Neuroinformatics最新文献

Neurotransmitter receptor densities are relevant for understanding the molecular architecture of brain regions. Quantitative in vitro receptor autoradiography, has been introduced to map neurotransmitter receptor distributions of brain areas. However, it is very time and cost-intensive, which makes it challenging to obtain whole-brain distributions. At the same time, high-throughput light microscopy and 3D reconstructions have enabled high-resolution brain maps capturing measures of cell density across the whole human brain. Aiming to bridge gaps in receptor measurements for building detailed whole-brain atlases, we study the feasibility of predicting realistic neurotransmitter density distributions from cell-body stainings. Specifically, we utilize conditional Generative Adversarial Networks (cGANs) to predict the density distributions of the M2 receptor of acetylcholine and the kainate receptor for glutamate in the macaque monkey's primary visual (V1) and motor cortex (M1), based on light microscopic scans of cell-body stained sections. Our model is trained on corresponding patches from aligned consecutive sections that display cell-body and receptor distributions, ensuring a mapping between the two modalities. Evaluations of our cGANs, both qualitative and quantitative, show their capability to predict receptor densities from cell-body stained sections while maintaining cortical features such as laminar thickness and curvature. Our work underscores the feasibility of cross-modality image translation problems to address data gaps in multi-modal brain atlases.

The increasing use of neuroimaging in clinical research has driven the creation of many large imaging datasets. However, these datasets often rely on inconsistent naming conventions in image file headers to describe acquisition, and time-consuming manual curation is necessary. Therefore, we sought to automate the process of classifying and organizing magnetic resonance imaging (MRI) data according to acquisition types common to the clinical routine, as well as automate the transformation of raw, unstructured images into Brain Imaging Data Structure (BIDS) datasets. To do this, we trained an XGBoost model to classify MRI acquisition types using relatively few acquisition parameters that are automatically stored by the MRI scanner in image file metadata, which are then mapped to the naming conventions prescribed by BIDS to transform the input images to the BIDS structure. The model recognizes MRI types with 99.475% accuracy, as well as a micro/macro-averaged precision of 0.9995/0.994, a micro/macro-averaged recall of 0.9995/0.989, and a micro/macro-averaged F1 of 0.9995/0.991. Our approach accurately and quickly classifies MRI types and transforms unstructured data into standardized structures with little-to-no user intervention, reducing the barrier of entry for clinical scientists and increasing the accessibility of existing neuroimaging data.

Sensorimotor computation integrates bottom-up world state information with top-down knowledge and task goals to form action plans. In the rodent whisker system, a prime model of active sensing, evidence shows neuromodulatory neurotransmitters shape whisker control, affecting whisking frequency and amplitude. Since neuromodulatory neurotransmitters are mostly released from subcortical nuclei and have long-range projections that reach the rest of the central nervous system, mapping the circuits of top-down neuromodulatory control of sensorimotor nuclei will help to systematically address the mechanisms of active sensing. Therefore, we developed a neuroinformatic target discovery pipeline to mine the Allen Institute's Mouse Brain Connectivity Atlas. Using network connectivity analysis, we identified new putative connections along the whisker system and anatomically confirmed the existence of 42 previously unknown monosynaptic connections. Using this data, we updated the sensorimotor connectivity map of the mouse whisker system and developed the first cell-type-specific map of the network. The map includes 157 projections across 18 principal nuclei of the whisker system and neuromodulatory neurotransmitter-releasing. Performing a graph network analysis of this connectome, we identified cell-type specific hubs, sources, and sinks, provided anatomical evidence for monosynaptic inhibitory projections into all stages of the ascending pathway, and showed that neuromodulatory projections improve network-wide connectivity. These results argue that beyond the modulatory chemical contributions to information processing and transfer in the whisker system, the circuit connectivity features of the neuromodulatory networks position them as nodes of sensory and motor integration.

CADENCE is an open Python 3-written neuroinformatics tool with Qt6 graphic user interface for supervised calcium events detection. In neuronal ensembles recording during calcium imaging experiments, the output of instruments such as Celena X, Zeiss LSM 5 Live confocal microscope and Miniscope is a movie showing flashing cells somata. There are few pipelines to convert video to relative fluorescence ΔF/F, from simplest ImageJ plugins to sophisticated tools like MiniAn (Dong et al. in Elife 11, https://doi.org/10.7554/eLife.70661 , 2022). Minian, an open-source miniscope analysis pipeline. Elife, 11.). While in some areas of study relative fluorescence ΔF/F may be the desired result in itself, researchers of neuronal ensembles are typically interested in a more detailed analysis of calcium events as indirect proxy of neuronal electrical activity. For such analyses, researchers need a tool to infer calcium events from the continuous ΔF/F curve in order to create a raster representation of calcium events for later use in analysis software, such as Elephant (Denker, M., Yegenoglu, A., & Grün, S. (2018). Collaborative HPC-enabled workflows on the HBP Collaboratory using the Elephant framework. Neuroinformatics, 19.). Here we present such an open tool with supervised calcium events detection.

Growth-associated protein 43 (GAP-43) is found in the axonal terminal of neurons in the limbic system, which is affected in people with Alzheimer’s disease (AD). We assumed GAP-43 may contribute to AD progression and serve as a biomarker. So, in a two-year follow-up study, we assessed GAP-43 changes and whether they are correlated with tensor-based morphometry (TBM) findings in patients with mild cognitive impairment (MCI). We included MCI and cognitively normal (CN) people with available baseline and follow-up cerebrospinal fluid (CSF) GAP-43 and TBM findings from the ADNI database. We assessed the difference between the two groups and correlations in each group at each time point. CSF GAP-43 and TBM measures were similar in the two study groups in all time points, except for the accelerated anatomical region of interest (ROI) of CN subjects that were significantly greater than those of MCI. The only significant correlations with GAP-43 observed were those inverse correlations with accelerated and non-accelerated anatomical ROI in MCI subjects at baseline. Plus, all TBM metrics decreased significantly in all study groups during the follow-up in contrast to CSF GAP-43 levels. Our study revealed significant associations between CSF GAP-43 levels and TBM indices among people of the AD spectrum.

Abstract

Multimodal neuroimaging grants a powerful in vivo window into the structure and function of the human brain. Recent methodological and conceptual advances have enabled investigations of the interplay between large-scale spatial trends – or gradients – in brain structure and function, offering a framework to unify principles of brain organization across multiple scales. Strong community enthusiasm for these techniques has been instrumental in their widespread adoption and implementation to answer key questions in neuroscience. Following a brief review of current literature on this framework, this perspective paper will highlight how pragmatic steps aiming to make gradient methods more accessible to the community propelled these techniques to the forefront of neuroscientific inquiry. More specifically, we will emphasize how interest for gradient methods was catalyzed by data sharing, open-source software development, as well as the organization of dedicated workshops led by a diverse team of early career researchers. To this end, we argue that the growing excitement for brain gradients is the result of coordinated and consistent efforts to build an inclusive community and can serve as a case in point for future innovations and conceptual advances in neuroinformatics. We close this perspective paper by discussing challenges for the continuous refinement of neuroscientific theory, methodological innovation, and real-world translation to maintain our collective progress towards integrated models of brain organization.

Deep learning approaches are state-of-the-art for semantic segmentation of medical images, but unlike many deep learning applications, medical segmentation is characterized by small amounts of annotated training data. Thus, while mainstream deep learning approaches focus on performance in domains with large training sets, researchers in the medical imaging field must apply new methods in creative ways to meet the more constrained requirements of medical datasets. We propose a framework for incrementally fine-tuning a multi-class segmentation of a high-resolution multiplex (multi-channel) immuno-flourescence image of a rat brain section, using a minimal amount of labelling from a human expert. Our framework begins with a modified Swin-UNet architecture that treats each biomarker in the multiplex image separately and learns an initial "global" segmentation (pre-training). This is followed by incremental learning and refinement of each class using a very limited amount of additional labeled data provided by a human expert for each region and its surroundings. This incremental learning utilizes the multi-class weights as an initialization and uses the additional labels to steer the network and optimize it for each region in the image. In this way, an expert can identify errors in the multi-class segmentation and rapidly correct them by supplying the model with additional annotations hand-picked from the region. In addition to increasing the speed of annotation and reducing the amount of labelling, we show that our proposed method outperforms a traditional multi-class segmentation by a large margin.

The delineation of cortical areas on magnetic resonance images (MRI) is important for understanding the complexities of the developing human brain. The previous version of the Melbourne Children's Regional Infant Brain (M-CRIB-S) (Adamson et al. Scientific Reports, 10(1), 10, 2020) is a software package that performs whole-brain segmentation, cortical surface extraction and parcellation of the neonatal brain. Available cortical parcellation schemes in the M-CRIB-S are the adult-compatible 34- and 31-region per hemisphere Desikan-Killiany (DK) and Desikan-Killiany-Tourville (DKT), respectively. We present a major update to the software package which achieves two aims: 1) to make the voxel-based segmentation outputs derived from the Freesurfer-compatible M-CRIB scheme, and 2) to improve the accuracy of whole-brain segmentation and cortical surface extraction. Cortical surface extraction has been improved with additional steps to improve penetration of the inner surface into thin gyri. The improved cortical surface extraction is shown to increase the robustness of measures such as surface area, cortical thickness, and cortical volume.

Performing group analysis on magnetic resonance imaging (MRI) data with linear mixed-effects (LME) models is challenging due to its large dimensionality and inherent multi-level covariance structure. In addition, as large-scale collaborative projects become commonplace in neuroimaging, data must increasingly be stored and analyzed from different locations. In such settings, substantial overhead can occur in terms of data transfer and coordination between participating research groups. In some cases, data cannot be pooled together due to privacy or regulatory concerns. In this work, we propose a decentralized LME model to perform a large-scale analysis of data from different collaborations without data pooling. This method is efficient as it overcomes the hurdles of data sharing and has lower bandwidth and memory requirements for analysis than the centralized modeling approach. We evaluate our model using features extracted from structural magnetic resonance imaging (sMRI) data. Results highlight gray matter reductions in the temporal lobe/insula and medial frontal regions in schizophrenia, consistent with prior studies. Our analysis also demonstrates that decentralized LME models achieve similar performance compared to the models trained with all the data in one location. We also implement the decentralized LME approach in COINSTAC, an open source, decentralized platform for federating neuroimaging analysis, providing an easy to use tool for dissemination to the neuroimaging community.

The increasing lifespan and large individual differences in cognitive capability highlight the importance of comprehending the aging process of the brain. Contrary to visible signs of bodily ageing, like greying of hair and loss of muscle mass, the internal changes that occur within our brains remain less apparent until they impair function. Brain age, distinct from chronological age, reflects our brain's health status and may deviate from our actual chronological age. Notably, brain age has been associated with mortality and depression. The brain is plastic and can compensate even for severe structural damage by rewiring. Functional characterization offers insights that structural cannot provide. Contrary to the multitude of studies relying on structural magnetic resonance imaging (MRI), we utilize resting-state functional MRI (rsfMRI). We also address the issue of inclusion of subjects with abnormal brain ageing through outlier removal. In this study, we employ the Least Absolute Shrinkage and Selection Operator (LASSO) to identify the 39 most predictive correlations derived from the rsfMRI data. The data is from a cohort of 176 healthy right-handed volunteers, aged 18-78 years (95/81 male/female, mean age 48, SD 17) collected at the Mind Research Imaging Center at the National Cheng Kung University. We establish a normal reference model by excluding 68 outliers, which achieves a leave-one-out mean absolute error of 2.48 years. By asking which additional features that are needed to predict the chronological age of the outliers with a smaller error, we identify correlations predictive of abnormal aging. These are associated with the Default Mode Network (DMN). Our normal reference model has the lowest prediction error among published models evaluated on adult subjects of almost all ages and is thus a candidate for screening for abnormal brain aging that has not yet manifested in cognitive decline. This study advances our ability to predict brain aging and provides insights into potential biomarkers for assessing brain age, suggesting that the role of DMN in brain aging should be studied further.