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.

Group characterization of impact-induced, in vivo human brain kinematics.

Brain movement during an impact can elicit a traumatic brain injury, but tissue kinematics vary from person to person and knowledge regarding this variability is limited. This study examines spatio-temporal brain-skull displacement and brain tissue deformation across groups of subjects during a mild impact in vivo. The heads of two groups of participants were imaged while subjected to a mild (less than 350 rad s-2) impact during neck extension (NE, n = 10) and neck rotation (NR, n = 9). A kinematic atlas of displacement and strain fields averaged across all participants was constructed and compared against individual participant data. The atlas-derived mean displacement magnitude was 0.26 ± 0.13 mm for NE and 0.40 ± 0.26 mm for NR, which is comparable to the displacement magnitudes from individual participants. The strain tensor from the atlas displacement field exhibited maximum shear strain (MSS) of 0.011 ± 0.006 for NE and 0.017 ± 0.009 for NR and was lower than the individual MSS averaged across participants. The atlas illustrates common patterns, containing some blurring but visible relationships between anatomy and kinematics. Conversely, the direction of the impact, brain size, and fluid motion appear to underlie kinematic variability. These findings demonstrate the biomechanical roles of key anatomical features and illustrate common features of brain response for model evaluation.

Analysis of fiber strain in the human tongue during speech.

This study investigates mechanical cooperation among tongue muscles. Five volunteers were imaged using tagged magnetic resonance imaging to quantify spatiotemporal kinematics while speaking. Waveforms of strain in the line of action of fibers (SLAF) were estimated by projecting strain tensors onto a model of fiber directionality. SLAF waveforms were temporally aligned to determine consistency across subjects and correlation across muscles. The cohort exhibited consistent patterns of SLAF, and muscular extension-contraction was correlated. Volume-preserving tongue movement in speech generation can be achieved through multiple paths, but the study reveals similarities in motion patterns and muscular action-despite anatomical (and other) dissimilarities.

Quantitative Validation of MRI-Based Motion Estimation for Brain Impact Biomechanics.

Head impact can cause traumatic brain injury (TBI) through axonal overstretch or subsequent inflammation and understanding the biomechanics of the impact event is useful for TBI prevention research. Tagged magnetic resonance imaging (MRI) acquired during a mild-acceleration impact has enabled measurement and visualization of brain deformation in vivo. However, measurements using MRI are subject to error, and having independent validation while imaging in vivo is very difficult. Thus, characterizing the accuracy of these measurements needs to be done in a separate experiment using a phantom where a gold standard is available. This study describes a method for error quantification using a calibration phantom compatible with MRI and high-speed video (the gold standard). During linear acceleration, the maximum shear strain (MSS) in the phantom ranged from 0 to 12%, which is similar to in vivo brain deformation at a similar acceleration. The mean displacement error against video was 0.3±0.3 mm, and the MSS error was 1.4±0.3%. To match resolutions, video data was filtered temporally using an averaging filter. Compared to the unfiltered results, resolution matching improved the agreement between MRI and video results by 15%. In conclusion, tagged MRI analysis compares well to video data provided that resolutions are matched-a finding that is also applicable when using MRI to validate simulations.

In vivo estimates of axonal stretch and 3D brain deformation during mild head impact.

The rapid deformation of brain tissue in response to head impact can lead to traumatic brain injury. In vivo measurements of brain deformation during non-injurious head impacts are necessary to understand the underlying mechanisms of traumatic brain injury and compare to computational models of brain biomechanics. Using tagged magnetic resonance imaging (MRI), we obtained measurements of three-dimensional strain tensors that resulted from a mild head impact after neck rotation or neck extension. Measurements of maximum principal strain (MPS) peaked shortly after impact, with maximal values of 0.019-0.053 that correlated strongly with peak angular velocity. Subject-specific patterns of MPS were spatially heterogeneous and consistent across subjects for the same motion, though regions of high deformation differed between motions. The largest MPS values were seen in the cortical gray matter and cerebral white matter for neck rotation and the brainstem and cerebellum for neck extension. Axonal fiber strain (Ef) was estimated by combining the strain tensor with diffusion tensor imaging data. As with MPS, patterns of Ef varied spatially within subjects, were similar across subjects within each motion, and showed group differences between motions. Values were highest and most strongly correlated with peak angular velocity in the corpus callosum for neck rotation and in the brainstem for neck extension. The different patterns of brain deformation between head motions highlight potential areas of greater risk of injury between motions at higher loading conditions. Additionally, these experimental measurements can be directly compared to predictions of generic or subject-specific computational models of traumatic brain injury.

Quantifying Tongue Tip Shape in Apical and Laminal /s/: Contributions of Palate Shape.

Purpose Anterior tongue shape during /s/ production is often described as "tip-up" or apical, versus "tip-down" or laminal. Typically, this is determined by observing the shape of the anterior midline tongue. The purpose of this study was to identify methods of curvature calculation that quantify the observed shape differences and to examine whether the shape differences were affected by palate shape. Previous work shows that palate height has some effect (Grimm et al., 2017). Method Four curvature-based measures were applied to a series of points selected along the tongue surface in midsagittal cine magnetic resonance images during speech. The measures were minimal curvature, averaged largest curvature (ALC), normalized ALC, and interpolated normalized ALC. These measures were compared to visual judgments of apical and laminal /s/. Anterior palate shape was measured from dental casts. Results The apical /s/ contained a flat or concave region in the anterior tongue, while the laminal /s/ had a convex shape along the entire tongue. Thus, the laminal shape was less complex than the apical. The last 2 metrics, based on averages of multiple normalized curvatures, captured this complexity difference. Subjects with a more steeply sloped anterior palate tended to use laminal /s/. Conclusions The tongue shape for the 2 /s/ types was best defined by complexity of the shape, rather than local anterior shape. Statistical quantities that measured curvature in multiple locations, and normalized across subjects, were best at distinguishing the 2 /s/ shapes. Interpolating additional points between the manually selected ones did not improve the method. Supplemental Material https://doi.org/10.23641/asha.9733709.

A Sparse Non-Negative Matrix Factorization Framework for Identifying Functional Units of Tongue Behavior From MRI.

Muscle coordination patterns of lingual behaviors are synergies generated by deforming local muscle groups in a variety of ways. Functional units are functional muscle groups of local structural elements within the tongue that compress, expand, and move in a cohesive and consistent manner. Identifying the functional units using tagged-magnetic resonance imaging (MRI) sheds light on the mechanisms of normal and pathological muscle coordination patterns, yielding improvement in surgical planning, treatment, or rehabilitation procedures. In this paper, to mine this information, we propose a matrix factorization and probabilistic graphical model framework to produce building blocks and their associated weighting map using motion quantities extracted from tagged-MRI. Our tagged-MRI imaging and accurate voxel-level tracking provide previously unavailable internal tongue motion patterns, thus revealing the inner workings of the tongue during speech or other lingual behaviors. We then employ spectral clustering on the weighting map to identify the cohesive regions defined by the tongue motion that may involve multiple or undocumented regions. To evaluate our method, we perform a series of experiments. We first use two-dimensional images and synthetic data to demonstrate the accuracy of our method. We then use three-dimensional synthetic and in vivo tongue motion data using protrusion and simple speech tasks to identify subject-specific and data-driven functional units of the tongue in localized regions.

3-D Measurements of Acceleration-Induced Brain Deformation via Harmonic Phase Analysis and Finite-Element Models.

OBJECTIVE: To obtain dense spatiotemporal measurements of brain deformation from two distinct but complementary head motion experiments: linear and rotational accelerations. METHODS: This study introduces a strategy for integrating harmonic phase analysis of tagged magnetic resonance imaging (MRI) and finite-element models to extract mechanically representative deformation measurements. The method was calibrated using simulated as well as experimental data, demonstrated in a phantom including data with image artifacts, and used to measure brain deformation in human volunteers undergoing rotational and linear acceleration. RESULTS: Evaluation methods yielded a displacement error of 1.1 mm compared to human observers and strain errors between [Formula: see text] for linear acceleration and [Formula: see text] for rotational acceleration. This study also demonstrates an approach that can reduce error by 86% in the presence of corrupted data. Analysis of results shows consistency with 2-D motion estimation, agreement with external sensors, and the expected physical behavior of the brain. CONCLUSION: Mechanical regularization is useful for obtaining dense spatiotemporal measurements of in vivo brain deformation under different loading regimes. SIGNIFICANCE: The measurements suggest that the brain's 3-D response to mild accelerations includes distinct patterns observable using practical MRI resolutions. This type of measurement can provide validation data for computer models for the study of traumatic brain injury.

A Novel Filtering Approach for 3D Harmonic Phase Analysis of Tagged MRI.

Harmonic phase analysis has been used to perform noninvasive organ motion and strain estimation using tagged magnetic resonance imaging (MRI). The filtering process, which is used to produce harmonic phase images used for tissue tracking, influences the estimation accuracy. In this work, we evaluated different filtering approaches, and propose a novel high-pass filter for volumes tagged in individual directions. Testing was done using an open benchmarking dataset and synthetic images obtained using a mechanical model. We compared estimation results from our filtering approach with results from the traditional filtering approach. Our results indicate that 1) the proposed high-pass filter outperforms the traditional filtering approach reducing error by as much as 50% and 2) the accuracy improvements are especially marked in complex deformations.

Inverse Biomechanical Modeling of the Tongue via Machine Learning and Synthetic Training Data.

The tongue's deformation during speech can be measured using tagged magnetic resonance imaging, but there is no current method to directly measure the pattern of muscles that activate to produce a given motion. In this paper, the activation pattern of the tongue's muscles is estimated by solving an inverse problem using a random forest. Examples describing different activation patterns and the resulting deformations are generated using a finite-element model of the tongue. These examples form training data for a random forest comprising 30 decision trees to estimate contractions in 262 contractile elements. The method was evaluated on data from tagged magnetic resonance data from actual speech and on simulated data mimicking flaps that might have resulted from glossectomy surgery. The estimation accuracy was modest (5.6% error), but it surpassed a semi-manual approach (8.1% error). The results suggest that a machine learning approach to contraction pattern estimation in the tongue is feasible, even in the presence of flaps.

Laplace-based modeling of fiber orientation in the tongue.

Mechanical modeling of tongue deformation plays a significant role in the study of breathing, swallowing, and speech production. In the absence of internal joints, fiber orientations determine the direction of sarcomeric contraction and have great influence over real and simulated tissue motion. However, subject-specific experimental observations of fiber distribution are difficult to obtain; thus, models of fiber distribution are generally used in mechanical simulations. This paper describes modeling of fiber distribution using solutions of Laplace equations and compares the effectiveness of this approach against tractography from diffusion tensor magnetic resonance imaging. The experiments included qualitative comparison of streamlines from the fiber model against experimental tractography, as well as quantitative differences between biomechanical simulations focusing in the region near the genioglossus. The model showed good overall agreement in terms of fiber directionality and muscle positioning when compared to subject-specific imaging results and the literature. The angle between the fiber distribution model against tractography in the genioglossus and geniohyoid muscles averaged [Formula: see text] likely due to experimental noise. However, kinematic responses were similar between simulations with modeled fibers versus experimentally obtained fibers; average discrepancy in surface displacement ranged from 1 to 7 mm, and average strain residual magnitude ranged from [Formula: see text] to 0.2. The results suggest that, for simulation purposes, the modeled fibers can act as a reasonable approximation for the tongue's fiber distribution. Also, given its agreement with the global tongue anatomy, the approach may be used in model-based reconstruction of displacement tracking and diffusion results.

Quantifying Tensor Field Similarity With Global Distributions and Optimal Transport.

Strain tensor fields quantify tissue deformation and are important for functional analysis of moving organs such as the heart and the tongue. Strain data can be readily obtained using medical imaging. However, quantification of similarity between different data sets is difficult. Strain patterns vary in space and time, and are inherently multidimensional. Also, the same type of mechanical deformation can be applied to different shapes; hence, automatic quantification of similarity should be unaffected by the geometry of the objects being deformed. This work introduces the application of global distributions used to classify shapes and vector fields in the pattern recognition literature, in the context of tensorial strain data. In particular, the distribution of mechanical properties of a field are approximated using a 3D histogram, and the Wasserstein distance from optimal transport theory is used to measure the similarity between histograms. To measure the method's consistency in matching deformations across different objects, the proposed approach was evaluated by sorting strain fields according to their similarity. Performance was compared to sorting via maximum shear distribution (a 1D histogram) and tensor residual magnitude (in perfectly registered objects). The technique was also applied to correlate muscle activation to muscular contraction observed via tagged MRI. The results show that the proposed approach accurately matches deformation regardless of the shape of the object being deformed. Sorting accuracy surpassed 1D shear distribution and was on par with residual magnitude, but without the need for registration between objects.

Test Suite for Image-Based Motion Estimation of the Brain and Tongue.

Noninvasive analysis of motion has important uses as qualitative markers for organ function and to validate biomechanical computer simulations relative to experimental observations. Tagged MRI is considered the gold standard for noninvasive tissue motion estimation in the heart, and this has inspired multiple studies focusing on other organs, including the brain under mild acceleration and the tongue during speech. As with other motion estimation approaches, using tagged MRI to measure 3D motion includes several preprocessing steps that affect the quality and accuracy of estimation. Benchmarks, or test suites, are datasets of known geometries and displacements that act as tools to tune tracking parameters or to compare different motion estimation approaches. Because motion estimation was originally developed to study the heart, existing test suites focus on cardiac motion. However, many fundamental differences exist between the heart and other organs, such that parameter tuning (or other optimization) with respect to a cardiac database may not be appropriate. Therefore, the objective of this research was to design and construct motion benchmarks by adopting an "image synthesis" test suite to study brain deformation due to mild rotational accelerations, and a benchmark to model motion of the tongue during speech. To obtain a realistic representation of mechanical behavior, kinematics were obtained from finite-element (FE) models. These results were combined with an approximation of the acquisition process of tagged MRI (including tag generation, slice thickness, and inconsistent motion repetition). To demonstrate an application of the presented methodology, the effect of motion inconsistency on synthetic measurements of head-brain rotation and deformation was evaluated. The results indicated that acquisition inconsistency is roughly proportional to head rotation estimation error. Furthermore, when evaluating non-rigid deformation, the results suggest that inconsistent motion can yield "ghost" shear strains, which are a function of slice acquisition viability as opposed to a true physical deformation.

Phase Vector Incompressible Registration Algorithm for Motion Estimation From Tagged Magnetic Resonance Images.

Tagged magnetic resonance imaging has been used for decades to observe and quantify motion and strain of deforming tissue. It is challenging to obtain 3-D motion estimates due to a tradeoff between image slice density and acquisition time. Typically, interpolation methods are used either to combine 2-D motion extracted from sparse slice acquisitions into 3-D motion or to construct a dense volume from sparse acquisitions before image registration methods are applied. This paper proposes a new phase-based 3-D motion estimation technique that first computes harmonic phase volumes from interpolated tagged slices and then matches them using an image registration framework. The approach uses several concepts from diffeomorphic image registration with a key novelty that defines a symmetric similarity metric on harmonic phase volumes from multiple orientations. The material property of harmonic phase solves the aperture problem of optical flow and intensity-based methods and is robust to tag fading. A harmonic magnitude volume is used in enforcing incompressibility in the tissue regions. The estimated motion fields are dense, incompressible, diffeomorphic, and inverse-consistent at a 3-D voxel level. The method was evaluated using simulated phantoms, human brain data in mild head accelerations, human tongue data during speech, and an open cardiac data set. The method shows comparable accuracy to three existing methods while demonstrating low computation time and robustness to tag fading and noise.

Right Ventricular Fiber Structure as a Compensatory Mechanism in Pressure Overload: A Computational Study.

Right ventricular failure (RVF) is a lethal condition in diverse pathologies. Pressure overload is the most common etiology of RVF, but our understanding of the tissue structure remodeling and other biomechanical factors involved in RVF is limited. Some remodeling patterns are interpreted as compensatory mechanisms including myocyte hypertrophy, extracellular fibrosis, and changes in fiber orientation. However, the specific implications of these changes, especially in relation to clinically observable measurements, are difficult to investigate experimentally. In this computational study, we hypothesized that, with other variables constant, fiber orientation alteration provides a quantifiable and distinct compensatory mechanism during RV pressure overload (RVPO). Numerical models were constructed using a rabbit model of chronic pressure overload RVF based on intraventricular pressure measurements, CINE magnetic resonance imaging (MRI), and diffusion tensor MRI (DT-MRI). Biventricular simulations were conducted under normotensive and hypertensive boundary conditions using variations in RV wall thickness, tissue stiffness, and fiber orientation to investigate their effect on RV pump function. Our results show that a longitudinally aligned myocardial fiber orientation contributed to an increase in RV ejection fraction (RVEF). This effect was more pronounced in response to pressure overload. Likewise, models with longitudinally aligned fiber orientation required a lesser contractility for maintaining a target RVEF against elevated pressures. In addition to increased wall thickness and material stiffness (diastolic compensation), systolic mechanisms in the forms of myocardial fiber realignment and changes in contractility are likely involved in the overall compensatory responses to pressure overload.

Accurate high-resolution measurements of 3-D tissue dynamics with registration-enhanced displacement encoded MRI.

Displacement fields are important to analyze deformation, which is associated with functional and material tissue properties often used as indicators of health. Magnetic resonance imaging (MRI) techniques like DENSE and image registration methods like Hyperelastic Warping have been used to produce pixel-level deformation fields that are desirable in high-resolution analysis. However, DENSE can be complicated by challenges associated with image phase unwrapping, in particular offset determination. On the other hand, Hyperelastic Warping can be hampered by low local image contrast. The current work proposes a novel approach for measuring tissue displacement with both DENSE and Hyperelastic Warping, incorporating physically accurate displacements obtained by the latter to improve phase characterization in DENSE. The validity of the proposed technique is demonstrated using numerical and physical phantoms, and in vivo small animal cardiac MRI.