A Tensor Decomposition Uncovers the Effect of Ageing on Muscle and Grip-Load Force Couplings During Grasping

Chang Ye, Seyed Saman Saboksayr, William J. Shaw, Rachel O. Coats, Sarah Astill, Gonzalo Mateos, Ioannis DelisbioRxiv 2022 · DOI

TL;DR

This paper uses a novel tensor decomposition approach to analyze the complex interplay between muscle activity and grip/load forces during grasping in both young and older adults. They discovered that aging leads to altered muscle coordination patterns, suggesting a compensatory strategy for declining manual dexterity.

What Problem Does This Paper Address?

As we age, our motor control abilities naturally decline, impacting everyday tasks like grasping and lifting objects. While previous research has identified age-related differences in grip and load forces, the underlying neuromuscular mechanisms – how muscles coordinate to produce these forces – have remained largely unclear. This is particularly true for bimanual tasks, which are ecologically relevant but less studied than unimanual ones. The authors aim to bridge this gap by investigating how muscle activation patterns and their coupling with grip and load forces change with age during bimanual grasping.

Methods

The core of this study is the application of Non-negative Canonical Polyadic (NCP) decomposition, a tensor factorization technique. They collected data from 30 participants (15 young, 15 old) performing bimanual grasp-to-lift tasks with objects of different weights. This data, comprising muscle activity (EMG) from several upper limb muscles and grip/load forces, was organized into a high-order tensor (up to 5-way: spatial, temporal, object, participant, trial).

NCP decomposition breaks down this complex tensor into a set of "components" or "factors." Each component represents a fundamental motor pattern, characterized by:

  • Spatial component: Which muscles and forces are involved.
  • Temporal component: How the activation unfolds over time.
  • Object component: How the pattern is used for different object weights.
  • Participant component: How strongly each participant recruits that pattern.
  • Trial component (in the 5-way model): How the pattern varies across repetitions.

The researchers then used these participant-specific components to predict age and manual dexterity using linear regression and classification analyses.

Datasets

The study utilized a dataset collected from 30 human participants performing a bimanual grasp-to-lift task. This dataset includes:

  • Muscle activity (EMG): From Anterior Deltoid, Flexor Carpi Radialis (FCR), Extensor Carpi Radialis (ECR), and Abductor Pollicis Brevis (APB) of both arms.
  • Kinetic variables: Grip force (GF) and load force (LF) for both hands.
  • Object properties: Two different weights (heavy and light).
  • Participant characteristics: Age and manual dexterity (measured by the Purdue Pegboard test).

No new benchmark datasets were introduced; the study used data collected specifically for this research.

Key Insights and Ideas

  • Aging alters muscle synergy "recipes": The study suggests that aging doesn't necessarily mean muscles stop working, but rather the way they are combined changes. This is seen in the differential recruitment of specific motor components by older adults.
  • Compensatory strategies for dexterity loss: The findings point to a potential compensatory mechanism where older adults might increase muscle coupling (e.g., between forearm and hand muscles) and exert higher grip forces to compensate for reduced manual dexterity. This is a less "economic" strategy but helps maintain task performance.
  • Tensor decomposition as a powerful tool: The paper highlights the efficacy of tensor decomposition in unraveling complex, multi-modal biological data, revealing underlying patterns that would be missed by traditional, lower-dimensional analyses.

Key Results

  • Age-discriminating components: A classification analysis successfully predicted age group (young vs. old) with high accuracy (AUC = 0.87, Acc = 0.70) using a combination of the first two identified motor components.
  • Component 1: The "aging" component: This component, characterized by coupled forearm (ECR) and hand (APB) muscle activation and higher grip forces, was more strongly recruited by older adults. It was also negatively correlated with manual dexterity (rho = -0.57) and positively correlated with age in the older adult group (rho = 0.63).
  • Component 2: The "younger" component: This component showed higher activation in younger adults and a more economical grip-to-load force ratio.
  • Predictive power: Advanced age and poorer manual dexterity could be predicted by the coupled activation of forearm and hand muscles associated with high grip force.

Limitations and Future Work

  • Limitations: The study acknowledges that the tensor decomposition has inherent scaling ambiguities, which were addressed through normalization. While the 4-way model (trial-averaged) provided robust results, the 5-way model (including trials) showed slightly noisier data, as expected.
  • Future Work: The authors suggest that this tensor decomposition methodology can be extended to analyze other motor behavioral experiments and to investigate differences in motor signals between healthy and impaired individuals (e.g., stroke patients), potentially informing rehabilitation strategies. They also propose using tensor structure to test specific hypotheses about motor control.

Summary updated

Abstract

ABSTRACT Do motor patterns of grasp-to-lift movements change as a result of ageing? Previous studies often relied on simple temporal and kinetic variables to unveil differences caused by ageing, yet their neuromuscular origins remain largely unknown. Here we employed a bimanual grasping protocol with younger and older adults and combined measurements of muscle activity with grip and load forces to provide a window into the neuromuscular strategies underlying effective grasping. We introduced a tensor decomposition to identify patterns of muscle activity and grip-load force ratios while also characterising their temporal profiles and relative activation across object weights and participants of different age groups. This approach extracted the motor components underpinning object grasping across participants. We then probed age-induced changes in these components. A classification analysis revealed two motor components that are differentially recruited between the two age groups. Linear regression analyses further showed that advanced age and poorer manual dexterity can be predicted by the coupled activation of forearm and hand muscles which is associated with high levels of grip force. Our findings suggest that ageing may induce stronger muscle couplings in distal aspects of the upper limbs, and a less economic grasping strategy to overcome age-related decline in manual dexterity.
BibTeX
@misc{ye2022tensor,
  author = {Chang Ye and Seyed Saman Saboksayr and William J. Shaw and Rachel O. Coats and Sarah Astill and Gonzalo Mateos and Ioannis Delis},
  title = {{A Tensor Decomposition Uncovers the Effect of Ageing on Muscle and Grip-Load Force Couplings During Grasping}},
  year = {2022},
  doi = {10.1101/2022.02.25.482040},
  howpublished = {bioRxiv},
}
Figure