Abstract
Part IStudy 1: Effect of Sample Frequency and Filter Frequency on the Approximate Entropy Values for Isometric Force Records
ApEn has been used to quantify signal complexity in isometric contractions and distinguish between patient groups. Sampling frequencies, ‘r’ values (a parameter needed in the ApEn algorithm which essentially accounts for the noise in the signal (Pincus 1991) and filters may affect signal characteristics reflected in an alteration in ApEn values and subsequent patterns of results. However, there is little standardisation of such procedures for this measure. While the ‘true’ ApEn value cannot be known the approach taken here was to start with the highest resolution signal and to identify the pattern of ApEn results across different percentages of maximum voluntary contraction and then assess the effect of the post-processing changes on this pattern. Isometric contractions of the first dorsal interosseus for 10 seconds at 5, 10, 25, 50, and 75% of maximum voluntary contraction (MVC) were collected at 1200 Hz. Signals were post processed to determine the effect of different filter cutoff frequencies, ‘r’ values, sample rates (by resampling) and lengths of data series. Resampling of the data changed the pattern of ApEn results across the percentages of MVC range dramatically, whereas shortening the length of the time series had no effect on this pattern. This suggests that it is the sample rate but not the number of samples that affects the pattern of ApEn results across the percentages of MVC range. ‘r’ values representing the noise in the signal that were scaled using the SD of each force record flattened the pattern of ApEn values across the percentage of MVC range, whereas ‘r’ values derived from the transducer noise led to a generally monotonic trend across the %MVC range. The filter cutoff frequency did not alter the pattern of ApEn values across effort levels, though frequencies in the signal above 20 Hz were found during spectral analysis that may well be biological in origin. The results show the choice of sample rate and ‘r’ is critical in reliably determining changes in ApEn with effort level. It is suggested that ‘r’ used should be, if possible, a measured estimate of noise, sample rates should be at least 600 Hz and filter cut-offs should not be lower than 40 Hz.
Study 2: Effect of Bimodal Stimulus on Force Control of Elderly and Young Adults
Research shows that elderly adults often exhibit reduced irregularity in force signal data during isometric contractions (e.g. Sosnoff and Newell (2006c). Previous findings revealed that the post-processing of data effects Approximate Entropy (ApEn) values, used to measure the regularity of the force signal. The purpose of the present study was to examine magnitude of variability (SD, CV, RMSE) and structure of isometric force data using appropriate post-processing methods previously identified. It was hypothesised that in-line with much of the literature there would be differences between young and elderly adults’ force output. Differences in structure of force output between elderly and young adults may be as a result of elderly adults having reduced visuomotor processing capabilities, to assess this, target feedback was provided in the form of three different types of stimuli: 1) audio 2) both audio and visual (bimodal) 3) visual. The audio stimulus changed pitch depending on whether the force trajectory was too high or low and was silenced when force level was correct. This was included with the visual target (on-screen target trajectory) during the bimodal trial. Audio stimulus
was used in order to gain more information as to whether differences between elderly and young adults are due to visuomotor processing differences. Young (18-25) and elderly (65-72) neurologically healthy adults produced isometric force contractions using abduction of the FDI at six force levels (5, 10, 25, 40, 50 and 75% MVC) during each of the three conditions. There were no differences found in magnitudes of variability between the age groups nor was there any alteration in force output in the bimodal condition compared to the visual condition for either groups. The audio condition altered all indexes of force structure and variability significantly. ApEn values were significantly higher in younger adults at force levels >25% MVC during all conditions indicating higher irregularity than elderly adults (mean across all conditions and force: elderly = 0.15, SD 0.11, young = 0.19, SD 0.13). In contrast with younger subjects, elderly adults exhibited a higher percentage of relative power in the 0-0.5 Hz frequency bands and exhibited limited alteration in the percentage power in the 0-4 Hz bands with change of force requirement or condition. This suggests differences in processing and behaviour between the two groups, but as there are still differences in audio condition it suggests that the differences in processing are not related to visual processing alone. This research has introduced a novel audio technique in order to compare groups without visual processing contribution. These results support the postulation that reduced complexity occurs with ageing. As the force signal is more pattern like it may result in a reduced ability to alter force production when
required leading to a lower level of functionality.
Study 3:Decreased ApEn values in older adults are associated with increased time to achieve steady muscle force following a change in required force.
Previous findings indicate that elderly adults exhibit reduced irregularity in the force signal compared to young adults when completing an isometric force matching task. Functional significance of lower ApEn values is investigated as reduced irregularity is considered to result in less adaptability (Lipsitz and Goldberger 1992). It was hypothesised that lower ApEn values would be associated with a reduced ability to adapt to a required force change. ApEn of the force signal, functional reaction time, and time to reach a steady state at the new force target was measured. Two different types of stimulus were presented to participants as force targets, bimodal and visual. The audio stimulus in the bimodal task changed pitch depending on whether the force trajectory was too high or low and was silenced when force level was correct. Audio stimulus was used in order to gain more information as to whether differences between elderly and young adults are due to visuomotor processing differences. Young (18-25) and elderly (65-72) neurologically healthy adults produced isometric force contractions using abduction of the FDI at six force levels that either increased, or decreased at a
random interval. Increasing force levels were 5 to 25%, 25 to 50% and 25 to 75% of MVC and decreasing force levels were 25 to 5%, 50 to 25% and 75 to 25% of MVC. Each force level was attempted in random order under both visual and bimodal conditions. Results were compared with data collected from a continuous force task under the same conditions. There were no differences found in magnitudes of variability between the age groups nor was there any significant difference in findings during the bimodal condition compared to the visual condition for either groups. Confirming previous findings, ApEn values were higher in younger adults at force levels >25% MVC during both conditions indicating higher irregularity. However elderly adults displayed increased ApEn values compared to the continuous force trial which suggests that they are able to increase irregularity in force output during certain tasks. Confirming our hypothesis, elderly adults exhibited longer times to reach steady state, even after removal of reaction time at force levels initiated above 5% MVC. Mean time to reach steady state (minus reaction time) was 2.83 s for young subjects and 3.23 s for elderly adults [t=2.14, p=0.03]. These results did not depend on whether the force target moved up or down. These results provide evidence to support the concept that reduced irregularity leads to decreased adaptability to task alterations. This knowledge may be beneficial when modelling ageing movement and force production or used as a pre-clinical tool for identifying those at risk of falls etc.
Part II
Study 4: The determination of Body Segment Inertial Parameters of young female club level athletes
Body segment inertial parameters (BSIPs) must be determined prior to performing any biomechanical analyses. Geometric BSIP models are cost effective, yet collecting the anthropometric data necessary is time consuming and time with athletes is often limited. Also, few anthropometric models have been validated for female athletes. Previous work suggested that modelling limb segments as two instead of four truncated cones per segment produces a negligible difference in predicted segment mass (Forrest 2008) yet whole body volume was overestimated due to inadequate modelling of the trunk segment. The present study aimed to confirm earlier findings using a refined trunk segment model. Thirty females provided written informed consent. A total of 118 anthropometric measurements were taken from each participant. The upper arms, forearms, hands, thighs, shank and feet were each modelled using four shapes per segment in the full model, and two shapes per segment in the reduced model. The trunk segment was modelled as a series of ten stadium solids in both models. Further refinements of the present model addressed the shoulder area reducing overlap of trunk and upper arm segments. The geometric model predicted segment volume and which was multiplied by cadaver derived density functions (Clauser et al. 1969) to determine segment mass. The root mean square error between actual Whole Body Volumes (WBV), determined using a hydrostatic weighing tank, and predicted WBV was 2.37%, 3.03% and 2.34% of WBV for the full, reduced and basic models respectively. Although the basic model produced the lowest WBV and whole body mass errors, the model had lower correlation than the full model with DXA derived segment masses and as a result is likely to be not such a good BSIP predictor. The model predicted trunk mass with MSE of just 3.49% of segment mass compared to DXA measured trunk mass. Pearsons correlation showed high correlation between the segment masses predicted by the full model and DXA measured mass [r values ranged from 0.727-0.893, p<0.001] for the upper arms, forearms, thighs, shanks and feet. The full and reduced model showed high correlation for all segments [mean r=0.9100, p<0.001] which confirms that reducing the number of anthropometric measurements taken from the limb segments
(reducing required measures from 118 measures to 94) causes little difference in the predicted mass for limb segments. These results are of interest to sports biomechanists who are without access to direct imaging techniques, but who wish to compute subject specific BSIPs.
| Date of Award | 23 Apr 2012 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Mark Burnley (Supervisor) & Samantha Lee Winter (Supervisor) |