Evaluation of matrix factorisation approaches for muscle synergy extraction

Ebied, Ahmed; Kinney‐Lang, Eli; Spyrou, Loukianos; Escudero, Javier

doi:10.1016/j.medengphy.2018.04.003

Cited by 61 publications

(61 citation statements)

References 58 publications

(86 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nonnegative matrix factorization was introduced as a tool to recognize patterns for the decomposition of images into meaningful features (Lee and Seung, 1999), spectral data analysis or denoising of audio signals. In biology, it has been applied to a variety of datasets ranging from EEG based imaging (Lu and Yin, 2015; Delis et al, 2016) and muscle electrographs to transcriptomic datasets from microarrays or RNA-Seq experiments (Kong et al, 2011; Zhang et al, 2016; Shao and Höfer, 2017; Duren et al, 2018; Ebied et al, 2018). Regarding transcriptome data, the initial idea was published by Brunet et al, (2004).…”

Section: Discussionmentioning

confidence: 99%

Transcriptome dynamics in developing leaves from C₃and C₄Flaveriaspecies reveal determinants of Kranz anatomy

Billakurthi

Wrobel

Bräutigam

et al. 2018

Preprint

View full text Add to dashboard Cite

C4 species have evolved more than 60 times independently from C3 ancestors. This multiple and parallel evolution of the complex C4 trait indicates common underlying evolutionary mechanisms that might be identified by comparative analysis of closely related C3 and C4 species. Efficient C4 function depends on a distinctive leaf anatomy that is characterized by enlarged, chloroplast rich bundle sheath cells and a narrow vein spacing. To elucidate molecular mechanisms generating this so called Kranz anatomy, we analyzed a developmental series of leaves from the C4 plant Flaveria bidentis and the closely related C3 species Flaveria robusta using leaf clearing and whole transcriptome sequencing. Applying non-negative matrix factorization on the data identified four different zones with distinct transcriptome patterns in growing leaves of both species. Comparing these transcriptome patterns revealed an important role of auxin metabolism and especially auxin homeostasis for establishing the high vein density typical for C4 leaves.2012). Under current ambient CO2 concentrations (405 ppm) at 25°C, photorespiration is estimated to decrease the yield of soybean or wheat in the US by 36% and 20%, respectively (Walker et al., 2016). Environmental constrains such as high temperatures and drought further increases Rubisco oxygenase activity (Laing et al., 1974; Jordan and Ogren, 1984; Brooks and Farquhar, 1985; Parry et al., 2007).In most C4 plants CO2 fixation is compartmentalized between two cell types, the bundle sheath (BS) and the mesophyll (M) cells. In the mesophyll phosphoenolpyruvate (PEP) is carboxylated by phosphoenolpyruvate carboxylase (PEPC), resulting in the 4-carbon compound oxaloacetate (OAA). OAA is converted to malate and/or aspartate, which is then transferred to the bundle sheath. Here the 4-carbon compounds are decarboxylated and the released CO2 is assimilated in the Calvin-Benson-Bassham cycle (CBB). The resulting pyruvate is transferred back to the mesophyll where the primary CO2 acceptor PEP is regenerated by pyruvate orthophosphate dikinase (PPDK) (Hatch, 1987).C4 photosynthesis requires a particular leaf anatomy. As BS and M cells operate as a photosynthetic unit, direct contact of both cell types is necessary to ensure efficient photosynthesis.The BS is composed of the cells directly adjacent to the vasculature. Therefore, leaves of most C4 species exhibit high vein densities with a characteristic pattern in which two veins, each surrounded by BS cells, are separated by only two layers of M cells in a vein-bundle sheathmesophyll-mesophyll-bundle sheath-vein layout. Bundle sheath cells of C4 plants often appear larger in cross-section compared to C3 species and contain more chloroplasts. This character-

show abstract

Section: Discussionmentioning

confidence: 99%

Transcriptome dynamics in developing leaves from C₃and C₄Flaveriaspecies reveal determinants of Kranz anatomy

Billakurthi

Wrobel

Bräutigam

et al. 2018

Preprint

View full text Add to dashboard Cite

show abstract

“…filtering parameters, normalization methods), assumptions about neural control complexity (e.g. number of synergies, number and choice of muscles, synergy vector variability across trials), numerical analysis approaches (e.g., matrix decomposition algorithm, selection criteria for number of synergies), and post-processing of results (Tresch et al, 2006;Hug et al, 2012;Steele et al, 2013;Oliveira et al, 2014;Shourijeh et al, 2016;Banks et al, 2017;Shuman et al, 2017;Ebied et al, 2018;Gallina et al, 2018;Mehryar et al, 2020). In the present study, for each subset of measured muscles, we performed synergy extrapolation using a total of 80 methodological combinations comprised of 2 algorithms for matrix factorization, 5 methods for EMG normalization, and 8 choices for number of muscle synergies.…”

Section: Methodological Choices For Synergy Extrapolationmentioning

confidence: 99%

“…number of synergies), and numerical analysis approaches (e.g. matrix decomposition algorithm (Tresch et al, 2006;Hug et al, 2012;Steele et al, 2013;Oliveira et al, 2014;Banks et al, 2017;Shuman et al, 2017;Ebied et al, This is a provisional file, not the final typeset article 2018; Gallina et al, 2018)), we also evaluated how each of these choices affects synergy extrapolation results. The evaluated methodological decisions for MSA process included matrix factorization algorithm (PCA and NMF), EMG normalization approach, and number of synergies.…”

Section: Introductionmentioning

confidence: 99%

Evaluation of Synergy Extrapolation for Predicting Unmeasured Muscle Excitations from Measured Muscle Synergies

Shourijeh

Patten

et al. 2020

Preprint

View full text Add to dashboard Cite

Electromyography (EMG)-driven musculoskeletal modeling relies on high-quality measurements of muscle electrical activity to estimate muscle forces. However, a critical challenge for practical deployment of this approach is missing EMG data from muscles that contribute substantially to joint moments. This situation may arise due to either the inability to measure deep muscles with surface electrodes or the lack of a sufficient number of EMG electrodes. Muscle synergy analysis is a dimensionality-reduction approach to decompose a large number of muscle excitations into a small number of time-varying synergy excitations along with time-invariant synergy weights that define the contribution of each corresponding synergy excitation to a specific muscle excitation. This study evaluates how accurately missing muscle excitations can be predicted using synergy excitations extracted from muscles with available EMGs (henceforth called “synergy extrapolation”). The results were reported on a gait dataset collected from a stroke survivor walking on an instrumented treadmill at self-selected and fastest-comfortable speeds. The evaluation process started with full calibration of a lower-body EMG-driven model using 16-channel EMGs (including surface and indwelling) in each leg. One indwelling EMG (either iliopsoas or adductor longus) was then treated as unmeasured at a time. The synergy weights associated with the unmeasured muscle were predicted through solving a nonlinear optimization problem where the errors between inverse dynamics and EMG-driven joint moments were minimized. We also quantitatively evaluated how synergy analysis algorithms (principal component analysis (PCA) and non-negative matrix factorization (NMF)), EMG normalization methods, and number of synergies affect the accuracy of the predicted unmeasured muscle excitation. Synergy extrapolation performance was most influenced by the choice of synergy analysis algorithm and number of synergies. PCA with 5 or 6 synergies consistently predicted unmeasured muscle excitations most accurately and with greatest robustness to choice of EMG normalization method. Furthermore, the associated joint moment matching accuracy was comparable to that produced by the full EMG-driven calibration. The synergy extrapolation method described in this study may facilitate the assessment of human neuromuscular control and biomechanics in response to surgical or rehabilitation treatment when important EMG signals are missing.

show abstract

“…On the other hand, the number of synergies for 2 ndorder model extracted via matrix factorisation methods have been determined using two main approaches: a functional approach, and a mathematical approach [53]. The functional approach relies on prior knowledge of data structure and myoelectric control requirements to choose the appropriate number of synergies.…”

Section: Tensor Decomposition Models For Muscle Synergy Analysis 1mentioning

confidence: 99%

Muscle Activity Analysis Using Higher-Order Tensor Decomposition: Application to Muscle Synergy Extraction

et al. 2019

Self Cite

View full text Add to dashboard Cite

Higher-order tensor decompositions have hardly been used in muscle activity analysis despite multichannel electromyography (EMG) datasets naturally occurring as multi-way structures. Here, we seek to demonstrate and discuss the potential of tensor decompositions as a framework to estimate muscle synergies from 3 rd -order EMG tensors built by stacking repetitions of multi-channel EMG for several tasks. We compare the two most widespread tensor decomposition models -Parallel Factor Analysis (PARAFAC) and Tucker -in muscle synergy analysis of the wrist's three main Degree of Freedoms (DoFs) using the public first Ninapro database. Furthermore, we proposed a constrained Tucker decomposition (consTD) method for efficient synergy extraction building on the power of tensor decompositions. This method is proposed as a direct novel approach for shared and task-specific synergy estimation from two biomechanically related tasks. Our approach is compared with the current standard approach of repetitively applying non-negative matrix factorisation (NMF) to a series of the movements. The results show that the consTD method is suitable for synergy extraction compared to PARAFAC and Tucker. Moreover, exploiting the multi-way structure of muscle activity, the proposed methods successfully identified shared and taskspecific synergies for all three DoFs tensors. These were found to be robust to disarrangement with regard to task-repetition information, unlike the commonly used NMF. In summary, we demonstrate how to use tensors to characterise muscle activity and develop a new consTD method for muscle synergy extraction that could be used for shared and task-specific synergies identification. We expect that this study will pave the way for the development of novel muscle activity analysis methods based on higher-order techniques.

show abstract

Evaluation of matrix factorisation approaches for muscle synergy extraction

Cited by 61 publications

References 58 publications

Transcriptome dynamics in developing leaves from C₃and C₄Flaveriaspecies reveal determinants of Kranz anatomy

Transcriptome dynamics in developing leaves from C₃and C₄Flaveriaspecies reveal determinants of Kranz anatomy

Evaluation of Synergy Extrapolation for Predicting Unmeasured Muscle Excitations from Measured Muscle Synergies

Muscle Activity Analysis Using Higher-Order Tensor Decomposition: Application to Muscle Synergy Extraction

Contact Info

Product

Resources

About

Evaluation of matrix factorisation approaches for muscle synergy extraction

Cited by 61 publications

References 58 publications

Transcriptome dynamics in developing leaves from C3and C4Flaveriaspecies reveal determinants of Kranz anatomy

Transcriptome dynamics in developing leaves from C3and C4Flaveriaspecies reveal determinants of Kranz anatomy

Evaluation of Synergy Extrapolation for Predicting Unmeasured Muscle Excitations from Measured Muscle Synergies

Muscle Activity Analysis Using Higher-Order Tensor Decomposition: Application to Muscle Synergy Extraction

Contact Info

Product

Resources

About

Transcriptome dynamics in developing leaves from C₃and C₄Flaveriaspecies reveal determinants of Kranz anatomy

Transcriptome dynamics in developing leaves from C₃and C₄Flaveriaspecies reveal determinants of Kranz anatomy