Improving multiple sclerosis lesion segmentation across clinical sites: A federated learning approach with noise-resilient training.

Accurately measuring the evolution of Multiple Sclerosis (MS) with magnetic resonance imaging (MRI) critically informs understanding of disease progression and helps to direct therapeutic strategy. Deep learning models have shown promise for automatically segmenting MS lesions, but the scarcity of accurately annotated data hinders progress in this area. Obtaining sufficient data from a single clinical site is challenging and does not address the heterogeneous need for model robustness. Conversely, the collection of data from multiple sites introduces data privacy concerns and potential label noise due to varying annotation standards. To address this dilemma, we explore the use of the federated learning framework while considering label noise. Our approach enables collaboration among multiple clinical sites without compromising data privacy under a federated learning paradigm that incorporates a noise-robust training strategy based on label correction. Specifically, we introduce a Decoupled Hard Label Correction (DHLC) strategy that considers the imbalanced distribution and fuzzy boundaries of MS lesions, enabling the correction of false annotations based on prediction confidence. We also introduce a Centrally Enhanced Label Correction (CELC) strategy, which leverages the aggregated central model as a correction teacher for all sites, enhancing the reliability of the correction process. Extensive experiments conducted on two multi-site datasets demonstrate the effectiveness and robustness of our proposed methods, indicating their potential for clinical applications in multi-site collaborations to train better deep learning models with lower cost in data collection and annotation.

Data-driven normative values based on generative manifold learning for quantitative MRI.

In medicine, abnormalities in quantitative metrics such as the volume reduction of one brain region of an individual versus a control group are often provided as deviations from so-called normal values. These normative reference values are traditionally calculated based on the quantitative values from a control group, which can be adjusted for relevant clinical co-variables, such as age or sex. However, these average normative values do not take into account the globality of the available quantitative information. For example, quantitative analysis of T1-weighted magnetic resonance images based on anatomical structure segmentation frequently includes over 100 cerebral structures in the quantitative reports, and these tend to be analyzed separately. In this study, we propose a global approach to personalized normative values for each brain structure using an unsupervised Artificial Intelligence technique known as generative manifold learning. We test the potential benefit of these personalized normative values in comparison with the more traditional average normative values on a population of patients with drug-resistant epilepsy operated for focal cortical dysplasia, as well as on a supplementary healthy group and on patients with Alzheimer's disease.

A novel method for PET connectomics guided by fibre-tracking MRI: Application to Alzheimer's disease.

This study introduces a novel brain connectome matrix, track-weighted PET connectivity (twPC) matrix, which combines positron emission tomography (PET) and diffusion magnetic resonance imaging data to compute a PET-weighted connectome at the individual subject level. The new method is applied to characterise connectivity changes in the Alzheimer's disease (AD) continuum. The proposed twPC samples PET tracer uptake guided by the underlying white matter fibre-tracking streamline point-to-point connectivity calculated from diffusion MRI (dMRI). Using tau-PET, dMRI and T1-weighted MRI from the Alzheimer's Disease Neuroimaging Initiative database, structural connectivity (SC) and twPC matrices were computed and analysed using the network-based statistic (NBS) technique to examine topological alterations in early mild cognitive impairment (MCI), late MCI and AD participants. Correlation analysis was also performed to explore the coupling between SC and twPC. The NBS analysis revealed progressive topological alterations in both SC and twPC as cognitive decline progressed along the continuum. Compared to healthy controls, networks with decreased SC were identified in late MCI and AD, and networks with increased twPC were identified in early MCI, late MCI and AD. The altered network topologies were mostly different between twPC and SC, although with several common edges largely involving the bilateral hippocampus, fusiform gyrus and entorhinal cortex. Negative correlations were observed between twPC and SC across all subject groups, although displaying an overall reduction in the strength of anti-correlation with disease progression. twPC provides a new means for analysing subject-specific PET and MRI-derived information within a hybrid connectome using established network analysis methods, providing valuable insights into the relationship between structural connections and molecular distributions. PRACTITIONER POINTS: New method is proposed to compute patient-specific PET connectome guided by MRI fibre-tracking. Track-weighted PET connectivity (twPC) matrix allows to leverage PET and structural connectivity information. twPC was applied to dementia, to characterise the PET nework abnormalities in Alzheimer's disease and mild cognitive impairment.

High angular diffusion tensor imaging estimation from minimal evenly distributed diffusion gradient directions.

Diffusion-weighted Imaging (DWI) is a non-invasive imaging technique based on Magnetic Resonance Imaging (MRI) principles to measure water diffusivity and reveal details of the underlying brain micro-structure. By fitting a tensor model to quantify the directionality of water diffusion a Diffusion Tensor Image (DTI) can be derived and scalar measures, such as fractional anisotropy (FA), can then be estimated from the DTI to summarise quantitative microstructural information for clinical studies. In particular, FA has been shown to be a useful research metric to identify tissue abnormalities in neurological disease (e.g. decreased anisotropy as a proxy for tissue damage). However, time constraints in clinical practice lead to low angular resolution diffusion imaging (LARDI) acquisitions that can cause inaccurate FA value estimates when compared to those generated from high angular resolution diffusion imaging (HARDI) acquisitions. In this work, we propose High Angular DTI Estimation Network (HADTI-Net) to estimate an enhanced DTI model from LARDI with a set of minimal and evenly distributed diffusion gradient directions. Extensive experiments have been conducted to show the reliability and generalisation of HADTI-Net to generate high angular DTI estimation from any minimal evenly distributed diffusion gradient directions and to explore the feasibility of applying a data-driven method for this task. The code repository of this work and other related works can be found at https://mri-synthesis.github.io/.

Parallel processing relies on a distributed, low-dimensional cortico-cerebellar architecture.

A characteristic feature of human cognition is our ability to 'multi-task'-performing two or more tasks in parallel-particularly when one task is well learned. How the brain supports this capacity remains poorly understood. Most past studies have focussed on identifying the areas of the brain-typically the dorsolateral prefrontal cortex-that are required to navigate information-processing bottlenecks. In contrast, we take a systems neuroscience approach to test the hypothesis that the capacity to conduct effective parallel processing relies on a distributed architecture that interconnects the cerebral cortex with the cerebellum. The latter structure contains over half of the neurons in the adult human brain and is well suited to support the fast, effective, dynamic sequences required to perform tasks relatively automatically. By delegating stereotyped within-task computations to the cerebellum, the cerebral cortex can be freed up to focus on the more challenging aspects of performing the tasks in parallel. To test this hypothesis, we analysed task-based fMRI data from 50 participants who performed a task in which they either balanced an avatar on a screen (balance), performed serial-7 subtractions (calculation) or performed both in parallel (dual task). Using a set of approaches that include dimensionality reduction, structure-function coupling, and time-varying functional connectivity, we provide robust evidence in support of our hypothesis. We conclude that distributed interactions between the cerebral cortex and cerebellum are crucially involved in parallel processing in the human brain.

MRI tractography-guided PET image reconstruction regularisation using connectome-based nonlocal means filtering.

Positron emission tomography (PET) molecular biomarkers and diffusion magnetic resonance imaging (dMRI) derived information show associations and highly complementary information in a number of neurodegenerative conditions, including Alzheimer's disease. Diffusion MRI provides valuable information about the microstructure and structural connectivity (SC) of the brain which could guide and improve the PET image reconstruction when such associations exist. However, this potental has not been previously explored. In the present study, we propose a CONNectome-based non-local means one-atep late maximuma posteriori(CONN-NLM-OSLMAP) method, which allows diffusion MRI-derived connectivity information to be incorporated into the PET iterative image reconstruction process, thus regularising the estimated PET images. The proposed method was evaluated using a realistic tau-PET/MRI simulated phantom, demonstrating more effective noise reduction and lesion contrast improvement, as well as the lowest overall bias compared with both a median filter applied as an alternative regulariser and CONNectome-based non-local means as a post-reconstruction filter. By adding complementary SC information from diffusion MRI, the proposed regularisation method offers more useful and targeted denoising and regularisation, demonstrating the feasibility and effectiveness of integrating connectivity information into PET image reconstruction.

Multiple sclerosis lesion segmentation: revisiting weighting mechanisms for federated learning.

Background and introduction: Federated learning (FL) has been widely employed for medical image analysis to facilitate multi-client collaborative learning without sharing raw data. Despite great success, FL's applications remain suboptimal in neuroimage analysis tasks such as lesion segmentation in multiple sclerosis (MS), due to variance in lesion characteristics imparted by different scanners and acquisition parameters. Methods: In this work, we propose the first FL MS lesion segmentation framework via two effective re-weighting mechanisms. Specifically, a learnable weight is assigned to each local node during the aggregation process, based on its segmentation performance. In addition, the segmentation loss function in each client is also re-weighted according to the lesion volume for the data during training. Results: The proposed method has been validated on two FL MS segmentation scenarios using public and clinical datasets. Specifically, the case-wise and voxel-wise Dice score of the proposed method under the first public dataset is 65.20 and 74.30, respectively. On the second in-house dataset, the case-wise and voxel-wise Dice score is 53.66, and 62.31, respectively. Discussions and conclusions: The Comparison experiments on two FL MS segmentation scenarios using public and clinical datasets have demonstrated the effectiveness of the proposed method by significantly outperforming other FL methods. Furthermore, the segmentation performance of FL incorporating our proposed aggregation mechanism can achieve comparable performance to that from centralized training with all the raw data.

Tremor suppression following treatment with MRgFUS: skull density ratio consistency and degree of posterior dentatorubrothalamic tract lesioning predicts long-term clinical outcomes in essential tremor.

Objectives: Magnetic resonance-guided focussed ultrasound (MRgFUS) is an incisionless ablative procedure, widely used for treatment of Parkinsonian and Essential Tremor (ET). Enhanced understanding of the patient- and treatment-specific factors that influence sustained long-term tremor suppression could help clinicians achieve superior outcomes via improved patient screening and treatment strategy. Methods: We retrospectively analysed data from 31 subjects with ET, treated with MRgFUS at a single centre. Tremor severity was assessed with parts A, B and C of the Clinical Rating Scale for Tremor (CRST) as well as the combined CRST. Tremor in the dominant and non-dominant hand was assessed with Hand Tremor Scores (HTS), derived from the CRST. Pre- and post-treatment imaging data were analysed to determine ablation volume overlap with automated thalamic segmentations, and the dentatorubrothalamic tract (DRTT) and compared with percentage change in CRST and HTS following treatment. Results: Tremor symptoms were significantly reduced following treatment. Combined pre-treatment CRST (mean: 60.7 ± 17.3) and HTS (mean: 19.2 ± 5.7) improved by an average of 45.5 and 62.6%, respectively. Percentage change in CRST was found to be significantly negatively associated with age (ß = -0.375, p = 0.015), and SDR standard deviation (SDRSD; ß = -0.324, p = 0.006), and positively associated with ablation overlap with the posterior DRTT (ß = 0.535, p < 0.001). Percentage HTS improvement in the dominant hand decreased significantly with older age (ß = -0.576, p < 0.01). Conclusion: Our results suggest that increased lesioning of the posterior region of the DRTT could result in greater improvements in combined CRST and non-dominant hand HTS, and that subjects with lower SDR standard deviation tended to experience greater improvement in combined CRST.

MFCSC: Novel method to calculate mismatch between functional and structural brain connectomes, and its application for detecting hemispheric functional specialisations.

We introduce a novel connectomics method, MFCSC, that integrates information on structural connectivity (SC) from diffusion MRI tractography and functional connectivity (FC) from functional MRI, at individual subject level. The MFCSC method is based on the fact that SC only broadly predicts FC, and for each connection in the brain, the method calculates a value that quantifies the mismatch that often still exists between the two modalities. To capture underlying physiological properties, MFCSC minimises biases in SC and addresses challenges with the multimodal analysis, including by using a data-driven normalisation approach. We ran MFCSC on data from the Human Connectome Project and used the output to detect pairs of left and right unilateral connections that have distinct relationship between structure and function in each hemisphere; we suggest that this reflects cases of hemispheric functional specialisation. In conclusion, the MFCSC method provides new information on brain organisation that may not be inferred from an analysis that considers SC and FC separately.

Individualised profiling of white matter organisation in moderate-to-severe traumatic brain injury patients.

BACKGROUND AND PURPOSE: Approximately 65% of moderate-to-severe traumatic brain injury (m-sTBI) patients present with poor long-term behavioural outcomes, which can significantly impair activities of daily living. Numerous diffusion-weighted MRI studies have linked these poor outcomes to decreased white matter integrity of several commissural tracts, association fibres and projection fibres in the brain. However, most studies have focused on group-based analyses, which are unable to deal with the substantial between-patient heterogeneity in m-sTBI. As a result, there is increasing interest and need in conducting individualised neuroimaging analyses. MATERIALS AND METHODS: Here, we generated a detailed subject-specific characterisation of microstructural organisation of white matter tracts in 5 chronic patients with m-sTBI (29 - 49y, 2 females), presented as a proof-of-concept. We developed an imaging analysis framework using fixel-based analysis and TractLearn to determine whether the values of fibre density of white matter tracts at the individual patient level deviate from the healthy control group (n = 12, 8F, Mage = 35.7y, age range 25 - 64y). RESULTS: Our individualised analysis revealed unique white matter profiles, confirming the heterogenous nature of m-sTBI and the need of individualised profiles to properly characterise the extent of injury. Future studies incorporating clinical data, as well as utilising larger reference samples and examining the test-retest reliability of the fixel-wise metrics are warranted. CONCLUSIONS: Individualised profiles may assist clinicians in tracking recovery and planning personalised training programs for chronic m-sTBI patients, which is necessary to achieve optimal behavioural outcomes and improved quality of life.

Building a tissue-unbiased brain template of fiber orientation distribution and tractography with multimodal registration.

PURPOSE: Brain templates provide an essential standard space for statistical analysis of brain structure and function. Despite recent advances, diffusion MRI still lacks a template of fiber orientation distribution (FOD) and tractography that is unbiased for both white and gray matter. Therefore, we aim to build up a set of such templates for better white-matter analysis and joint structural and functional analysis. METHODS: We have developed a multimodal registration method to leverage the complementary information captured by T1 -weighted, T2 -weighted, and diffusion MRI, so that a coherent transformation is generated to register FODs into a common space and average them into a template. Consequently, the anatomically constrained fiber-tracking method was applied to the FOD template to generate a tractography template. Fiber-centered functional connectivity analysis was then performed as an example of the benefits of such an unbiased template. RESULTS: Our FOD template preserves fine structural details in white matter and also, importantly, clear folding patterns in the cortex and good contrast in the subcortex. Quantitatively, our templates show better individual-template agreement at the whole-brain scale and segmentation scale. The tractography template aligns well with the gray matter, which led to fiber-centered functional connectivity showing high cross-group consistency. CONCLUSION: We have proposed a novel methodology for building a tissue-unbiased FOD and anatomically constrained tractography template based on multimodal registration. Our templates provide a standard space and statistical platform for not only white-matter analysis but also joint structural and functional analysis, therefore filling an important gap in multimodal neuroimage analysis.

Hippocampal Neuronal Integrity and Functional Connectivity Within the Default Mode Network in Mild Cognitive Impairment: A Multimodal Investigation.

Background: In older people with mild cognitive impairment (MCI), the relationship between early changes in functional connectivity and in vivo changes in key neurometabolites is not known. Two established correlates of MCI diagnosis are decreased N-acetylaspartate (NAA) in the hippocampus, indicative of decreased neuronal integrity, and changes in the default mode network (DMN) functional network. If and how these measures interrelate is yet to be established, and such understanding may provide insight into the processes underpinning observed cognitive decline. Objectives: To determine the relationship between NAA levels in the left hippocampus and functional connectivity within the DMN in an aging cohort. Methods: In a sample of 51 participants with MCI and 30 controls, hippocampal NAA was determined using magnetic resonance spectroscopy, and DMN connectivity was quantified using resting-state functional MRI. The association between hippocampal NAA and the DMN functional connectivity was tested within the MCI group and separately within the control group. Results: In the DMN, we showed a significant inverse association between functional connectivity and hippocampal NAA in 20 specific brain connections for patients with MCI. This was despite no evidence of any associations in the healthy control group or group differences in either of these measures alone. Conclusions: This study suggests that decreased neuronal integrity in the hippocampus is associated with functional change within the DMN for those with MCI, in contrast to healthy older adults. These results highlight the potential of multimodal investigations to better understand the processes associated with cognitive decline. Impact statement This study measured activity within the default mode network (DMN) and quantified N-acetylaspartate (NAA), a measure of neuronal integrity, within the hippocampus in participants with mild cognitive impairment (MCI) and healthy controls. In participants with MCI, NAA levels were inversely associated with connectivity between specific regions of the DMN, a relationship not evident in healthy controls. This association was present even in the absence of group differences in DMN connectivity or NAA levels. This research illustrates the possibility of using multiple magnetic resonance modalities for more sensitive measures of early cognitive decline to identify and intervene earlier.

New insights into anatomical connectivity along the anterior-posterior axis of the human hippocampus using in vivo quantitative fibre tracking.

The hippocampus supports multiple cognitive functions including episodic memory. Recent work has highlighted functional differences along the anterior-posterior axis of the human hippocampus, but the neuroanatomical underpinnings of these differences remain unclear. We leveraged track-density imaging to systematically examine anatomical connectivity between the cortical mantle and the anterior-posterior axis of the in vivo human hippocampus. We first identified the most highly connected cortical areas and detailed the degree to which they preferentially connect along the anterior-posterior axis of the hippocampus. Then, using a tractography pipeline specifically tailored to measure the location and density of streamline endpoints within the hippocampus, we characterised where these cortical areas preferentially connect within the hippocampus. Our results provide new and detailed insights into how specific regions along the anterior-posterior axis of the hippocampus are associated with different cortical inputs/outputs and provide evidence that both gradients and circumscribed areas of dense extrinsic anatomical connectivity exist within the human hippocampus. These findings inform conceptual debates in the field and emphasise the importance of considering the hippocampus as a heterogeneous structure. Overall, our results represent a major advance in our ability to map the anatomical connectivity of the human hippocampus in vivo and inform our understanding of the neural architecture of hippocampal-dependent memory systems in the human brain.

The brain allows us to perceive and interact with our environment and to create and recall memories about our day-to-day lives. A sea-horse shaped structure in the brain, called the hippocampus, is critical for translating our perceptions into memories, and it does so in coordination with other brain regions. For example, different regions of the cerebral cortex (the outer layer of the brain) support different aspects of cognition, and pathways of information flow between the cerebral cortex and hippocampus underpin the healthy functioning of memory. Decades of research conducted into the brains of non-human primates show that specific regions of the cerebral cortex anatomically connect with different parts of the hippocampus to support this information flow. These insights form the foundation for existing theoretical models of how networks of neurons in the hippocampus and the cerebral cortex are connected. However, the human cerebral cortex has greatly expanded during our evolution, meaning that patterns of connectivity in the human brain may diverge from those in the brains of non-human primates. Deciphering human brain circuits in greater detail is crucial if we are to gain a better understanding of the structure and operation of the healthy human brain. However, obtaining comprehensive maps of anatomical connections between the hippocampus and cerebral cortex has been hampered by technical limitations. For example, magnetic resonance imaging (MRI), an approach that can be used to study the living human brain, suffers from insufficient image resolution. To overcome these issues, Dalton et al. used an imaging technique called diffusion weighted imaging which is used to study white matter pathways in the brain. They developed a tailored approach to create high-resolution maps showing how the hippocampus anatomically connects with the cerebral cortex in the healthy human brain. Dalton et al. produced detailed maps illustrating which areas of the cerebral cortex have high anatomical connectivity with the hippocampus and how different parts of the hippocampus preferentially connect to different neural circuits in the cortex. For example, the experiments demonstrate that highly connected areas in a cortical region called the temporal cortex connect to very specific, circumscribed regions within the hippocampus. These findings suggest that the hippocampus may consist of different neural circuits, each preferentially linked to defined areas of the cortex which are, in turn, associated with specific aspects of cognition. These observations further our knowledge of hippocampal-dependant memory circuits in the human brain and provide a foundation for the study of memory decline in aging and neurodegenerative diseases.

Gradual changes in microarchitectural properties of cortex and juxtacortical white matter: Observed by anatomical and diffusion MRI.

PURPOSE: Characterization of cerebral cortex is challenged by the complexity and heterogeneity of its cyto- and myeloarchitecture. This study evaluates quantitative MRI metrics, measured across two cortical depths and in subcortical white matter (WM) adjacent to cortex (juxtacortical WM), indicative of myelin content, neurite density, and diffusion microenvironment, for a comprehensive characterization of cortical microarchitecture. METHODS: High-quality structural and diffusion MRI data (N = 30) from the Human Connectome Project were processed to compute myelin index, neurite density index, fractional anisotropy, mean diffusivity, axial diffusivity, and radial diffusivity from superficial cortex, deep cortex, and juxtacortical WM. The distributional patterns of these metrics were analyzed individually, correlated to one another, and were compared to established parcellations. RESULTS: Our results supported that myeloarchitectonic and the coexisting cytoarchitectonic structures influence the diffusion properties of water molecules residing in cortex. Full cortical thickness showed myelination patterns similar to those previously observed in humans. Higher myelin indices with similar distributional patterns were observed in deep cortex whereas lower myelin indices were observed in superficial cortex. Neurite density index and other diffusion MRI derived parameters provided complementary information to myelination. Reliable and reproducible correlations were identified among the cortical microarchitectural properties and fiber distributional patterns in proximal WM structures. CONCLUSION: We demonstrated gradual changes across the cortical sheath by assessing depth-specific cortical micro-architecture using anatomical and diffusion MRI. Mutually independent but coexisting features of cortical layers and juxtacortical WM provided new insights towards structural organizational units and variabilities across cortical regions and through depth.

CONN-NLM: A Novel CONNectome-Based Non-local Means Filter for PET-MRI Denoising.

Background: Advancements in hybrid positron emission tomography-magnetic resonance (PET-MR) systems allow for combining the advantages of each modality. Integrating information from MRI and PET can be valuable for diagnosing and treating neurological disorders. However, combining diffusion MRI (dMRI) and PET data, which provide highly complementary information, has rarely been exploited in image post-processing. dMRI has the ability to investigate the white matter pathways of the brain through fibre tractography, which enables comprehensive mapping of the brain connection networks (the "connectome"). Novel methods are required to combine information present in the connectome and PET to increase the full potential of PET-MRI. Methods: We developed a CONNectome-based Non-Local Means (CONN-NLM) filter to exploit synergies between dMRI-derived structural connectivity and PET intensity information to denoise PET images. PET-MR data are parcelled into a number of regions based on a brain atlas, and the inter-regional structural connectivity is calculated based on dMRI fibre-tracking. The CONN-NLM filter is then implemented as a post-reconstruction filter by combining the nonlocal means filter and a connectivity-based cortical smoothing. The effect of this approach is to weight voxels with similar PET intensity and highly connected voxels higher when computing the weighted-average to perform more informative denoising. The proposed method was first evaluated using a novel computer phantom framework to simulate realistic hybrid PET-MR images with different lesion scenarios. CONN-NLM was further assessed with clinical dMRI and tau PET examples. Results: The results showed that CONN-NLM has the capacity to improve the overall PET image quality by reducing noise while preserving lesion contrasts, and it outperformed a range of filters that did not use dMRI information. The simulations demonstrate that CONN-NLM can handle various lesion contrasts consistently, as well as lesions with different levels of inter-connectivity. Conclusion: CONN-NLM has unique advantages of providing more informative and accurate PET smoothing by adding complementary structural connectivity information from dMRI, representing a new avenue to exploit synergies between MRI and PET.

Network communication models narrow the gap between the modular organization of structural and functional brain networks.

Structural and functional brain networks are modular. Canonical functional systems, such as the default mode network, are well-known modules of the human brain and have been implicated in a large number of cognitive, behavioral and clinical processes. However, modules delineated in structural brain networks inferred from tractography generally do not recapitulate canonical functional systems. Neuroimaging evidence suggests that functional connectivity between regions in the same systems is not always underpinned by anatomical connections. As such, direct structural connectivity alone would be insufficient to characterize the functional modular organization of the brain. Here, we demonstrate that augmenting structural brain networks with models of indirect (polysynaptic) communication unveils a modular network architecture that more closely resembles the brain's established functional systems. We find that diffusion models of polysynaptic connectivity, particularly communicability, narrow the gap between the modular organization of structural and functional brain networks by 20-60%, whereas routing models based on single efficient paths do not improve mesoscopic structure-function correspondence. This suggests that functional modules emerge from the constraints imposed by local network structure that facilitates diffusive neural communication. Our work establishes the importance of modeling polysynaptic communication to understand the structural basis of functional systems.

FOD-Net: A deep learning method for fiber orientation distribution angular super resolution.

Mapping the human connectome using fiber-tracking permits the study of brain connectivity and yields new insights into neuroscience. However, reliable connectome reconstruction using diffusion magnetic resonance imaging (dMRI) data acquired by widely available clinical protocols remains challenging, thus limiting the connectome/tractography clinical applications. Here we develop fiber orientation distribution (FOD) network (FOD-Net), a deep-learning-based framework for FOD angular super-resolution. Our method enhances the angular resolution of FOD images computed from common clinical-quality dMRI data, to obtain FODs with quality comparable to those produced from advanced research scanners. Super-resolved FOD images enable superior tractography and structural connectome reconstruction from clinical protocols. The method was trained and tested with high-quality data from the Human Connectome Project (HCP) and further validated with a local clinical 3.0T scanner as well as with another public available multicenter-multiscanner dataset. Using this method, we improve the angular resolution of FOD images acquired with typical single-shell low-angular-resolution dMRI data (e.g., 32 directions, b=1000s/mm2) to approximate the quality of FODs derived from time-consuming, multi-shell high-angular-resolution dMRI research protocols. We also demonstrate tractography improvement, removing spurious connections and bridging missing connections. We further demonstrate that connectomes reconstructed by super-resolved FODs achieve comparable results to those obtained with more advanced dMRI acquisition protocols, on both HCP and clinical 3.0T data. Advances in deep-learning approaches used in FOD-Net facilitate the generation of high quality tractography/connectome analysis from existing clinical MRI environments. Our code is freely available at https://github.com/ruizengalways/FOD-Net.