Simple, Scalable Proteomic Imaging for High-Dimensional Profiling of Intact Systems.

Combined measurement of diverse molecular and anatomical traits that span multiple levels remains a major challenge in biology. Here, we introduce a simple method that enables proteomic imaging for scalable, integrated, high-dimensional phenotyping of both animal tissues and human clinical samples. This method, termed SWITCH, uniformly secures tissue architecture, native biomolecules, and antigenicity across an entire system by synchronizing the tissue preservation reaction. The heat- and chemical-resistant nature of the resulting framework permits multiple rounds (>20) of relabeling. We have performed 22 rounds of labeling of a single tissue with precise co-registration of multiple datasets. Furthermore, SWITCH synchronizes labeling reactions to improve probe penetration depth and uniformity of staining. With SWITCH, we performed combinatorial protein expression profiling of the human cortex and also interrogated the geometric structure of the fiber pathways in mouse brains. Such integrated high-dimensional information may accelerate our understanding of biological systems at multiple levels.

Lipid and smooth muscle architectural pathology in the rabbit atherosclerotic vessel wall using Q-space cardiovascular magnetic resonance.

BACKGROUND: Atherosclerosis is an arterial vessel wall disease characterized by slow, progressive lipid accumulation, smooth muscle disorganization, and inflammatory infiltration. Atherosclerosis often remains subclinical until extensive inflammatory injury promotes vulnerability of the atherosclerotic plaque to rupture with luminal thrombosis, which can cause the acute event of myocardial infarction or stroke. Current bioimaging techniques are unable to capture the pathognomonic distribution of cellular elements of the plaque and thus cannot accurately define its structural disorganization. METHODS: We applied cardiovascular magnetic resonance spectroscopy (CMRS) and diffusion weighted CMR (DWI) with generalized Q-space imaging (GQI) analysis to architecturally define features of atheroma and correlated these to the microscopic distribution of vascular smooth muscle cells (SMC), immune cells, extracellular matrix (ECM) fibers, thrombus, and cholesteryl esters (CE). We compared rabbits with normal chow diet and cholesterol-fed rabbits with endothelial balloon injury, which accelerates atherosclerosis and produces advanced rupture-prone plaques, in a well-validated rabbit model of human atherosclerosis. RESULTS: Our methods revealed new structural properties of advanced atherosclerosis incorporating SMC and lipid distributions. GQI with tractography portrayed the locations of these components across the atherosclerotic vessel wall and differentiated multi-level organization of normal, pro-inflammatory cellular phenotypes, or thrombus. Moreover, the locations of CE were differentiated from cellular constituents by their higher restrictive diffusion properties, which permitted chemical confirmation of CE by high field voxel-guided CMRS. CONCLUSIONS: GQI with tractography is a new method for atherosclerosis imaging that defines a pathological architectural signature for the atheromatous plaque composed of distributed SMC, ECM, inflammatory cells, and thrombus and lipid. This provides a detailed transmural map of normal and inflamed vessel walls in the setting of atherosclerosis that has not been previously achieved using traditional CMR techniques. Although this is an ex-vivo study, detection of micro and mesoscale level vascular destabilization as enabled by GQI with tractography could increase the accuracy of diagnosis and assessment of treatment outcomes in individuals with atherosclerosis.

Longitudinal predictive modeling of tau progression along the structural connectome.

Tau neurofibrillary tangles, a pathophysiological hallmark of Alzheimer's disease (AD), exhibit a stereotypical spatiotemporal trajectory that is strongly correlated with disease progression and cognitive decline. Personalized prediction of tau progression is, therefore, vital for the early diagnosis and prognosis of AD. Evidence from both animal and human studies is suggestive of tau transmission along the brains preexisting neural connectivity conduits. We present here an analytic graph diffusion framework for individualized predictive modeling of tau progression along the structural connectome. To account for physiological processes that lead to active generation and clearance of tau alongside passive diffusion, our model uses an inhomogenous graph diffusion equation with a source term and provides closed-form solutions to this equation for linear and exponential source functionals. Longitudinal imaging data from two cohorts, the Harvard Aging Brain Study (HABS) and the Alzheimer's Disease Neuroimaging Initiative (ADNI), were used to validate the model. The clinical data used for developing and validating the model include regional tau measures extracted from longitudinal positron emission tomography (PET) scans based on the 18F-Flortaucipir radiotracer and individual structural connectivity maps computed from diffusion tensor imaging (DTI) by means of tractography and streamline counting. Two-timepoint tau PET scans were used to assess the goodness of model fit. Three-timepoint tau PET scans were used to assess predictive accuracy via comparison of predicted and observed tau measures at the third timepoint. Our results show high consistency between predicted and observed tau and differential tau from region-based analysis. While the prognostic value of this approach needs to be validated in a larger cohort, our preliminary results suggest that our longitudinal predictive model, which offers an in vivo macroscopic perspective on tau progression in the brain, is potentially promising as a personalizable predictive framework for AD.

Diffusion interactions between crossing fibers of the brain.

PURPOSE: Recent observations of several preferred orientations of diffusion in deep white matter may indicate either (a) that axons in different directions are independently bundled in thick sheets and function noninteractively, or more interestingly, (b) that the axons are closely interwoven and would exhibit branching and sharp turns. This study aims to investigate whether the dependence of dMRI Q-ball signal on the interpulse time Δ can decode the smaller-than-voxel-size brain structure, in particular, to distinguish scenarios (a) and (b). METHODS: High-resolution Q-ball images of a healthy brain taken with b=8000  s/mm2 for 3 different values of Δ were analyzed. The exchange of water molecules between crossing fibers was characterized by the fourth Fourier coefficient f4(Δ) of the signal profile in the plane of crossing. To interpret the empirical results, a model consisting of differently oriented parallel sheets of cylinders was developed. Diffusion of water molecules inside and outside cylinders was simulated by the Monte Carlo method. RESULTS: Simulations predict that f4(Δ) , agreeing with the empirical results, must increase with Δ for large b-values, but may peak at a typical Δ that depends on the thickness of the cylinder sheets for intermediate b-values. Thus, the thickness of axon layers in voxels with 2 predominant orientations can be detected from empirical f4(Δ) taken at smaller b-values. CONCLUSION: Based on the simulation results, recommendations are made on how to design a dMRI experiment with optimal b-value and range of Δ in order to measure the thickness of axon sheets in the white matter, hence to distinguish (a) and (b).

Geometric Navigation of Axons in a Cerebral Pathway: Comparing dMRI with Tract Tracing and Immunohistochemistry.

Brain fiber pathways are presumed to follow smooth curves but recent high angular resolution diffusion MRI (dMRI) suggests that instead they follow 3 primary axes often nearly orthogonal. To investigate this, we analyzed axon pathways under monkey primary motor cortex with (1) dMRI tractography, (2) axon tract tracing, and (3) axon immunohistochemistry. dMRI tractography shows the predicted crossings of axons in mediolateral and dorsoventral orientations and does not show axon turns in this region. Axons labeled with tract tracer in the motor cortex dispersed in the centrum semiovale by microscopically sharp axonal turns and/or branches (radii ≤15 µm) into 2 sharply defined orientations, mediolateral and dorsoventral. Nearby sections processed with SMI-32 antibody to label projection axons and SMI-312 antibody to label all axons revealed axon distributions parallel to the tracer axons. All 3 histological methods confirmed preponderant axon distributions parallel with dMRI axes with few axons (<20%) following smooth curves or diagonal orientations. These findings indicate that axons navigate deep white matter via microscopic sharp turns and branches between primary axes. They support dMRI observations of primary fiber axes, as well as the prediction that fiber crossings include navigational events not yet directly resolved by dMRI. New methods will be needed to incorporate coherent microscopic navigation into dMRI of connectivity.

Differentiating post-cancer from healthy tongue muscle coordination patterns during speech using deep learning.

The ability to differentiate post-cancer from healthy tongue muscle coordination patterns is necessary for the advancement of speech motor control theories and for the development of therapeutic and rehabilitative strategies. A deep learning approach is presented to classify two groups using muscle coordination patterns from magnetic resonance imaging (MRI). The proposed method uses tagged-MRI to track the tongue's internal tissue points and atlas-driven non-negative matrix factorization to reduce the dimensionality of the deformation fields. A convolutional neural network is applied to the classification task yielding an accuracy of 96.90%, offering the potential to the development of therapeutic or rehabilitative strategies in speech-related disorders.

Magnetic resonance imaging based anatomical assessment of tongue impairment due to amyotrophic lateral sclerosis: A preliminary study.

Amyotrophic Lateral Sclerosis (ALS) is a neurological disorder, which impairs tongue function for speech and swallowing. A widely used Diffusion Tensor Imaging (DTI) analysis pipeline is employed for quantifying differences in tongue fiber myoarchitecture between controls and ALS patients. This pipeline uses both high-resolution magnetic resonance imaging (hMRI) and DTI. hMRI is used to delineate tongue muscles, while DTI provides indices to reveal fiber connectivity within and between muscles. The preliminary results using five controls and two patients show quantitative differences between the groups. This work has the potential to provide insights into the detrimental effects of ALS on speech and swallowing.

HIgh b-value and high Resolution Integrated Diffusion (HIBRID) imaging.

The parameter selection for diffusion MRI experiments is dominated by the "k-q tradeoff" whereby the Signal to Noise Ratio (SNR) of the images is traded for either high spatial resolution (determined by the maximum k-value collected) or high diffusion sensitivity (effected by b-value or the q vector) but usually not both. Furthermore, different brain regions (such as gray matter and white matter) likely require different tradeoffs between these parameters due to the size of the structures to be visualized or the length-scale of the microstructure being probed. In this case, it might be advantageous to combine information from two scans - a scan with high q but low k (high angular resolution in diffusion but low spatial resolution in the image domain) to provide maximal information about white matter fiber crossing, and one low q but high k (low angular resolution but high spatial resolution) for probing the cortex. In this study, we propose a method, termed HIgh b-value and high Resolution Integrated Diffusion (HIBRID) imaging, for acquiring and combining the information from these two complementary types of scan with the goal of studying diffusion in the cortex without compromising white matter fiber information. The white-gray boundary and pial surface obtained from anatomical scans are incorporated as prior information to guide the fusion. We study the complementary advantages of the fused datasets, and assess the quality of the HIBRID data compared to either alone.

MGH-USC Human Connectome Project datasets with ultra-high b-value diffusion MRI.

The MGH-USC CONNECTOM MRI scanner housed at the Massachusetts General Hospital (MGH) is a major hardware innovation of the Human Connectome Project (HCP). The 3T CONNECTOM scanner is capable of producing a magnetic field gradient of up to 300 mT/m strength for in vivo human brain imaging, which greatly shortens the time spent on diffusion encoding, and decreases the signal loss due to T2 decay. To demonstrate the capability of the novel gradient system, data of healthy adult participants were acquired for this MGH-USC Adult Diffusion Dataset (N=35), minimally preprocessed, and shared through the Laboratory of Neuro Imaging Image Data Archive (LONI IDA) and the WU-Minn Connectome Database (ConnectomeDB). Another purpose of sharing the data is to facilitate methodological studies of diffusion MRI (dMRI) analyses utilizing high diffusion contrast, which perhaps is not easily feasible with standard MR gradient system. In addition, acquisition of the MGH-Harvard-USC Lifespan Dataset is currently underway to include 120 healthy participants ranging from 8 to 90 years old, which will also be shared through LONI IDA and ConnectomeDB. Here we describe the efforts of the MGH-USC HCP consortium in acquiring and sharing the ultra-high b-value diffusion MRI data and provide a report on data preprocessing and access. We conclude with a demonstration of the example data, along with results of standard diffusion analyses, including q-ball Orientation Distribution Function (ODF) reconstruction and tractography.

Microstructural impact of ischemia and bone marrow-derived cell therapy revealed with diffusion tensor magnetic resonance imaging tractography of the heart in vivo.

BACKGROUND: The arrangement of myofibers in the heart is highly complex and must be replicated by injected cells to produce functional myocardium. A novel approach to characterize the microstructural response of the myocardium to ischemia and cell therapy, with the use of serial diffusion tensor magnetic resonance imaging tractography of the heart in vivo, is presented. METHODS AND RESULTS: Validation of the approach was performed in normal (n=6) and infarcted mice (n=6) as well as healthy human volunteers. Mice (n=12) were then injected with bone marrow mononuclear cells 3 weeks after coronary ligation. In half of the mice the donor and recipient strains were identical, and in half the strains were different. A positive response to cell injection was defined by a decrease in mean diffusivity, an increase in fractional anisotropy, and the appearance of new myofiber tracts with the correct orientation. A positive response to bone marrow mononuclear cell injection was seen in 1 mouse. The response of the majority of mice to bone marrow mononuclear cell injection was neutral (9/12) or negative (2/12). The in vivo tractography findings were confirmed with histology. CONCLUSIONS: Diffusion tensor magnetic resonance imaging tractography was able to directly resolve the ability of injected cells to generate new myofiber tracts and provided a fundamental readout of their regenerative capacity. A highly novel and translatable approach to assess the efficacy of cell therapy in the heart is thus presented.

Blockface histology with optical coherence tomography: a comparison with Nissl staining.

Spectral domain optical coherence tomography (SD-OCT) is a high resolution imaging technique that generates excellent contrast based on intrinsic optical properties of the tissue, such as neurons and fibers. The SD-OCT data acquisition is performed directly on the tissue block, diminishing the need for cutting, mounting and staining. We utilized SD-OCT to visualize the laminar structure of the isocortex and compared cortical cytoarchitecture with the gold standard Nissl staining, both qualitatively and quantitatively. In histological processing, distortions routinely affect registration to the blockface image and prevent accurate 3D reconstruction of regions of tissue. We compared blockface registration to SD-OCT and Nissl, respectively, and found that SD-OCT-blockface registration was significantly more accurate than Nissl-blockface registration. Two independent observers manually labeled cortical laminae (e.g. III, IV and V) in SD-OCT images and Nissl stained sections. Our results show that OCT images exhibit sufficient contrast in the cortex to reliably differentiate the cortical layers. Furthermore, the modalities were compared with regard to cortical laminar organization and showed good agreement. Taken together, these SD-OCT results suggest that SD-OCT contains information comparable to standard histological stains such as Nissl in terms of distinguishing cortical layers and architectonic areas. Given these data, we propose that SD-OCT can be used to reliably generate 3D reconstructions of multiple cubic centimeters of cortex that can be used to accurately and semi-automatically perform standard histological analyses.

Probing dynamic myocardial microstructure with cardiac magnetic resonance diffusion tensor imaging.

This article is an invited editorial comment on the paper entitled "In vivo cardiovascular magnetic resonance diffusion tensor imaging shows evidence of abnormal myocardial laminar orientations and mobility in hypertrophic cardiomyopathy" by Ferreira et al., and published as Journal of Cardiovascular Magnetic Resonance 2014; 16:87.

Improved diffusion imaging through SNR-enhancing joint reconstruction.

Quantitative diffusion imaging is a powerful technique for the characterization of complex tissue microarchitecture. However, long acquisition times and limited signal-to-noise ratio represent significant hurdles for many in vivo applications. This article presents a new approach to reduce noise while largely maintaining resolution in diffusion weighted images, using a statistical reconstruction method that takes advantage of the high level of structural correlation observed in typical datasets. Compared to existing denoising methods, the proposed method performs reconstruction directly from the measured complex k-space data, allowing for gaussian noise modeling and theoretical characterizations of the resolution and signal-to-noise ratio of the reconstructed images. In addition, the proposed method is compatible with many different models of the diffusion signal (e.g., diffusion tensor modeling and q-space modeling). The joint reconstruction method can provide significant improvements in signal-to-noise ratio relative to conventional reconstruction techniques, with a relatively minor corresponding loss in image resolution. Results are shown in the context of diffusion spectrum imaging tractography and diffusion tensor imaging, illustrating the potential of this signal-to-noise ratio-enhancing joint reconstruction approach for a range of different diffusion imaging experiments.

Blipped-controlled aliasing in parallel imaging for simultaneous multislice echo planar imaging with reduced g-factor penalty.

Simultaneous multislice Echo Planar Imaging (EPI) acquisition using parallel imaging can decrease the acquisition time for diffusion imaging and allow full-brain, high-resolution functional MRI (fMRI) acquisitions at a reduced repetition time (TR). However, the unaliasing of simultaneously acquired, closely spaced slices can be difficult, leading to a high g-factor penalty. We introduce a method to create interslice image shifts in the phase encoding direction to increase the distance between aliasing pixels. The shift between the slices is induced using sign- and amplitude-modulated slice-select gradient blips simultaneous with the EPI phase encoding blips. This achieves the desired shifts but avoids an undesired "tilted voxel" blurring artifact associated with previous methods. We validate the method in 3× slice-accelerated spin-echo and gradient-echo EPI at 3 T and 7 T using 32-channel radio frequency (RF) coil brain arrays. The Monte-Carlo simulated average g-factor penalty of the 3-fold slice-accelerated acquisition with interslice shifts is <1% at 3 T (compared with 32% without slice shift). Combining 3× slice acceleration with 2× inplane acceleration, the g-factor penalty becomes 19% at 3 T and 10% at 7 T (compared with 41% and 23% without slice shift). We demonstrate the potential of the method for accelerating diffusion imaging by comparing the fiber orientation uncertainty, where the 3-fold faster acquisition showed no noticeable degradation.

Developing neocortex organization and connectivity in cats revealed by direct correlation of diffusion tractography and histology.

The immature cortex (cortical plate [CP]) and underlying subplate (SP), a transient cell layer just below the CP, play critical roles in the formation of intracerebral connections. The purpose of this study was to examine the diffusion characteristics of the developing cortex and subcortical structures and compare to histology. We obtained high-resolution diffusion spectrum images of postnatal day (P) 0 (newborn), P35 (pediatric), and P100 (adult) cat brains, performed tractography analysis, and correlated with histological findings. Tractography revealed radial organization and radial afferent/efferent tracts not only in the CP but also in external SP at P0. Radial organization persisted only in the cortex but decreased at P35 and P100. Radial organization correlated with radial cellular organization, with highest cellular density at P0 (Cresyl Violet staining). At P0, the internal SP contained abundant corticocortical and projection tractography pathways, crossing at wide angles in areas with no myelination by Luxol Fast Blue staining. At P35 and P100, increased directional coherence of white matter was observed, with fewer local tracts, but more long association pathways. These results suggest that diffusion tractography can differentially characterize internal and external SP zones and their transition into mature cortical pathways.

Measuring Strain in Diffusion-Weighted Data Using Tagged Magnetic Resonance Imaging.

Accurate strain measurement in a deforming organ has been essential in motion analysis using medical images. In recent years, internal tissue's in vivo motion and strain computation has been mostly achieved through dynamic magnetic resonance (MR) imaging. However, such data lack information on tissue's intrinsic fiber directions, preventing computed strain tensors from being projected onto a direction of interest. Although diffusion-weighted MR imaging excels at providing fiber tractography, it yields static images unmatched with dynamic MR data. This work reports an algorithm workflow that estimates strain values in the diffusion MR space by matching corresponding tagged dynamic MR images. We focus on processing a dataset of various human tongue deformations in speech. The geometry of tongue muscle fibers is provided by diffusion tractography, while spatiotemporal motion fields are provided by tagged MR analysis. The tongue's deforming shapes are determined by segmenting a synthetic cine dynamic MR sequence generated from tagged data using a deep neural network. Estimated motion fields are transformed into the diffusion MR space using diffeomorphic registration, eventually leading to strain values computed in the direction of muscle fibers. The method was tested on 78 time volumes acquired during three sets of specific tongue deformations including both speech and protrusion motion. Strain in the line of action of seven internal tongue muscles was extracted and compared both intra- and inter-subject. Resulting compression and stretching patterns of individual muscles revealed the unique behavior of individual muscles and their potential activation pattern.

Mapping the structural core of human cerebral cortex.

Structurally segregated and functionally specialized regions of the human cerebral cortex are interconnected by a dense network of cortico-cortical axonal pathways. By using diffusion spectrum imaging, we noninvasively mapped these pathways within and across cortical hemispheres in individual human participants. An analysis of the resulting large-scale structural brain networks reveals a structural core within posterior medial and parietal cerebral cortex, as well as several distinct temporal and frontal modules. Brain regions within the structural core share high degree, strength, and betweenness centrality, and they constitute connector hubs that link all major structural modules. The structural core contains brain regions that form the posterior components of the human default network. Looking both within and outside of core regions, we observed a substantial correspondence between structural connectivity and resting-state functional connectivity measured in the same participants. The spatial and topological centrality of the core within cortex suggests an important role in functional integration.

A deep joint sparse non-negative matrix factorization framework for identifying the common and subject-specific functional units of tongue motion during speech.

Intelligible speech is produced by creating varying internal local muscle groupings-i.e., functional units-that are generated in a systematic and coordinated manner. There are two major challenges in characterizing and analyzing functional units. First, due to the complex and convoluted nature of tongue structure and function, it is of great importance to develop a method that can accurately decode complex muscle coordination patterns during speech. Second, it is challenging to keep identified functional units across subjects comparable due to their substantial variability. In this work, to address these challenges, we develop a new deep learning framework to identify common and subject-specific functional units of tongue motion during speech. Our framework hinges on joint deep graph-regularized sparse non-negative matrix factorization (NMF) using motion quantities derived from displacements by tagged Magnetic Resonance Imaging. More specifically, we transform NMF with sparse and graph regularizations into modular architectures akin to deep neural networks by means of unfolding the Iterative Shrinkage-Thresholding Algorithm to learn interpretable building blocks and associated weighting map. We then apply spectral clustering to common and subject-specific weighting maps from which we jointly determine the common and subject-specific functional units. Experiments carried out with simulated datasets show that the proposed method achieved on par or better clustering performance over the comparison methods.Experiments carried out with in vivo tongue motion data show that the proposed method can determine the common and subject-specific functional units with increased interpretability and decreased size variability.

Development of cerebral fiber pathways in cats revealed by diffusion spectrum imaging.

Examination of the three-dimensional axonal pathways in the developing brain is key to understanding the formation of cerebral connectivity. By tracing fiber pathways throughout the entire brain, diffusion tractography provides information that cannot be achieved by conventional anatomical MR imaging or histology. However, standard diffusion tractography (based on diffusion tensor imaging, or DTI) tends to terminate in brain areas with low water diffusivity, indexed by low diffusion fractional anisotropy (FA), which can be caused by crossing fibers as well as fibers with less myelin. For this reason, DTI tractography is not effective for delineating the structural changes that occur in the developing brain, where the process of myelination is incomplete, and where crossing fibers exist in greater numbers than in the adult brain. Unlike DTI, diffusion spectrum imaging (DSI) can define multiple directions of water diffusivity; as such, diffusion tractography based on DSI provides marked flexibility for delineation of fiber tracts in areas where the fiber architecture is complex and multidirectional, even in areas of low FA. In this study, we showed that FA values were lower in the white matter of newborn (postnatal day 0; P0) cat brains than in the white matter of infant (P35) and juvenile (P100) cat brains. These results correlated well with histological myelin stains of the white matter: the newborn kitten brain has much less myelin than that found in cat brains at later stages of development. Using DSI tractography, we successfully identified structural changes in thalamo-cortical and cortico-cortical association tracts in cat brains from one stage of development to another. In newborns, the main body of the thalamo-cortical tract was smooth, and fibers branching from it were almost straight, while the main body became more complex and branching fibers became curved reflecting gyrification in the older cats. Cortico-cortical tracts in the temporal lobe were smooth in newborns, and they formed a sharper angle in the later stages of development. The cingulum bundle and superior longitudinal fasciculus became more visible with time. Within the first month after birth, structural changes occurred in these tracts that coincided with the formation of the gyri. These results show that DSI tractography has the potential for mapping morphological changes in low FA areas associated with growth and development. The technique may also be applicable to the study of other forms of brain plasticity, including future studies in vivo.

Identifying the Common and Subject-specific Functional Units of Speech Movements via a Joint Sparse Non-negative Matrix Factorization Framework.

The tongue is capable of producing intelligible speech because of successful orchestration of muscle groupings-i.e., functional units-of the highly complex muscles over time. Due to the different motions that tongues produce, functional units are transitional structures which transform muscle activity to surface tongue geometry and they vary significantly from one subject to another. In order to compare and contrast the location and size of functional units in the presence of such substantial inter-person variability, it is essential to study both common and subject-specific functional units in a group of people carrying out the same speech task. In this work, a new normalization technique is presented to simultaneously identify the common and subject-specific functional units defined in the tongue when tracked by tagged magnetic resonance imaging. To achieve our goal, a joint sparse non-negative matrix factorization framework is used, which learns a set of building blocks and subject-specific as well as common weighting matrices from motion quantities extracted from displacements. A spectral clustering technique is then applied to the subject-specific and common weighting matrices to determine the subject-specific functional units for each subject and the common functional units across subjects. Our experimental results using in vivo tongue motion data show that our approach is able to identify the common and subject-specific functional units with reduced size variability of tongue motion during speech.