Structure of force as a predictor of oral motor learning in healthy younger and older adults

adult;complexity;dysphagia;entropy;dynamical systems;learning;lip;motor control;motor learning;oral motor;rehabilitation;tongue;speech;swallow;variability

Bronson-Lowe, Christina R

关键词: adult;    complexity;    dysphagia;    entropy;    dynamical systems;    learning;    lip;    motor control;    motor learning;    oral motor;    rehabilitation;    tongue;    speech;    swallow;    variability;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/99351/BRONSON-LOWE-DISSERTATION-2017.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This study examined the relationship of lingual and labial force structure to learning of oral motor continuous fine force tasks using a pursuit tracking paradigm. It investigated how error and the temporal and frequency structure of force during baseline performance predicted oral motor learning in healthy younger and older adults, drawing on dynamical systems (Bernstein, 1967, as cited in Newell et al., 2003), bidirectional complexity change (Vaillancourt & Newell, 2002) and optimal variability theories(Stergiou, Harbourne, & Cavanaugh, 2006) to explain interacting effects of age and task demand.Right-handed younger (18-28 years of age, N = 20) and older (71-79 years of age, N = 21) adults participated in 2 days’ practice matching constant, 0.75-Hz sinusoidal, and complex periodic (hereafter “multicosine”) visual targets by pursing the lips or elevating the tongue to exert submaximal force whose magnitude controlled the height of a visual trace. Targets were centered at 15% of maximal voluntary force (MVF) determined individually per participant and effector. Over the two days, participants practiced matching each target a total of 35 times with each effector. On the third day, learning was assessed in retention trials (unmodified tasks) and transfer trials (multiple task characteristics individually, systematically modified; only transfer to 10% and 20% MVF target force levels is reported here).Measures of force structure (approximate entropy, ApEn; fuzzy measure entropy, FuzzyMEn; proportion of power, PoP, in 0-1 Hz, 1-2 Hz and 2-3 Hz bands) and error (normalized root mean square error, NRMSE) at baseline (day 1, first trial of each effector x task condition) were related to measurements of reduction in error vs. baseline in retention and transfer trials on the third day ({delta}finalNRMSEret, {delta}finalNRMSEtrn), using primarily linear mixed effects modeling. Results are presented organized by hypotheses within specific aims. Because each hypothesis was assessed using multiple measures, only a selection of the results is covered here for brevity.Specific aim 1. Assess applicability of previous findings on effects of age and task to oral effectors.Hypothesis 1a. Older adults’ force structure will differ task-dependently from younger adults’ (lower entropy and a greater proportion of low-frequency power when the task demands high entropy and reduced low-frequency power, and vice versa).At baseline, task and age group interacted (ApEn: F(2, 205) = 9.555; FuzzyMEn: F(2, 205) = 9.515; both p < 0.0005). Follow-up analysis showed that only younger adults altered entropy across task, (ApEn: F(2, 100) = 17.173; FuzzyMEn, F(2, 100) = 20.492, both p <0.0005). Younger adults produced higher-entropy force than older adults only on the constant task (ApEn: F(1, 41) = 9.407, p = 0.004; FuzzyMEn, F(1, 41) = 10.297, p = 0.003).In retention trials, younger adults’ entropy was higher than older adults’ in the lip x constant force condition (ApEn: t(39) = -4.339, p = 0.002; FuzzyMEn: t(39) = -4.295; p = 0.001) and lower in the tongue x sine condition (ApEn: t(39) = 4.741; FuzzyMEn: t(39) = 4.191; both p = 0.001). These effects suggest younger adults adapted structure of output to task demand more closely than the older adults.Hypothesis 1b. Adaptability (immediate): Older adults will change structure of force to meet task demands less effectively than younger adults, comparing trial 2 to trial 1 on day 1 within each effector x task combination.There was no significant effect of age or age-task interaction with this minimal practice.Hypothesis 1c. Adaptability (after practice): Older adults will change structure of force to meet task demands less effectively than younger adults, comparing day 1 trial 1 to day 3 retention trial 1 within each effector x task combination.Entropy change with practice depended upon an interaction of age group and task (ApEn: F(2, 205) = 5.890, p = 0.003; FuzzyMEn: F(2, 205) = 4.950, p = 0.008). For the multicosine task, younger adults did not change entropy with practice, while older adults increased it (ApEn: t(67.503) = 2.675, p = 0.014; FuzzyMEn: F(1, 41) = 4.559, p = 0.039, NS). For the sine task, younger adults decreased entropy with practice, while older adults increased it (ApEn: t(75.08) = 4.308, p = 0.001; FuzzyMEn: F(1, 41) = 9.657, p = 0.003).Hypothesis 1d. Older adults’ reduction in error vs. baseline on retention and transfer trials after two days’ practice will be less than younger adults’.On retention trials, age group interacted significantly with both effector (F(1, 205) = 5.702, p = 0.018) and task (F(2, 205) = 3.871, p = 0.022). Younger adults showed greater reduction in NRMSE than older adults only with the tongue (F(1, 41) = 7.38, p = 0.009) and on the sine task (F(1, 41) = 11.288, p = 0.002).Hypothesis 1e. Structure of force will differ by task. The constant task will elicit the highest entropy, lowest proportion of low-frequency power, and greatest proportion of higher-frequency power. The sine task will elicit the lowest entropy, greatest proportion of low-frequency power, and lowest proportion of higher-frequency power. The multicosine task will be intermediate.At baseline, both younger and older adults responded to the increased high-frequency content of the multicosine target compared to the sine target by decreasing power in the 0-1 Hz band and increasing it in the 1-2 Hz band (all p ≤ 0.002; see Table 20). On retention trials, entropy had increased with practice to a greater degree for the constant task than for both variable tasks (ApEn, F(2, 205) = 34.918; FuzzyMEn, F(2, 205) = 32.121; main effects and pairwise comparisons of constant to sine and multicosine, all p < 0.0005).Specific aim 2. Assess differences in motor variability between oral effectors.Hypothesis 2a. The tongue will produce less complex force than the lip (lower-entropy, greater dominance of low-frequency power).Hypothesis 2b. The effects of age group and effector on entropy will interact.At baseline, effector and age group interacted (ApEn: F(1, 205) = 10.806, p < 0.0005; FuzzyMEn, F(1, 205) = 9.769, p = 0.002). For older adults only, entropy was higher for the tongue (ApEn: F(1, 105) = 23.591; FuzzyMEn, F(1, 105) = 20.794; both p < 0.0005). On retention trials, older adults still produced higher-entropy force with the tongue (ApEn: F(1, 105) = 22.708; FuzzyMEn, F(1, 105) = 22.767; both p < 0.0005). Younger adults’ force production on retention trials showed higher entropy with the lip for the constant task (which demands high-entropy force; ApEn: F(1, 20) = 11.950, p = 0.002; FuzzyMEn: F(1, 20) = 10.768) and slightly lower entropy with the lip for the more structured variable tasks (significant only for multicosine, ApEn: F(1, 20) = 10.821, p = 0.004; FuzzyMEn: F(1, 20) = 11.775, p = 0.003), suggesting that adaptation to task demand with practice may have been better with the lip.Specific aim 3. Assess utility of baseline performance measures in predicting de novo learning of fine-force pursuit tracking tasks in oral effectors.Hypothesis 3. (a) Error at baseline (NRMSEinitial) and a measure of temporal structure, (b) higher maximal force entropy (maxApEn or maxFuzzyMEn) at baseline or (c) greater adaptability of entropy at baseline, will predict reduction in error vs. baseline on retention and transfer trials ({delta}finalNRMSEret, {delta}finalNRMSEtrn) in pursuit tracking tasks after controlling for age group, effector and task.NRMSEinitial was a significant predictor of reduced error compared to baseline for both retention and transfer trials in all pairings with the various entropy-based predictors (all p < 0.0005). Parameter estimates ranged from -0.89 to -0.91, suggesting that poor initial performance functioned as a marker of greater room for improvement.Maximum entropy also significantly predicted {delta}finalNRMSEret (maxApEn model: F(1, 245.948) = 7.005, p = 0.009; maxFuzzyMEn model: F(1, 245.823) = 5.414, p = 0.021). 1-unit increases in maxApEn and maxFuzzyMEn were estimated to predict changes in {delta}finalNRMSEret of 0.32 and 0.14 respectively (worsening of performance), after controlling for age group, task, effector and NRMSEinitial.The predictive value of initial change in entropy varied by task for retention trials ({delta}initialApEn: F(2, 236.302) = 3.514, p = 0.031; {delta}initialFuzzyMEn: F(2, 236.860) = 5.229, p = 0.006). For the constant task, which demands force output of high entropy, higher entropy on trial 2 than trial 1 on day 1 predicted a decrease in {delta}finalNRMSEret (i.e. a greater reduction in error by day 3). For the sine target, which requires force output of low entropy, higher entropy on trial 2 than trial 1 on day 1 predicted an increase in {delta}finalNRMSEret (i.e. a lesser reduction in error by day 3).Initial change in entropy significantly predicted transfer of learning to tasks with a higher target force level in both models ({delta}initial ApEn model: F(1, 232.077) = 11.853; {delta}initial FuzzyMEn model: F(1, 234.461) = 10.437, both p = 0.001). 1-unit increases in {delta}initial ApEn and {delta}initial FuzzyMEn were estimated to predict changes in {delta}finalNRMSEtrn of -0.17 [-0.27, -0.07] and -0.09 [-0.15, -0.04] respectively (improved transfer), after controlling for age group, task, effector and NRMSEinitial.These and other results from this study suggest that (Aim 1) task-dependent effects on force structure and the bidirectional complexity hypothesis of healthy aging developed in non-oral systems can be applied to fine-force control in pursuit tracking tasks using the lip and tongue; (Aim 2) oral effectors’ structure of force differs, influenced by age, and (Aim 3) baseline behavioral measures can predict learning (measured as reduction in error) after two days’ practice.Initial adaptability of entropy predicted better performance on retention trials if the direction of change was in line with task demand, and worse performance if the direction of change was counter to task demand. This effect comports with the idea of variability in early learning as an exploration of task space (Dhawale, Smith, & Ölveczky, 2017; Stergiou, Harbourne, & Cavanaugh, 2006; Wu, Miyamoto, Gonzalez Castro, Ölveczky, & Smith, 2014) and therefore an active support of learning, rather than a hindrance to be suppressed. Optimal variability in a learning context suggests the ability to shift temporal/frequency structure of force output in the direction demanded by a goal or task. The reduction in adaptability of force structure seen in the older adult participants may play a role in changes in learning with aging.This prediction can be made from a small enough data set to have potential clinical applicability. Older adults remain robustly able to learn and to adjust complexity of oral force output, though with limitations most consistent with the loss of adaptability hypothesis.