Frontiers in neuroimaging最新文献

Automatic brain tumor segmentation is particularly challenging on magnetic resonance imaging (MRI) with marked pathologies, such as brain tumors, which usually cause large displacement, abnormal appearance, and deformation of brain tissue. Despite an abundance of previous literature on learning-based methodologies for MRI segmentation, few works have focused on tackling MRI skull stripping of brain tumor patient data. This gap in literature can be associated with the lack of publicly available data (due to concerns about patient identification) and the labor-intensive nature of generating ground truth labels for model training. In this retrospective study, we assessed the performance of Dense-Vnet in skull stripping brain tumor patient MRI trained on our large multi-institutional brain tumor patient dataset. Our data included pretreatment MRI of 668 patients from our in-house institutional review board-approved multi-institutional brain tumor repository. Because of the absence of ground truth, we used imperfect automatically generated training labels using SPM12 software. We trained the network using common MRI sequences in oncology: T1-weighted with gadolinium contrast, T2-weighted fluid-attenuated inversion recovery, or both. We measured model performance against 30 independent brain tumor test cases with available manual brain masks. All images were harmonized for voxel spacing and volumetric dimensions before model training. Model training was performed using the modularly structured deep learning platform NiftyNet that is tailored toward simplifying medical image analysis. Our proposed approach showed the success of a weakly supervised deep learning approach in MRI brain extraction even in the presence of pathology. Our best model achieved an average Dice score, sensitivity, and specificity of, respectively, 94.5, 96.4, and 98.5% on the multi-institutional independent brain tumor test set. To further contextualize our results within existing literature on healthy brain segmentation, we tested the model against healthy subjects from the benchmark LBPA40 dataset. For this dataset, the model achieved an average Dice score, sensitivity, and specificity of 96.2, 96.6, and 99.2%, which are, although comparable to other publications, slightly lower than the performance of models trained on healthy patients. We associate this drop in performance with the use of brain tumor data for model training and its influence on brain appearance.

Recent advances in neuroimaging create groundbreaking opportunities to better understand human neurological and psychiatric diseases, but also bring new challenges. With the advent of more and more sophisticated and efficient multimodal image processing software, we can now study much larger populations and integrate information from multiple modalities. In consequence, investigators that use neuroimaging techniques must also understand and apply principles of population sampling and contemporary data analytic techniques. The next generation of neuroimaging researchers must be skilled in numerous previously distinct disciplines and so a new integrated model of training is needed. This tutorial presents the rationale for such a new training model and presents the results from the first years of the training program focused on population neuroimaging of Alzheimer's Disease. This approach is applicable to other areas of population neuroimaging.

Background: In Alzheimer's disease (AD), the deposition of β-amyloid (Aβ) plaques is closely associated with the neuronal apoptosis and activation of microglia, which may result in the functional impairment of neurons through pro-inflammation and over-pruning of the neurons. Photobiomodulation (PBM) is a non-invasive therapeutic approach without any conspicuous side effect, which has shown promising attributes in the treatment of chronic brain diseases such as AD by reducing the Aβ burden. However, neither the optimal parameters for PBM treatment nor its exact role in modulating the microglial functions/activities has been conclusively established yet.

Methods: An inflammatory stimulation model of Alzheimer's disease (AD) was set up by activating microglia and neuroblastoma with fibrosis β-amyloid (fAβ) in a transwell insert system. SH-SY5Y neuroblastoma cells and BV2 microglial cells were irradiated with the 808- and 1,064-nm lasers, respectively (a power density of 50 mW/cm² and a dose of 10 J/cm²) to study the PBM activity. The amount of labeled fAβ phagocytosed by microglia was considered to assess the microglial phagocytosis. A PBM-induced neuroprotective study was conducted with the AD model under different laser parameters to realize the optimal condition. Microglial phenotype, microglial secretions of the pro-inflammatory and anti-inflammatory factors, and the intracellular Ca²⁺ levels in microglia were studied in detail to understand the structural and functional changes occurring in the microglial cells of AD model upon PBM treatment.

Conclusion: A synergistic PBM effect (with the 808- and 1,064-nm lasers) effectively inhibited the fAβ-induced neurotoxicity of neuroblastoma by promoting the viability of neuroblastoma and regulating the intracellular Ca²⁺ levels of microglia. Moreover, the downregulation of Ca²⁺ led to microglial polarization with an M2 phenotype, which promotes the fAβ phagocytosis, and resulted in the upregulated expression of anti-inflammatory factors and downregulated expression of inflammatory factors.

Registration methods facilitate the comparison of multiparametric magnetic resonance images acquired at different stages of brain tumor treatments. Image-based registration solutions are influenced by the sequences chosen to compute the distance measure, and the lack of image correspondences due to the resection cavities and pathological tissues. Nonetheless, an evaluation of the impact of these input parameters on the registration of longitudinal data is still missing. This work evaluates the influence of multiple sequences, namely T1-weighted (T1), T2-weighted (T2), contrast enhanced T1-weighted (T1-CE), and T2 Fluid Attenuated Inversion Recovery (FLAIR), and the exclusion of the pathological tissues on the non-rigid registration of pre- and post-operative images. We here investigate two types of registration methods, an iterative approach and a convolutional neural network solution based on a 3D U-Net. We employ two test sets to compute the mean target registration error (mTRE) based on corresponding landmarks. In the first set, markers are positioned exclusively in the surroundings of the pathology. The methods employing T1-CE achieves the lowest mTREs, with a improvement up to 0.8 mm for the iterative solution. The results are higher than the baseline when using the FLAIR sequence. When excluding the pathology, lower mTREs are observable for most of the methods. In the second test set, corresponding landmarks are located in the entire brain volumes. Both solutions employing T1-CE obtain the lowest mTREs, with a decrease up to 1.16 mm for the iterative method, whereas the results worsen using the FLAIR. When excluding the pathology, an improvement is observable for the CNN method using T1-CE. Both approaches utilizing the T1-CE sequence obtain the best mTREs, whereas the FLAIR is the least informative to guide the registration process. Besides, the exclusion of pathology from the distance measure computation improves the registration of the brain tissues surrounding the tumor. Thus, this work provides the first numerical evaluation of the influence of these parameters on the registration of longitudinal magnetic resonance images, and it can be helpful for developing future algorithms.

The detection of new multiple sclerosis (MS) lesions is an important marker of the evolution of the disease. The applicability of learning-based methods could automate this task efficiently. However, the lack of annotated longitudinal data with new-appearing lesions is a limiting factor for the training of robust and generalizing models. In this study, we describe a deep-learning-based pipeline addressing the challenging task of detecting and segmenting new MS lesions. First, we propose to use transfer-learning from a model trained on a segmentation task using single time-points. Therefore, we exploit knowledge from an easier task and for which more annotated datasets are available. Second, we propose a data synthesis strategy to generate realistic longitudinal time-points with new lesions using single time-point scans. In this way, we pretrain our detection model on large synthetic annotated datasets. Finally, we use a data-augmentation technique designed to simulate data diversity in MRI. By doing that, we increase the size of the available small annotated longitudinal datasets. Our ablation study showed that each contribution lead to an enhancement of the segmentation accuracy. Using the proposed pipeline, we obtained the best score for the segmentation and the detection of new MS lesions in the MSSEG2 MICCAI challenge.

We reviewed fMRI experiments from our previous work in conscious rabbits, an experimental preparation that is advantageous for measuring brain activation that is free of anesthetic modulation and which can address questions in a variety of areas in sensory, cognitive, and pharmacological neuroscience research. Rabbits do not struggle or move for several hours while sitting with their heads restrained inside the horizontal bore of a magnet. This greatly reduces movement artifacts in magnetic resonance (MR) images in comparison to other experimental animals such as rodents, cats, and monkeys. We have been able to acquire high-resolution anatomic as well as functional images that are free of movement artifacts during several hours of restraint. Results from conscious rabbit fMRI studies with whisker stimulation are provided to illustrate the feasibility of this conscious animal model for functional MRI and the reproducibility of data gained with it.

Diffusion tensor imaging (DTI) is considered feasible for the nerve plexuses' imaging and quantitative evaluation but its value in the clinical practice is still virtually unexplored. We present the DTI profile of a case of acute varicella-zoster virus (VZV)-related brachial plexopathy. A 72-year-old woman presented with left upper-limb segmental paresis involving the spinal metamers C6-C7, preceded by a painful dermatomal vesicular eruption in C5-T1 dermatomes. Clinical and electrophysiological findings and magnetic resonance imaging indicated a plexus involvement. DTI analysis showed decreased fractional anisotropy (FA) and an increase of all the other diffusivity indexes, i.e., mean, axial, and radial diffusivity. The mechanisms underlying DTI parameter differences between healthy and pathologic brachial plexus sides could be related to microstructural fiber damage. Water diffusion is affected within the nerve roots by increasing the diffusion distance, leading to increased diffusion perpendicular to the largest eigenvalue and therefore to decreased FA values The role of DTI in clinical practice has not been defined yet. Additional quantitative and qualitative DTI information could improve the assessment and follow-up of brachial plexopathy.

Neural activity can be readily and non-invasively recorded from the scalp using electromagnetic and optical signals, but unfortunately all scalp-based techniques have depth-dependent sensitivities. We hypothesize, though, that the cortex's connectivity with the rest of the brain could serve to construct proxy signals of deeper brain activity. For example, functional magnetic resonance imaging (fMRI)-derived models that link surface connectivity to deeper regions could subsequently extend the depth capabilities of other modalities. Thus, as a first step toward this goal, this study examines whether or not surface-limited support vector regression of resting-state fMRI can indeed track deeper regions and distributed networks in independent data. Our results demonstrate that depth-limited fMRI signals can in fact be calibrated to report ongoing activity of deeper brain structures. Although much future work remains to be done, the present study suggests that scalp recordings have the potential to ultimately overcome their intrinsic physical limitations by utilizing the multivariate information exchanged between the surface and the rest of the brain.

Aim: Resting-state fMRI (rs-fMRI) is often used to infer regional brain interactions from the degree of temporal correlation between spontaneous low-frequency fluctuations, thought to reflect local changes in the BOLD signal due to neuronal activity. One complication in the analysis and interpretation of rs-fMRI data is the existence of non-neuronal low frequency physiological noise (systemic low frequency oscillations; sLFOs) which occurs within the same low frequency band as the signal used to compute functional connectivity. Here, we demonstrate the use of a time lag mapping technique to estimate and mitigate the effects of the sLFO signal on resting state functional connectivity of awake squirrel monkeys.

Methods: Twelve squirrel monkeys (6 male/6 female) were acclimated to awake scanning procedures; whole-brain fMRI images were acquired with a 9.4 Tesla scanner. Rs-fMRI data was preprocessed using an in-house pipeline and sLFOs were detected using a seed regressor generated by averaging BOLD signal across all voxels in the brain, which was then refined recursively within a time window of -16-12 s. The refined regressor was then used to estimate the voxel-wise sLFOs; these regressors were subsequently included in the general linear model to remove these moving hemodynamic components from the rs-fMRI data using general linear model filtering. Group level independent component analysis (ICA) with dual regression was used to detect resting-state networks and compare networks before and after sLFO denoising.

Results: Results show sLFOs constitute ~64% of the low frequency fMRI signal in squirrel monkey gray matter; they arrive earlier in regions in proximity to the middle cerebral arteries (e.g., somatosensory cortex) and later in regions close to draining vessels (e.g., cerebellum). Dual regression results showed that the physiological noise was significantly reduced after removing sLFOs and the extent of reduction was determined by the brain region contained in the resting-state network.

Conclusion: These results highlight the need to estimate and remove sLFOs from fMRI data before further analysis.

Multimodal neuroimaging using EEG and fMRI provides deeper insights into brain function by improving the spatial and temporal resolution of the acquired data. However, simultaneous EEG-fMRI inevitably compromises the quality of the EEG and fMRI signals due to the high degree of interaction between the two systems. Fluctuations in the magnetic flux flowing through the participant and the EEG system, whether due to movement within the magnetic field of the scanner or to changes in magnetic field strength, induce electrical potentials in the EEG recordings that mask the much weaker electrical activity of the neuronal populations. A number of different methods have been proposed to reduce MR artifacts. We present an overview of the most commonly used methods and an evaluation of the methods using three sets of diverse EEG data. We limited the evaluation to open-access and easy-to-use methods and a reference signal regression method using a set of six carbon-wire loops (CWL), which allowed evaluation of their added value. The evaluation was performed by comparing EEG signals recorded outside the MRI scanner with artifact-corrected EEG signals recorded simultaneously with fMRI. To quantify and evaluate the quality of artifact reduction methods in terms of the spectral content of the signal, we analyzed changes in oscillatory activity during a resting-state and a finger tapping motor task. The quality of artifact reduction in the time domain was assessed using data collected during a visual stimulation task. In the study we utilized hierarchical Bayesian probabilistic modeling for statistical inference and observed significant differences between the evaluated methods in the success of artifact reduction and associated signal quality in both the frequency and time domains. In particular, the CWL system proved superior to the other methods evaluated in improving spectral contrast in the alpha and beta bands and in recovering visual evoked responses. Based on the results of the evaluation study, we proposed guidelines for selecting the optimal method for MR artifact reduction.