Contrasting action and posture coding with hierarchical deep neural network models of proprioception

Biological motor control is versatile, efficient, and depends on proprioceptive feedback. Muscles are flexible and undergo continuous changes, requiring distributed adaptive control mechanisms that continuously account for the body's state. The canonical role of proprioception is representing the body state. We hypothesize that the proprioceptive system could also be critical for high-level tasks such as action recognition. To test this theory, we pursued a task-driven modeling approach, which allowed us to isolate the study of proprioception. We generated a large synthetic dataset of human arm trajectories tracing characters of the Latin alphabet in 3D space, together with muscle activities obtained from a musculoskeletal model and model-based muscle spindle activity. Next, we compared two classes of tasks: trajectory decoding and action recognition, which allowed us to train hierarchical models to decode either the position and velocity of the end-effector of one's posture or the character (action) identity from the spindle firing patterns. We found that artificial neural networks could robustly solve both tasks, and the networks' units show tuning properties similar to neurons in the primate somatosensory cortex and the brainstem. Remarkably, we found uniformly distributed directional selective units only with the action-recognition-trained models and not the trajectory-decoding-trained models. This suggests that proprioceptive encoding is additionally associated with higher-level functions such as action recognition and therefore provides new, experimentally testable hypotheses of how proprioception aids in adaptive motor control.

Diverse and complex muscle spindle afferent firing properties emerge from multiscale muscle mechanics

Despite decades of research, we lack a mechanistic framework capable of predicting how movement-related signals are transformed into the diversity of muscle spindle afferent firing patterns observed experimentally, particularly in naturalistic behaviors. Here, a biophysical model demonstrates that well-known firing characteristics of mammalian muscle spindle Ia afferents - including movement history dependence, and nonlinear scaling with muscle stretch velocity - emerge from first principles of muscle contractile mechanics. Further, mechanical interactions of the muscle spindle with muscle-tendon dynamics reveal how motor commands to the muscle (alpha drive) versus muscle spindle (gamma drive) can cause highly variable and complex activity during active muscle contraction and muscle stretch that defy simple explanation. Depending on the neuromechanical conditions, the muscle spindle model output appears to 'encode' aspects of muscle force, yank, length, stiffness, velocity, and/or acceleration, providing an extendable, multiscale, biophysical framework for understanding and predicting proprioceptive sensory signals in health and disease.

Muscle Spindle Density of Lateral Rotators of the Thigh in Japanese Macaques and a Gibbon

We examined the six small lateral rotators of the hip joint, which is one of the most flexible joints and allows kinematically complex motions of the hindlimb, to elucidate the functional differentiation among these muscles and to test the hypothesis that species-specific characteristics in hindlimb use during locomotion are reflected in the muscle spindle density and in other parameters of the deep small hip joint rotators. For these purposes, we estimated the number of muscle spindles of the superior gemellus muscle (SG), inferior gemellus muscle, quadratus femoris muscle, obturator internus muscle (OI), obturator externus muscle, and piriformis muscle in three Japanese macaques and a gibbon, using 30-µm-thick serial sections throughout each muscle length after azan staining. The numbers of muscle spindles per 10,000 muscle fibers were determined to compare inter-muscle variation. The spindle density was highest in the SG and lowest in the OI in the Japanese macaques, suggesting that the SG, which is attached to the tendon of the OI, functions as a kinesiological monitor of the OI. On the other hand, SG the was missing in the gibbon, and the OI in the gibbon contained more spindles than that in the Japanese macaques. This suggests that the SG and the OI fused into one muscle in the gibbon. We postulate that the relative importance of the deep small hip rotator muscles differs between the Japanese macaques and gibbon and that the gibbon's muscles are less differentiated in terms of the spindle density, probably because this brachiating species uses its hindlimbs less frequently.

Corticospinal responses during passive shortening and lengthening of tibialis anterior and soleus in older compared to younger adults

NEW FINDINGS

What is the central question of this study? Are there age-related differences in corticospinal responses whilst passively changing muscle length? What is the main finding and its importance? In contrast to young, older adults exhibited no modulation of corticospinal excitability in tibialis anterior during passive ankle movement. These data show impaired sensorimotor response in older adults during length changes of tibialis anterior, thus contributing to our understanding of age-related changes in sensorimotor control.

ABSTRACT

Corticospinal responses have been shown to increase and decrease with passive muscle shortening and lengthening, respectively, as a result of changes in muscle spindle afferent feedback. The ageing sensory system is accompanied by a number of alterations that might influence the processing and integration of sensory information. Consequently, corticospinal excitability might be modulated differently whilst changing muscle length. In 10 older adults (66 ± 4 years), corticospinal responses (MEP/M_max ) were evoked in a static position, and during passive shortening and lengthening of soleus (SOL) and tibialis anterior (TA), and these data were compared to the re-analysed data pool of 18 younger adults (25 ± 4 years) published previously. Resting motor threshold was greater in SOL compared to TA (P < 0.001), but did not differ between young and older (P = 0.405). No differences were observed in MEP/M_max between the static position, passive shortening or lengthening in SOL (young: all 0.02 ± 0.01; older: 0.05 ± 0.04, 0.03 ± 0.02 and 0.04 ± 0.01, respectively; P = 0.298), and responses were not dependent on age (P = 0.090). Conversely, corticospinal responses in TA were modulated differently between the age groups (P = 0.002), with greater MEP/M_max during passive shortening (0.22 ± 0.12) compared to passive lengthening (0.13 ± 0.10) and static position (0.10 ± 0.05) in young (P < 0.001), but unchanged in older adults (0.19 ± 0.11, 0.22 ± 0.11 and 0.18 ± 0.07, respectively; P ≥ 0.867). The present experiment shows that length-dependent changes in corticospinal excitability in TA of the young are not evident in older adults. This suggests impaired sensorimotor response during muscle length changes in older age that might only be present in ankle flexors, but not extensors.

Corticospinal excitability of tibialis anterior and soleus differs during passive ankle movement

The purpose of this study was to assess corticospinal excitability of soleus (SOL) and tibialis anterior (TA) at a segmental level during passive ankle movement. Four experimental components were performed to assess the effects of passive ankle movement and muscle length on corticospinal excitability (MEP/M_max) at different muscle lengths, subcortical excitability at the level of lumbar spinal segments (LEP/M_max), intracortical inhibition (SICI) and facilitation (ICF), and H-reflex in SOL and TA. In addition, the degree of fascicle length changes between SOL and TA was assessed in a subpopulation during passive ankle movement. Fascicles shortened and lengthened with joint movement during passive shortening and lengthening of SOL and TA to a similar degree (p < 0.001). Resting motor threshold was greater in SOL compared to TA (p ≤ 0.014). MEP/M_max was facilitated in TA during passive shortening relative to the static position (p ≤ 0.023) and passive lengthening (p ≤ 0.001), but remained similar during passive ankle movement in SOL (p ≥ 0.497), regardless of muscle length at the point of stimulus (p = 0.922). LEP/M_max (SOL: p = 0.075, TA: p = 0.071), SICI (SOL: p = 0.427, TA: p = 0.540), and ICF (SOL: p = 0.177, TA: p = 0.777) remained similar during passive ankle movement. H-reflex was not different across conditions in TA (p = 0.258), but was reduced during passive lengthening compared to shortening in SOL (p = 0.048). These results suggest a differential modulation of corticospinal excitability between plantar and dorsiflexors during passive movement. The corticospinal behaviour observed might be mediated by an increase in corticospinal drive as a result of reduced afferent input during muscle shortening and appears to be flexor-biased.

Age-related changes in leg proprioception: implications for postural control

In addition to being a prerequisite for many activities of daily living, the ability to maintain steady upright standing is a relevant model to study sensorimotor integrative function. Upright standing requires managing multimodal sensory inputs to produce finely tuned motor output that can be adjusted to accommodate changes in standing conditions and environment. The sensory information used for postural control mainly arises from the vestibular system of the inner ear, vision, and proprioception. Proprioception (sense of body position and movement) encompasses signals from mechanoreceptors (proprioceptors) located in muscles, tendons, and joint capsules. There is general agreement that proprioception signals from leg muscles provide the primary source of information for postural control. This is because of their exquisite sensitivity to detect body sway during unperturbed upright standing that mainly results from variations in leg muscle length induced by rotations around the ankle joint. However, aging is associated with alterations of muscle spindles and their neural pathways, which induce a decrease in the sensitivity, acuity, and integration of the proprioceptive signal. These alterations promote changes in postural control that reduce its efficiency and thereby may have deleterious consequences for the functional independence of an individual. This narrative review provides an overview of how aging alters the proprioceptive signal from the legs and presents compelling evidence that these changes modify the neural control of upright standing.

Muscle thixotropy-where are we now?

Relaxed skeletal muscle has an inbuilt resistance to movement. In particular, the resistance manifests itself as a substantial stiffness for small movements. The stiffness is impermanent, because it forms only when the muscle is stationary for some time and is reduced upon active or passive movement. Because the resistance to movement increases with time at rest and is reduced by movement, this behavior has become known as muscle thixotropy. In this short review, we describe the phenomenon of thixotropy and illustrate its significance in postural control with particular emphasis on human standing. We show how thixotropy came to be unambiguously associated with muscle mechanics and we review present knowledge of the molecular basis of thixotropic behavior. Specifically, we examine how recent knowledge about titin, and about the control of cross-bridge cycling, has impacted on the role of non-cross-bridge mechanisms and cross-bridge mechanisms in explaining thixotropy. We describe how thixotropic changes in muscle stiffness that occur during transitions from posture to movement can be tracked by analyzing physiological tremor. Finally, because skeletal muscle contains sensory receptors, and because some of these receptors are themselves thixotropic, we outline some of the consequences of muscle thixotropy for proprioception.

Diet induced obesity alters muscle spindle afferent function in adult mice

Populations with obesity are more likely to fall and exhibit balance instability. The reason for this is likely multifactorial, but there is some evidence that sensory function is impaired during obesity. We tested the hypothesis that muscle proprioceptor function is compromised in a mouse model of diet induced obesity. An in vitro muscle-nerve preparation was used to record muscle spindle afferent responses to physiological stretch and sinusoidal vibration. We compared the responses of C57/Bl6 male and female mice on a control diet (10% kcal fat) with those eating a high fat diet (HFD; 60% kcal fat) for 10 weeks (final age 14–15 weeks old). Following HFD feeding, adult mice of both sexes exhibited decreased muscle spindle afferent responses to muscle movement. Muscle spindle afferent firing rates during the plateau phase of stretch were significantly lower in both male and female HFD animals as were two measures of dynamic sensitivity (dynamic peak and dynamic index). Muscle spindle afferents in male mice on a HFD were also significantly less likely to entrain to vibration. Due to the importance of muscle spindle afferents to proprioception and motor control, decreased muscle spindle afferent responsiveness may contribute to balance instability during obesity.

The influence of mechanical vibration on local and central balance control

Fall prevention has an indispensable role in enhancing life expectancy and quality of life among older adults. The first step to prevent falls is to devise reliable methods to identify individuals at high fall risk. The purpose of the current study was to assess alterations in local postural muscle and central sensory balance control mechanisms due to low-frequency externally applied vibration among elders at high fall risk, in comparison with healthy controls, as a potential tool for assessing fall risk. Three groups of participants were recruited: healthy young (n = 10; age = 23 ± 2 years), healthy elders (n = 10; age = 73 ± 3 years), and elders at high fall risk (n = 10; age = 84 ± 9 years). Eyes-open and eyes-closed upright standing balance performance was measured with no vibration, 30 Hz, and 40 Hz vibration of Gastrocnemius muscles. When vibratory stimulation was applied, changes in local-control performance manifested significant differences among the groups (p < 0.01). On average between conditions, we observed 97% and 92% less change among high fall risk participants when compared to healthy young and older adults, respectively. On the other hand, vibration-induced changes in the central-control performance were not significant between groups (p ≥ 0.19). Results suggest that local-control deficits are responsible for balance behavior alterations among elders at high fall risk and healthy individuals. This observation may be attributable to deterioration of short-latency reflexive loop in elders at high fall risk. On the other hand, we could not ascribe the balance alterations to problems related to central nervous system performance or long-latency responses.

Force encoding in muscle spindles during stretch of passive muscle

Muscle spindle proprioceptive receptors play a primary role in encoding the effects of external mechanical perturbations to the body. During externally-imposed stretches of passive, i.e. electrically-quiescent, muscles, the instantaneous firing rates (IFRs) of muscle spindles are associated with characteristics of stretch such as length and velocity. However, even in passive muscle, there are history-dependent transients of muscle spindle firing that are not uniquely related to muscle length and velocity, nor reproduced by current muscle spindle models. These include acceleration-dependent initial bursts, increased dynamic response to stretch velocity if a muscle has been isometric, and rate relaxation, i.e., a decrease in tonic IFR when a muscle is held at a constant length after being stretched. We collected muscle spindle spike trains across a variety of muscle stretch kinematic conditions, including systematic changes in peak length, velocity, and acceleration. We demonstrate that muscle spindle primary afferents in passive muscle fire in direct relationship to muscle force-related variables, rather than length-related variables. Linear combinations of whole muscle-tendon force and the first time derivative of force (dF/dt) predict the entire time course of transient IFRs in muscle spindle Ia afferents during stretch (i.e., lengthening) of passive muscle, including the initial burst, the dynamic response to lengthening, and rate relaxation following lengthening. Similar to acceleration scaling found previously in postural responses to perturbations, initial burst amplitude scaled equally well to initial stretch acceleration or dF/dt, though later transients were only described by dF/dt. The transient increase in dF/dt at the onset of lengthening reflects muscle short-range stiffness due to cross-bridge dynamics. Our work demonstrates a critical role of muscle cross-bridge dynamics in history-dependent muscle spindle IFRs in passive muscle lengthening conditions relevant to the detection and sensorimotor response to mechanical perturbations to the body, and to previously-described history-dependence in perception of limb position.

Acute effects of muscle stretching on physical performance, range of motion, and injury incidence in healthy active individuals: a systematic review

Recently, there has been a shift from static stretching (SS) or proprioceptive neuromuscular facilitation (PNF) stretching within a warm-up to a greater emphasis on dynamic stretching (DS). The objective of this review was to compare the effects of SS, DS, and PNF on performance, range of motion (ROM), and injury prevention. The data indicated that SS- (-3.7%), DS- (+1.3%), and PNF- (-4.4%) induced performance changes were small to moderate with testing performed immediately after stretching, possibly because of reduced muscle activation after SS and PNF. A dose-response relationship illustrated greater performance deficits with ≥60 s (-4.6%) than with <60 s (-1.1%) SS per muscle group. Conversely, SS demonstrated a moderate (2.2%) performance benefit at longer muscle lengths. Testing was performed on average 3-5 min after stretching, and most studies did not include poststretching dynamic activities; when these activities were included, no clear performance effect was observed. DS produced small-to-moderate performance improvements when completed within minutes of physical activity. SS and PNF stretching had no clear effect on all-cause or overuse injuries; no data are available for DS. All forms of training induced ROM improvements, typically lasting <30 min. Changes may result from acute reductions in muscle and tendon stiffness or from neural adaptations causing an improved stretch tolerance. Considering the small-to-moderate changes immediately after stretching and the study limitations, stretching within a warm-up that includes additional poststretching dynamic activity is recommended for reducing muscle injuries and increasing joint ROM with inconsequential effects on subsequent athletic performance.

Parallel neuronal mechanisms underlying physiological force tremor in steady muscle contractions of humans

We present results from a study of the 6-to 12-Hz force tremor in relation to motor unit (MU) firing synchrony. Our experimental observations from 32 subjects, 321 contractions, and 427 recorded MUs reveal that tremor is accompanied by corresponding, in-phase MU rhythms that are additional to the ones at the MU intrinsic firing rates. This rhythmical synchrony is widespread and has a uniform strength that ranges from near zero to very large (MU/MU coherence > 0.50) in different contractions. Both the synchrony and the tremor are suppressed during ischemia, and this strongly suggests an involvement of spindle feedback in their generation. Furthermore, in the presence of substantial synchrony, the tremor enhancement, relative to the minimal tremor of ischemia, reflects the strength of the synchrony. Theoretical considerations based on these observations indicate that the muscle force signal is expected to show 1) frequency components in the band of the firing rates of the last-recruited, large MUs, and 2) because of the synchronized MU rhythms, an additional, distinct component with a size reflecting the strength of synchrony. Furthermore, synchronized MU rhythms, with frequencies in the 6- to 12-Hz range, are expected to arise from self-oscillations in the monosynaptic stretch reflex loop, due primarily to the associated muscle delay (several tens of milliseconds). Our results therefore reveal the parallel action of two tremor mechanisms, one of which involves MU synchrony probably caused by loop action. Clearly, the results on the synchrony and its impact also apply to other possible generators of tremor synchrony, including supraspinal ones.

Posture-related oscillations in human cerebellar thalamus in essential tremor are enabled by voluntary motor circuits

The mechanism of essential tremor (ET) is unclear. Animal models of tremor and functional imaging studies in ET predict that the cerebellum and a cerebellar recipient thalamic nucleus (ventral intermediate, Vim) should exhibit oscillatory activity during rest and during tremor due to abnormal olivo-cerebellar activity. Physiologic responses of 152 single neurons were recorded during awake mapping of the ventral thalamus in seven patients with ET prior to thalamotomy. During postural tremor, spectral cross-correlation analysis demonstrated that 51% of the neurons studied exhibited a concentration of power at tremor frequency that was correlated with electromyography, i.e., tremor neurons. During rest, thalamic neurons did not exhibit tremor-frequency activity. Among the three thalamic nuclei surveyed, Vim had a significantly higher proportion of tremor neurons than did the principal somatic sensory nucleus (ventral caudal, Vc) or a pallidal recipient thalamic nucleus (ventral oral posterior, Vop). Neurons related to active movement (voluntary neurons) had significantly greater tremor-related activity than did nonvoluntary neurons. These findings are not consistent with a model of continuous olivo-cerebellar driving of the motor cortex through thalamic connections. Instead ET may be facilitated by motor circuits that enable tremor-related thalamic activity during voluntary movement. Additionally, a subgroup of tremor neurons with proprioceptive inputs were identified that may allow sensory feedback to access the central tremor network.

Improving signal reliability for on-line joint angle estimation from nerve cuff recordings of muscle afferents

Closed-loop functional electrical stimulation (FES) applications depend on sensory feedback, thus, it is important to continuously investigate new methods to obtain reliable feedback signals. The objective of the present paper was to examine the feasibility of using an artificial neural network (ANN) to predict joint angle from whole nerve cuff recordings of muscle afferent activity within a physiological range of motion. Furthermore, we estimated how small changes in joint angle that can be detected from the nerve cuff recordings. Neural networks were tested with data obtained from ten acute rabbit experiments in simulated, on-line experiments. The electroneurograms (ENG) of the tibial and peroneal nerves were recorded during passive ankle joint rotation. To decrease the joint angle prediction error with new rabbit data, we attempted to pretune the nerve signals and re-trained the ANNs with the pretuned data. With these procedures we were able to compensate for interrabbit variability. On average the mean prediction errors were less than 2.0 degrees (a total excursion of 20 degrees) and we were able to predict joint angles from muscle afferent activity with accuracy close to the best-estimated angular resolution. The angular resolution was found to depend on the initial joint angle and the actual step size taken and we found that there was a low probability of detecting joint angle changes less than 1.5 degrees. We thus suggest that muscle afferent activity is applicable as feedback in real-time closed-loop control, when the motion speed is restricted and when the movement is limited to a portion of the joint's physiological range.

Nerve cuff recordings of muscle afferent activity from tibial and peroneal nerves in rabbit during passive ankle motion

Activity from muscle afferents regarding ankle kinesthesia was recorded using cuff electrodes in a rabbit preparation in which tactile input from the foot was eliminated. The purpose was to determine if such activity can provide information useful in controlling functional electrical stimulation (FES) systems that restore mobility in spinal injured man. The rabbit's ankle was passively flexed and extended while the activity of the tibial and peroneal nerves was recorded. Responses to trapezoidal stimulus profiles were investigated for excursions from 10 degrees to 60 degrees using velocities from 5 degrees/s to 30 degrees/s and different initial ankle positions. The recorded signals mainly reflect activity from primary and secondary muscle afferents. Dorsiflexion stretched the ankle extensors and produced velocity dependent activity in the tibial nerve, and this diminished to a tonic level during the stimulus plateau. The peroneal nerve was silent during dorsiflexion, but was activated by stretch of the peroneal muscles during ankle extension. The responses of the two nerves behaved in a reciprocal manner, but exhibited considerable hysteresis, since motion that relaxed the stretch to the driving muscle produced an immediate cessation of the prior stretch induced activity. A noted difference between the tibial and peroneal nerve responses is that the range of joint position change that activated the flexor afferents was greater then for the extensor afferents. Ankle rotation at higher velocities increased the dynamic stretch evoked responses during the stimulus ramp but showed no effect on the tonic activity during the stimulus plateau. Prestretching the muscles by altering the initial position increased the response to the ramp movement, however, for the peroneal nerve, when the prestretch brought the flexor muscles near to their maximal lengths, the response to additional stretch provided by the ramp movement was diminished. The results indicate that the whole nerve recorded muscle afferent activity may be useful for control of FES assisted standing, because it can indicate the direction of rotation of the passively moved ankle joint, along with coarse information regarding the rate of movement and static joint position.

Simultaneous and nonlinear identification of mechanical and reflex properties of human elbow joint muscles

The naturally coexisting intrinsic mechanical and reflex properties of the human elbow joint were identified simultaneously using nonlinear, time-delay, continuous-time, and dynamic models. Angular random perturbations of small amplitude and low bandwidth were applied to the joint using a computer-controlled servomotor, while the subject maintained various levels of mean background muscle torque. Joint neuromuscular dynamics were identified from the measured elbow angle and torque. Stretch reflexes were modeled nonlinearly with both dynamic and static reflex gains. A continuous-time system identification method was developed to estimate parameters of the nonlinear models directly from sampled data while retaining realistic physical or physiological interpretations. Results from six subjects showed that dynamic stretch reflex gains, joint stiffness, and viscosity generally increased with mean background muscle torque; and that dynamic stretch reflex gain was higher during muscle stretch than that during muscle shortening. More importantly, the study provided realistic simultaneous estimates of the relative contributions of intrinsic mechanical and reflex actions to net joint torque. In particular, reflexively-mediated stiffness generated a significant portion of the total joint stiffness and the percentage varied systematically with background muscle torque.

Closed-loop control of ankle position using muscle afferent feedback with functional neuromuscular stimulation

This paper describes a closed-loop functional neuromuscular stimulation system that uses afferent neural activity from muscle spindle fibers as feedback for controlling position of the ankle joint. Ankle extension against a load was effected by neural stimulation through a dual channel intrafascicular electrode of a fascicle of the tibial nerve that innervated the gastrocnemius muscle. Ankle joint angle was estimated from recordings of tibialis anterior and lateral gastrocnemius spindle fiber activity made with dual channel intrafascicular electrodes. Experiments were conducted in neurally intact anesthetized cats and in unanesthetized decerebrate cats to demonstrate the feasibility of this system. The system was able to reach and maintain a fixed target ankle position in the presence of a varying external moment ranging in magnitude between 7.3 and 22 N-cm opposing the action of the ankle extensor, as well as track a sinusoidal target ankle position up to a frequency of 1 Hz in the presence of a constant magnitude 22- or 37-N-cm external moment.

Propriomuscular coding of kinaesthetic sensation. Experimental approach and mathematical modelling

The role of propriomuscular information in kinaesthetic sensation was studied. Experiments were carried out on human subjects in whom kinaesthetic illusions were induced by applying tendon vibration with a variable frequency. Six patterns of frequency modulation were used, four of which had an arbitrary form and the other two mimicked natural Ia discharges. The results show that the shape of the illusory movements recorded depended on the type of vibratory pattern used. A mathematical model for the propriomuscular information decoding process is proposed. It takes into account both the agonist and antagonist muscle spindle populations as sources of kinaesthetic information and is based on the assumption that position and velocity information are additively combined. The experimental data show a good fit with the theoretical data obtained by means of model simulation, thus validating our initial hypothesis. Various aspects of the experimental results and the hypotheses involved in the model are discussed.