general anxiety. Moreover, two sub-dimensions of social anxiety, i.e., social distress and social avoidance, showed distinct neural correlates, with social distress correlated with the positive network and social avoidance with the negative network. Together, these findings provide novel insights into the neural basis of social anxiety in nonclinical populations, highlighting specific functional connectivity associated with different facets of social anxiety.

Hierarchy is a fundamental principle of network organization in the human brain. Functional gradient introduces a new perspective in identifying hierarchy alterations by capturing major axes of functional connectivity (FC) in lowdimensional space. However, whether this gradient structure is disrupted in mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes (MELAS) patients with stroke-like episodes (SLE) and how this disruption modulated by gene expression profiles remain unknow. Thirty-one MELAS patients at acute stage (MELAS-acute) and 31 healthy controls (HC) underwent restingstate functional magnetic resonance imaging (rs-fMRI) scan. Based on the whole-brain voxel-wise FC patterns, functional gradient values were generated and group-averaged gradient values were further extracted and compared from global to voxel level. Combined with the Allen Human Brain Atlas, we then assessed the spatial correlations between MELAS-related gradient alterations and gene expression profiles. Relative to the HC, MELAS-acute patients exhibited global alterations in the principal gradient, including reduced gradient range and gradient variation. In addition, patients showed lower gradient values in the default mode network (DMN) but higher values in the ventral attention network (VAN) and sensorimotor network (SMN) at network and voxel level. Furthermore, we established a link between MELAS-acute related principal gradient and gene expression profiles, with two gene sets mainly enriched in mitochondrion, neuron, glutamatergic synapse, and ATPase activity. These results highlight the connectome gradient alterations in MELAS patients at acute stage and its linkage with gene expression profiles, providing insight into the neurobiological basis of functional alterations during the acute SLE stage in MELAS. AHBA = Allen Human Brain Atlas; DMN = default mode network; FPN = frontoparietal network; GO = Gene ontology; HC = healthy controls; MELAS = mitochondrial encephalomyopathy, lactic acidosis, and stroke-like episodes; MELAS-acute = MELAS patients at acute stage; MNI = Montreal Neurological Institute; SLE = stroke-like episodes; SMN = sensorimotor network; VAN = ventral attention network; VN = visual network.

Oligodendroglial pathology is a hallmark of multiple system atrophy (MSA), yet it remains unclear whether MRI-detected atrophy reflects underlying biological mechanisms. This study investigated whether regional atrophy aligns with gene expression and neurotransmitter systems. We recruited 65 MSA patients and derived brain atrophy measures from T1-weighted MRIs. Using postmortem data from the Allen Human Brain Atlas, partial least squares (PLS) regression identified gene expression components associated with brain atrophy. Gene enrichment analyses explored biological processes, and annotation mapping identified neurotransmitter systems matching atrophy patterns. Specificity was tested against 57 patients with Parkinson's disease. Atrophy primarily affected the cerebellar white matter, pons, putamen, olive, and substantia nigra. PLS revealed two latent variables explaining 27.5% of the covariance. Atrophic regions overexpressed genes linked to mitochondrial function and oligodendrocytes, showing patterns distinct from Parkinson's disease. These regions also exhibited lower serotonin and GABA levels, and higher acetylcholine and noradrenaline receptor densities. MRI-derived atrophy in MSA is biologically grounded and may inform future therapeutic studies.

BackgroundProtein deposition and neurodegeneration differently affect the brain spatially and temporally in Alzheimer's disease (AD). Here we used imaging transcriptomics to understand the biological and molecular properties underlying regional variability of neuroimaging phenotypes of amyloid, tau, and neurodegeneration assessed by PET and MRI.MethodBrain patterns were estimated by contrasting imaging data between AD patients and healthy controls from two independent cohorts for replication (Geneva Memory Clinic and ADNI). Regional gene expression profiles were derived from brain‐wide microarray measurements provided by the Allen human brain atlas (AHBA). Hypothesis‐driven analyses assessed the spatial association between neuroimaging patterns and gene expression (gene‐to‐biomarker associations) for selected candidate genes for AD. Over‐representation analysis (ORA) and gene set enrichment analysis (GSEA) were used to characterize molecular properties and biological pathways of genome‐wide gene sets associated with regional AD pathologies in a data‐driven manner.ResultRegional patterns showed the highest amyloid load in frontal, parietal, and lateral temporal lobes, whereas tau deposition was most pronounced in medial temporal lobes and lateral temporoparietal areas. Neurodegeneration patterns were instead less widespread, involving mainly temporoparietal areas. Specific patterns of amyloid, tau and neurodegeneration were differently associated with AD‐related genes. ORA and GSEA revealed that genes implicated in different aspects of protein synthesis (e.g. cytosolic ribosome, mitochondrion organization, and RNA metabolic processes) as well as immune regulation and neuroinflammation correlated exclusively with amyloid load, whereas genes involved in the synaptic organization, transmission, and function were associated to the severity of amyloid, tau, and neurodegeneration pathologies. GSEA confirmed that the gene‐to‐tau and gene‐to‐atrophy associations were related to similar biological pathways involving synaptic signaling and organization, while gene‐to‐hypometabolism associations were more related to cellular processes.ConclusionSelective AD vulnerabilities were differently related to specific gene expression and molecular‐biological properties, with a large set of genes associated with amyloid accumulation and a subset of genes conferring additional vulnerability to downstream tau. Our findings suggest that the spatial and temporal decoupling between amyloid deposition, tau deposition and neurodegeneration is explained by differential genetic expression but that shared mechanisms link upstream amyloid with subsequent tau pathology and loss of neuronal integrity.

The mechanism by which genetic risk leads to cortical vulnerability in drug-resistant epilepsy (DRE) remains unclear. This study used 7T structural and resting-state functional MRI to investigate cortical neural activity alterations in 105 DRE patients and 105 healthy controls (HCs), and to explore related genetic mechanisms. Vertex-wise analyses of mean amplitude of low-frequency fluctuation (mALFF) and regional homogeneity (ReHo) revealed that DRE patients primarily exhibited decreased mALFF and increased ReHo in the Cingulo-Opercular Network. Using the Allen Human Brain Atlas, we conducted spatial transcriptomic analysis via partial least squares (PLS) and gene enrichment analysis to identify gene categories associated with these functional changes. The results showed that cortical alterations were related to epilepsy-general genes (e.g., TMEM74, KCNN2, RBFOX1) and brain-relevant genes. Genes positively correlated with mALFF alterations enriched in mitochondrial inner membrane, matrix, and carboxylic acid metabolism; negatively in chromatin remodeling, binding, and postsynapse. Genes positively correlated with ReHo alterations enriched in nucleic acid-related catalytic activity, ribonucleoprotein granule, and centrosome; negatively in amyotrophic lateral sclerosis, mitochondrial membrane, and pyrophosphatase activity. These findings link spatial brain activity abnormalities in DRE to specific genetic signatures and biological pathways, suggesting new mechanistic insights and potential therapeutic targets for this difficult-to-treat condition.

Structure-function coupling alterations in adolescent depression correlate with neurotransmitter systems and cell-type-specific transcriptomics.

Structural brain alterations in bipolar disorder (BD) have been widely reported, yet the hierarchical organization of cortical morphometric networks and their molecular and cognitive underpinnings remain unclear. We applied the morphometric inverse divergence (MIND) network approach to structural MRI data from 49 BD patients and 119 healthy controls. Principal MIND gradients were derived using diffusion map embedding, followed by multiscale analyses linking gradient alterations to neurotransmitter systems, cognitive-behavioral domains, and transcriptomic profiles from the Allen Human Brain Atlas. Validation was performed in three independent, cross-scanner, cross-race, and cross-age validation datasets. Bipolar disorder patients showed significant principal gradient alterations in the left rostral middle frontal and lateral occipital cortices, with network-level decreases in the ventral attention and motor networks and increases in frontoparietal and visual networks. Gradient alterations spatially correlated with acetylcholine (VAChT) and GABA (GABAA/BZ) systems, and were associated with cognitive processes involving executive control and visual attention. Transcriptomic analyses identified gene sets enriched for BD-related GWAS loci, expressed predominantly in excitatory and inhibitory neurons, astrocytes, and oligodendrocytes, with preferential enrichment in cortical layers III-IV and developmental windows spanning early fetal to young adulthood. These findings reveal disrupted hierarchical cortical organization in BD and link macroscale morphometric alterations to specific neurotransmitter systems and transcriptional architectures. The MIND gradient emerges as a potential biomarker bridging structural disruptions with molecular and cognitive mechanisms in BD.

Increasing evidence indicates that catecholaminergic degeneration, especially in the ventral tegmental area (VTA) and locus coeruleus (LC), precedes classical Alzheimer’s disease (AD) pathology, a finding confirmed by structural and functional imaging in patient cohorts with amnestic MCI and AD. Given the central role of these nuclei in shaping motivation, arousal and memory, we developed a dedicated pipeline to reconstruct the catecholaminergic connectome at the whole-brain level. Tg2576 mice overexpressing human APP695 with the Swedish mutation were employed. Brains were cleared, immunolabeled for tyrosine hydroxylase (TH), and imaged using volumetric light-sheet microscopy. Tiles were stitched with BigStitcher, and Arivis Pro U-Net models segmented soma, axon hillocks, dendrites, and nuclei. 3D reconstruction and automated tracing were performed with Vaa3D APP2. Reconstructions were registered to the Allen Brain Atlas. NetworkX was used for data analysis, and in MATLAB the physiological laws of CNS impulse transmission were applied to the networks to perform simulations. The pipeline generated a whole-brain catecholaminergic connectomic model for Tg2576 and controls. A decrease in TH+ neuron counts was observed in both VTA and LC, leading to marked denervation of the hippocampus, amygdala, medial prefrontal cortex, and the entire limbic lobe. Alterations were also noted within and in between catecholaminergic nuclei, with changes in their intrinsic functional circuit units, modifications of dendritic arborizations, and a reduction in reverberant circuits that normally sustain continuous output activity. conclusions: This represents the first connectomic model of catecholaminergic architecture in both healthy and AD brains. TH+ fibers normally synapse onto GABAergic neurons; their denervation drives excitotoxicity and degeneration in target areas. These alterations may underlie prodromal psychiatric symptoms and, through hippocampal and cortical denervation, may be the cause of pathological alterations and subsequent cognitive decline.

Abstract Short-range association fibers (SAFs) are critical for cortical communications but are often underestimated in conventional resolution diffusion magnetic resonance imaging (dMRI) since they locate within a ~1.5 mm thin layer of superficial white matter. With the advent of high-resolution diffusion imaging techniques, this study evaluated the effects of image spatial resolution on SAF reconstruction using two datasets: (1) prospectively acquired dMRI data from 20 healthy subjects, each scanned at three resolutions (i.e., 2, 1.5, and 0.96 mm iso.), and (2) retrospectively down-sampled dMRI data from the Human Connectome Project dataset, as well as 20 representative MRtrix3-based tractography pipelines. It was found that lower resolution degraded superficial white matter model fitting, lowered the SAF streamline counts, and reduced global and regional short-range connectivity fraction (SCF), defined as the fraction of SAF connections among all association fiber connections, across all tested methods. Temporal lobe cortical regions exhibited the greatest SCF declines at lower resolutions. Tractography methods differed in resolution sensitivity, with diffusion tensor imaging (DTI)-based single-tissue single-fiber tractography showing greater decreases in SCF than constrained spherical deconvolution (CSD)-based multi-tissue multi-fiber tractography at lower resolutions. Probabilistic, anatomically-constrained tractography combined with spherical-deconvolution informed filtering of tractograms was more robust to decreases in resolution. Up-sampling to a nominally higher resolution partially improved model fitting and SCF accuracy across the evaluated pipelines, with the greatest effect observed for DTI. Using the 0.96 mm iso. gSlider data and optimized tractography pipelines from this study, we constructed the first human brain atlas of RSCF. In summary, this study provides a systematic and quantitative evaluation using MRtrix3 of how spatial resolution, fiber models, and tracking methodologies affect SAF reconstruction and structural connectivity estimation, serving as a reference framework for methodological choices. These advances may enhance the characterization of both healthy and diseased human brains across a wide range of neuroscientific and clinical applications.

Altered homotopic functional connectivity in primary angle-closure glaucoma (PACG) and their underlying molecular mechanisms remain poorly understood. In our study, we investigated voxel-mirrored homotopic connectivity (VMHC) alterations in patients with PACG and the molecular mechanisms of VMHC. In this study, we investigated alterations in VMHC among 47 patients with PACG and 45 matched healthy controls. We then integrated these spatial patterns with cortical transcriptomic data from the Allen Human Brain Atlas using partial least squares (PLS) regression to identify gene expression profiles associated with VMHC alterations. In this study, we identified widespread reductions in interhemispheric functional connectivity in patients with PACG using VMHC analysis. Multivariate spatial correlation with gene expression data revealed that VMHC alterations were significantly associated with a distinct transcriptional signature captured by the first PLS component. Functional enrichment of these genes indicated downregulation of pathways related to synaptic and metabolic maintenance, and upregulation of immune, stress, and chromatin regulatory processes. Cell-type analysis showed that astrocytes and endothelial cells were selectively enriched with VMHC-related genes, reflecting glial and vascular involvement. Moreover, spatial alignment with neurotransmitter receptor maps highlighted significant associations with serotonergic, dopaminergic, GABAergic, cholinergic, and opioidergic pathways, suggesting a neuromodulatory basis for VMHC disruption. Together, these findings suggest that interhemispheric dysconnectivity in PACG is not only a reflection of functional brain changes but is also grounded in molecular and cellular mechanisms. This integrative approach advances our understanding of PACG as a brain-wide neurodegenerative condition and offers new perspectives for targeting glial, vascular, and neuromodulatory pathways in future therapeutic interventions.

ObjectiveThis study seeks to identify brain regions with atypical neural connectivity in individuals suffering from arthritis-related chronic pain, compared to healthy controls, using resting-state functional magnetic resonance imaging (rs-fMRI).MethodsA seed-based connectivity analysis was conducted between the known pain-related regions of interest (ROIs), derived from the MNI (n = 76) and the Automated Anatomical Labeling (AAL) whole brain atlas (n = 116). We examined the connectivity differences in a cohort of 56 osteoarthritis patients and 20 healthy controls. Connectivity matrices were compared using permutation tests corrected for multiple comparisons, identifying statistically significant differences (p < 0.05). Subsequent network analysis resulted in hub scores, identifying the most central and influential brain regions within the altered connectivity network in patients experiencing pain.ResultsThe most significant atypical neural connections in osteoarthritis patients were identified in the cingulate gyrus, insula, inferior parietal lobe, and thalamus, with notable involvement of the occipital lobe, postcentral gyrus, inferior frontal gyrus, orbitofrontal cortex, temporal lobe, hippocampus, and basal ganglia. The thalamus, cingulate gyrus, and insula emerged as key hubs in the chronic pain network, reflecting disrupted sensory, emotional, and cognitive pain processing. No significant connectivity differences were found in the brainstem, cerebellum, superior parietal lobe, precentral gyrus, superior and middle frontal gyri, or amygdala.ConclusionOur data-driven approach reveals specific neural connectivity disruptions in OA, highlighting connections between the cingulate gyrus, temporal lobe, and thalamus. These findings identify specific network disruptions in OA-related pain, offering insight into altered brain connectivity and potential avenues for targeted interventions.Supplementary informationThe online version contains supplementary material available at 10.1186/s12880-025-02009-0.

To investigate hemodynamic and blood oxygen metabolism and their associations with disease progression, dopaminergic transporter (DAT) activity, and glucose uptake in patients with Parkinson's disease (PD). This cross-sectional study included 73 patients with PD (mean age: 61.10 years) and 67 healthy controls (mean age: 58.99 years). Oxygen metabolism parameters-deoxygenated hemoglobin (Cdeoxy), oxygen extraction fraction (OEFrel), deoxygenated cerebral blood volume (dCBV), and R2* were measured using qMRI. DAT availability and glucose metabolism were assessed using PET with [18F]FP-CIT and [18F]FDG, respectively. Regional analyses were conducted using standardized brain atlases. Compared with the controls patients with PD exhibited elevated Cdeoxy, OEFrel, and R2* in the substantia nigra, whereas Cdeoxy and dCBV levels were reduced in the bilateral caudate nucleus and frontal cortex (p < 0.05). The Hoehn-Yahr (H-Y) 2.5-3 subgroup exhibited higher levels of Cdeoxy and OEFrel in the left putamen than the H-Y 1-2 subgroup (p < 0.05). In the H-Y 1-2 subgroup, Cdeoxy, OEFrel, and R2* correlated with UPDRS scores in the substantia nigra and red nucleus (p < 0.05). In advanced stages (H-Y stages 2.5-3), significant correlations were observed in the striatal structures/the left dorsolateral putamen/posterior right caudate (p < 0.05). OEFrel and R2* values were positively correlated with glucose metabolism in the left putamen and right caudate. (p < 0.05). qMRI demonstrated alterations in hemodynamics and oxygen metabolism in patients with PD, particularly within the nigrostriatal system, suggesting that metabolic indicators could serve as supplementary biomarkers for diagnosing and monitoring the progression of PD.

The granin gene family of neuropeptides functions as peptide neurotransmitters in the brain for the regulation of neural functions that regulate behaviors. Granins are involved in regulating cognition, memory, depression, aggression, stress, energy expenditure, inflammation, and related. Development of the human brain involves formation of synapses and their spectrum of neurotransmitters to establish neural connections that are required for brain functions. Therefore, the goal of this study was to analyze the gene expression profiles of the granin neurotransmitter genes during human brain development at prenatal, infancy, childhood, adolescence, and adult stages. Granin gene expression in brain development was assessed by quantitative RNA sequencing data from the Allen Human Brain Atlas resource. VGF (neurosecretory protein VGF) expression was significantly increased during development during the prenatal to childhood through adult stages in the anterior cingulate cortex, dorsolateral prefrontal cortex, inferolateral temporal cortex, orbital frontal cortex, posteroventral parietal cortex, primary somatosensory cortex, and primary visual cortex regions. SCG2 (secretogranin 2) expression was also significantly increased from prenatal to infancy through adult stages in anterior cingulate cortex, dorsolateral prefrontal cortex, inferolateral temporal cortex, orbital frontal cortex, posterior superior temporal cortex, posteroventral parietal cortex, primary somatosensory cortex, and primary visual cortex. A modest number of brain regions showed increased CHGA, CHGB, and SCG3 expression in the postnatal periods compared to the prenatal periods. Further, the SCG5, PCSK1N, and GNAS genes displayed minimal changes throughout development. Overall, these results demonstrate developmental upregulation of VGF and SCG2 genes, with lesser upregulation of CHGA, CHGB, and SCG3 genes, and almost no changes in SCG5, PCSK1N, and GNAS genes during development. These findings illustrate the differential regulation of granin genes during human brain development.

BackgroundProcessing of large‐scale multimodal MRI and PET data is crucial for advancing Alzheimer's disease (AD) neuroimaging research. However, image processing requires strong expertise on different software packages (SPM/FSL/AFNI/FreeSurfer) and programming languages (R/MATLAB/Python/Bash) and lab‐specific image processing approaches are a major roadblock for harmonization across sites. Therefore, establishing uniform and user‐friendly image processing workflows is crucial for inter‐site standardization and harmonization of neuroimaging data and to reduce bias introduced by different preprocessing strategies. Therefore, we developed the containerized, state‐of‐the‐art, fully automated, neuroimaging toolbox ADprep that integrates robust preprocessing of structural/functional MRI, and multi‐tracer PET (amyloid/tau/FDG/TSPO), generating standardized nifti and atlas‐based spreadsheet outputs across a broad range of brain atlases. ADprep requires no programming expertise and can facilitate harmonized neuroimaging analyses and data sharing across the AD neuroimaging community and will be fully integrated into the cloud‐based GRIP platform.MethodsADprep works on bids‐formatted data and was fully developed in nipype (Figure 1). Preprocessing for structural MRI includes volumetric and cortical thickness assessments for widely used brain atlases (Desikan‐Killiany/Schaefer100‐600/LPBA/Hammers/Neuromorphometrics/Cobra/Destrieux), plus spatially normalized and smoothed tissue segments for voxel‐based morphometry analyses. Functional MRI processing includes slice‐timing and motion correction, nuisance regression, spatial normalization, and functional connectivity assessments for above‐ mentioned atlases. PET processing includes generation of spatially normalized SUVR images for different tracer‐specific reference regions, as well as extraction of atlas‐based SUVRs and partial‐volume correction. Standardized outputs in FreeSurfer space were benchmarked against data from the ADNI imaging core to illustrate comparability with existing pipelines. Cloud‐based GRIP and local cluster implementation is provided to ensure large‐scale data processing.ResultsADprep was tested successfully on large‐scale multimodal datasets, including several thousand scans from ADNI, ADNI‐DOD, and A4 with an overall processing failure rate of <4%. Using data from the ADNI PET core for benchmarking, ADprep closely reproduces openly available amyloid‐PET (r=0.99, p <0.001, Figure 2A) and tau‐PET SUVRs (r=0.98, p <0.001, Figure 2B). Runtime is ∼1h for a structural/functional MRI and ∼30min for a PET scan.ConclusionsADprep is user‐friendly and harmonized multimodal neuroimaging pipeline, that can be applied to different neuroimaging datasets by non‐expert users, providing outputs that can be directly used for statistical analyses.

BackgroundProcessing of large‐scale multimodal MRI and PET data is crucial for advancing Alzheimer’s disease (AD) neuroimaging research. However, image processing requires strong expertise on different software packages (SPM/FSL/AFNI/FreeSurfer) and programming languages (R/MATLAB/Python/Bash) and lab‐specific image processing approaches are a major roadblock for harmonization across sites. Therefore, establishing uniform and user‐friendly image processing workflows is crucial for inter‐site standardization and harmonization of neuroimaging data and to reduce bias introduced by different preprocessing strategies. Therefore, we developed the containerized, state‐of‐the‐art, fully automated, neuroimaging toolbox ADprep that integrates robust preprocessing of structural/functional MRI, and multi‐tracer PET (amyloid/tau/FDG/TSPO), generating standardized nifti and atlas‐based spreadsheet outputs across a broad range of brain atlases. ADprep requires no programming expertise and can facilitate harmonized neuroimaging analyses and data sharing across the AD neuroimaging community and will be fully integrated into the cloud‐based GRIP platform.MethodsADprep works on bids‐formatted data and was fully developed in nipype (Figure 1). Preprocessing for structural MRI includes volumetric and cortical thickness assessments for widely used brain atlases (Desikan‐Killiany/Schaefer100‐600/LPBA/Hammers/Neuromorphometrics/Cobra/Destrieux), plus spatially normalized and smoothed tissue segments for voxel‐based morphometry analyses. Functional MRI processing includes slice‐timing and motion correction, nuisance regression, spatial normalization, and functional connectivity assessments for above‐ mentioned atlases. PET processing includes generation of spatially normalized SUVR images for different tracer‐specific reference regions, as well as extraction of atlas‐based SUVRs and partial‐volume correction. Standardized outputs in FreeSurfer space were benchmarked against data from the ADNI imaging core to illustrate comparability with existing pipelines. Cloud‐based GRIP and local cluster implementation is provided to ensure large‐scale data processing.ResultsADprep was tested successfully on large‐scale multimodal datasets, including several thousand scans from ADNI, ADNI‐DOD, and A4 with an overall processing failure rate of <4%. Using data from the ADNI PET core for benchmarking, ADprep closely reproduces openly available amyloid‐PET (r=0.99, p <0.001, Figure 2A) and tau‐PET SUVRs (r=0.98, p <0.001, Figure 2B). Runtime is ∼1h for a structural/functional MRI and ∼30min for a PET scan.ConclusionsADprep is user‐friendly and harmonized multimodal neuroimaging pipeline, that can be applied to different neuroimaging datasets by non‐expert users, providing outputs that can be directly used for statistical analyses.

Advances in brain imaging genetics have propelled precision medicine by generating imaging-derived phenotypes linked to genetic variants, particularly in polygenic neurodegenerative diseases like Alzheimer's Disease (AD), where genetic variants exert pleiotropic effects across the genome. Using a representation learning framework, we provide interpretable summaries of brain structural networks guided by brain MRI and genetic variations. We applied a combination of two generative models-Autoencoder and Variational Autoencoder-to identify genetically influenced brain structural patterns from MRI data. This approach partitions the brain into anatomical brain elements (ABEs) that correlate with genetic variations, including disease-related ones, in a data-driven manner (Figure 1). The model was applied to clinical data from healthy controls, AD, and Mild Cognitive Impairment (MCI) patients (N=1564) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. Using voxel-wise brain volumetric measurements and 54 AD-related Single Nucleotide Polymorphisms (SNPs), we extracted genetic-correlated ABEs. The model robustly detected specific imaging and genetic patterns and elucidated their associations. When applied to ADNI data, it grouped brain regions based on structural and genetic covariance, creating a novel brain parcellation distinct from those based solely on structural covariance. Six genetically informed ABEs were extracted (Figure 2A). Notably, the subcortical regions and orbitofrontal cortex were clustered together, distinct from the rest of the frontal lobe. Most cerebellar regions were grouped together, except the inferior posterior lobe of the cerebellar vermis and hemisphere. These distinctions were absent when genetic information was excluded. We also pinpointed specific SNPs that were associated with each ABE. For instance, rs429358 (APOE) and rs41289512 (NECTIN2) were associated with temporal and hippocampal ABEs, aligning with existing literature and several ABCA7 polymorphisms (rs4147929, rs3752246, rs111278892) and CASS4 polymorphisms (rs7274581, rs6014724, rs6024870) were clustered together in association with specific ABEs respectively (Figure 2B). These insights highlight a novel brain atlas that captures the complexity of genetic and neuroanatomical heterogeneity, refining our understanding of how genetic factors influence brain anatomy.

BackgroundThis research aims to improve the prediction of Mild Cognitive Impairment (MCI) conversion to Alzheimer's disease. In order to achieve this, this study focuses on seven specific brain regions identified using the brain atlas. The regions are Hippocampus, Entorhinal cortex, Cerebral cortex, Frontal lobe, Temporal lobe, Parietal lobe, and Occipital lobe. The decay in the gray matter in these regions is associated with the cognitive impairment. This method proposes a novel feature extraction method based on Auto Encoders and then uses these feature to create a graph representing the association between these regions.MethodThe latent representation of the seven regions is found using a novel auto‐encoder based method. This is followed by the formation of a graph, where each of the above regions are nodes and the distance between these nodes is proportional to the inverse of the similarity between the latent representation of the regions.By examining the relationships between these regions, the study seeks to identify patterns associated with MCI conversion. The method involves flattening the above‐formed graph representation into a 1‐D vector, which serves as a unique feature representation for each brain volume. The classification is done using the Support Vector Machine Linear Kernel and forward feature selection is used for selecting the pertinent features.ResultThe method has been validated using the data has obtained from Alzheimer's Disease Neuroimaging Initiative (ADNI). We collected 75 s‐MRI scans of the patients suffering from MCI who converted to Alzheimer's (MCI‐Converts) and 112 s‐MRI scans of the patients suffering from MCI who did not convert to Alzheimer's (MCI‐Non Converts). The F‐score of the classification is 95.4%, which is better than the state of the art.ConclusionThe proposed method provides region‐specific insights, that is it allows for the identification of specific brain regions that are crucial in predicting MCI conversion. It also opens the doors for network analysis of the connections between regions and provides valuable information about the underlying networks involved in MCI conversion. Furthermore, the method also gives promising results and is more generalizable.