Multi-Modal Deformation Sensing for Evaluating Muscle Function

Abstract

In the evolving landscape of biomechanics and rehabilitation science, understanding muscle function and the mechanisms behind muscle contractile dynamics remains a cornerstone for advancing therapeutic interventions, developing new musculoskeletal sensing methodologies, and enhancing human motor performance. Central to all muscle function is the ability to generate force. This fundamental attribute not only coordinates movement but also underpins the many regulatory mechanisms that muscles support. Importantly, the concept of muscle force extends beyond simple motion, allowing us to explore more specific physiological parameters such as fatigue, which reflects a change in force over time, or central drive, which measures the ratio between volitionally-generated force and the maximum force-generating ability of a muscle. When muscles contract, they deform---and this deformation signal can be measured using ultrasound to estimate muscle force during a contraction. Some sensing approaches instead measure external, surface-level muscle deformation resulting from internal changes in muscle architecture propagating through soft tissue to the surface of the skin. This thesis seeks to bridge critical gaps in our current understanding of muscle function by developing innovative methodologies for analyzing and evaluating muscle dynamics. At the heart of this investigation is the premise that muscle deformation, especially when combined with electrical stimulation, can provide comprehensive insights into muscle function, offering a new perspective on the biomechanical and neural basis of muscle performance and fatigue.

This thesis starts by characterizing and validating surface-level muscle deformation, as a correlate for joint torque using a highly accurate, non-contact, 2D laser profilometer. We show that three metrics of surface-level muscle deformation---peak radial displacement: $r$ = 0.94$\pm$0.05, surface curvature: $r$ = 0.78$\pm$0.10, and surface strain: $r$ = 0.83$\pm$0.12---correlate with elbow joint torque across a range of elbow angles and measurement locations. We also find that surface-level muscle deformation leads torque generation but lags torque relaxation during volitional contractions. Finally, we demonstrate that unlike peak radial displacement and surface strain, surface curvature is not significantly affected by changes in joint angle or measurement location, exhibiting strong and consistent correlations with joint torque and range of responses across all conditions, underscoring its potential as a candidate signal for wearable estimation of joint torque during dynamic activities.

Building upon the systematic characterization of surface-level muscle deformation as a correlate for force generation, we next show how ultra-sensitive soft strain gauges can be used as wearable sensors to measure changes in surface curvature. We demonstrate how the soft sensors can track plantarflexion torque by monitoring deformation of the skin above the gastrocnemius and elbow flexion torque (monitoring the biceps brachii) during volitional contractions on a dynamometer. We then show how soft sensors strongly correlate to knee extension torque ($r^2$ = 0.93$\pm$0.05) when monitoring rectus femoris deformation on eight participants. Next, to explore the viability of using sensors in more real-world scenarios, we use the sensors to measure deformation to estimate muscle coordination (peak force timing) of the quadriceps muscles during stationary cycling at different cadences and to exhaustion. The results show that the deformation-derived timing metrics are more closely aligned with ground-truth peak force application measured by the cycle ergometer as compared to surface electromyography-derived coordination metrics.

Building on our research looking at surface measures of muscle deformation, we next explore the potential of using A-Mode ultrasound combined with electrical stimulation to estimate muscle fatigue. In particular, we look to address the volition-induced variability of traditional fatigue assessment by bypassing the central nervous system and directly activating muscles with electrical stimulation. We demonstrate how stimulation-induced changes in muscle thickness measured via A-Mode ultrasound correlate with simultaneously generated torque and fatigue-related deficits in overall volitional force. Through a dynamic, volitional knee extensor fatiguing protocol, we show how this combination can provide targeted and highly correlated measures of fatigue when compared to ground-truth torque ($r$ = 0.82$\pm$0.16).

Finally, building upon the application of electrical stimulation, we are interested in exploring how neurological control is impacted following a stroke, and seek to understand if force-derived measures can be clinically relevant. To this end, we develop a portable, neurostimulation integrated, force measurement platform to measure plantarflexion central drive (i.e.,\ the ratio of the forces produced without and with superimposed electrical stimulation). Although not wearable, this initial work reaffirms central drive as a clinically-relevant metric which is associated with post-stroke walking ability. Future work to explore how such metrics can be measured directly with deformation sensing is ongoing.

Together, this thesis contributes to a more thorough understanding of how muscle deformation can be used as a wearable measure of muscle force and demonstrates how electrical stimulation can advance muscle function assessment in both deformation-based and mechanical devices.