Machine learning algorithms can classify outdoor terrain types during running using accelerometry data.

BACKGROUND: Running is a popular physical activity that benefits health; however, running surface characteristics may influence loading impact and injury risk. Machine learning algorithms could automatically identify running surface from wearable motion sensors to quantify running exposures, and perhaps loading and injury risk for a runner. RESEARCH QUESTION: (1) How accurately can machine learning algorithms identify surface type from three-dimensional accelerometer sensors? (2) Does the sensor count (single or two-sensor setup) affect model accuracy? METHODS: Twenty-nine healthy adults (23.3 ±â€¯3.6 years, 1.8 ±â€¯0.1 m, and 63.6 ±â€¯8.5 kg) participated in this study. Participants ran on three different surfaces (concrete, synthetic, woodchip) while fit with two three-dimensional accelerometers (lower-back and right tibia). Summary features (n = 208) were extracted from the accelerometer signals. Feature-based Gradient Boosting (GB) and signal-based deep learning Convolutional Neural Network (CNN) models were developed. Models were trained on 90% of the data and tested on the remaining 10%. The process was repeated five times, with data randomly shuffled between train-test splits, to quantify model performance variability. RESULTS: All models and configurations achieved greater than 90% average accuracy. The highest performing models were the two-sensor GB and tibia-sensor CNN (average accuracy of 97.0 ±â€¯0.7 and 96.1 ±â€¯2.6%, respectively). SIGNIFICANCE: Machine learning algorithms trained on running data from a single- or dual-sensor accelerometer setup can accurately distinguish between surfaces types. Automatic identification of surfaces encountered during running activities could help runners and coaches better monitor training load, improve performance, and reduce injury rates.

Pulseless electrical activity following traumatic cardiac arrest: Sign of life or death?

BACKGROUND: Generally considered a sign of life, PEA is the most common arrhythmia encountered following pre-hospital traumatic cardiac arrest. Some recommend cardiac ultrasound (CUS) to determine cardiac wall motion (CWM) prior to terminating resuscitation efforts. This purpose of this study was to evaluate the outcomes of patients with traumatic cardiac arrest presenting with PEA, with and without CWM. METHODS: Trauma patients who underwent pre-hospital CPR were identified from the registries of two level-1 trauma centers. Pre-hospital management by emergency medical transport services was guided by advanced life support protocols. The on-duty trauma surgeon directed the resuscitations and performed or supervised CUS and determined CWM. RESULTS: Among 277 patients who underwent pre-hospital CPR, 110 patients had PEA on arrival to ED. 69 (62.7%) were injured by blunt mechanisms. Median CPR duration was 20.0 and 8.0 min for pre-hospital and ED, respectively. Sixty-three patients (22.7%) underwent resuscitative thoracotomy. One hundred seventy-two patients (62.1%) received CUS and of these 32 (18.6%) had CWM. CWM was significantly associated with survival to hospital admission (21.9% vs. 1.4%; P < 0.001); however, no patient with CUS survived to hospital discharge. Overall, only one patient with PEA on arrival survived to discharge. CONCLUSION: Following pre-hospital traumatic cardiac arrest, PEA on arrival portends death. Although CWM is associated with survival to admission, it is not associated with meaningful survival. Heroic resuscitative measures may be unwarranted for PEA following pre-hospital traumatic arrest, regardless of CWM.

Effect of walking surface, late-cueing, physiological characteristics of aging, and gait parameters on turn style preference in healthy, older adults.

Turning while walking is a crucial component of locomotion, often performed on irregular surfaces with little planning time. Turns can be difficult for some older adults due to physiological age-related changes. Two different turning strategies have been identified in the literature. During step turns, which are biomechanically stable, the body rotates about the outside limb, while for spin turns, generally performed with closer foot-to-foot distance, the inside limb is the main pivot point. Turning strategy preferences of older adults under challenging conditions remains unclear. The aim of this study was to determine how turning strategy preference in healthy older adults is modulated by surface features, cueing time, physiological characteristics of aging, and gait parameters. Seventeen healthy older adults (71.5 ±â€¯4.2 years) performed 90° turns for two surfaces (flat, uneven) and two cue conditions (pre-planned, late-cue). Gait parameters were identified from kinematic data. Measures of lower-limb strength, balance, and reaction-time were also recorded. Generalized linear (logistic) regression mixed-effects models examined the effect of (1) surface and cuing, (2) physiological characteristics of ageing, and (3) gait parameters on turn strategy preference. Step turns were preferred when the condition was pre-planned (p < 0.001) (model 1) and when the gait parameters of stride regularity and maximum acceleration decreased (p = 0.010 and p = 0.039, respectively) (model 3). Differences in turn strategy selection under dynamic conditions ought to be evaluated in future fall-risk research and rehabilitation utilizing real-world activity monitoring.

Machine learning algorithms based on signals from a single wearable inertial sensor can detect surface- and age-related differences in walking.

The aim of this study was to investigate if a machine learning algorithm utilizing triaxial accelerometer, gyroscope, and magnetometer data from an inertial motion unit (IMU) could detect surface- and age-related differences in walking. Seventeen older (71.5 ±â€¯4.2 years) and eighteen young (27.0 ±â€¯4.7 years) healthy adults walked over flat and uneven brick surfaces wearing an inertial measurement unit (IMU) over the L5 vertebra. IMU data were binned into smaller data segments using 4-s sliding windows with 1-s step lengths. Ninety percent of the data were used as training inputs and the remaining ten percent were saved for testing. A deep learning network with long short-term memory units was used for training (fully supervised), prediction, and implementation. Four models were trained using the following inputs: all nine channels from every sensor in the IMU (fully trained model), accelerometer signals alone, gyroscope signals alone, and magnetometer signals alone. The fully trained models for surface and age outperformed all other models (area under the receiver operator curve, AUC = 0.97 and 0.96, respectively; p ≤ .045). The fully trained models for surface and age had high accuracy (96.3, 94.7%), precision (96.4, 95.2%), recall (96.3, 94.7%), and f1-score (96.3, 94.6%). These results demonstrate that processing the signals of a single IMU device with machine-learning algorithms enables the detection of surface conditions and age-group status from an individual's walking behavior which, with further learning, may be utilized to facilitate identifying and intervening on fall risk.

Gait adaptations of older adults on an uneven brick surface can be predicted by age-related physiological changes in strength.

BACKGROUND: Outdoor falls in community-dwelling older adults are often triggered by uneven pedestrian walkways. It remains unclear how older adults adapt to uneven surfaces typically encountered in the outdoor built-environment and whether these adaptations are associated to age-related physiological changes. RESEARCH QUESTION: The aims of this study were to (1) compare gait parameters over uneven and flat brick walkways, (2) evaluate the differences between older and young adults for these two surfaces, and (3) assess if physiological characteristics could predict adaptations in older adults. METHODS: Balance, strength, reaction-time, full-body marker positions, and acceleration signals from a trunk-mounted inertial measurement unit were collected in seventeen older (71.5 ±â€¯4.2 years) and eighteen young (27.0 ±â€¯4.7 years) healthy adults to compute lower-limb joint kinematics, spatio-temporal parameters, dynamic stability, and accelerometry-derived metrics (symmetry, consistency, and smoothness). RESULTS: Both groups increased hip flexion at foot-strike, while decreasing ankle dorsiflexion, margin of stability, symmetry, and consistency on the uneven, compared to flat, surface. Older, compared to young, adults showed a larger increase in knee flexion at foot-strike and a larger decrease in smoothness on the uneven surface. Only young adults decreased hip abduction on the uneven surface. Strength, not balance nor reaction-time, was the main predictor of hip abduction in older adults on both surfaces. SIGNIFICANCE: While older adults may be especially vulnerable, uneven surfaces negatively impact gait, irrespective of age, and could represent a risk to all pedestrians.

Evidence for beta corticomuscular coherence during human standing balance: Effects of stance width, vision, and support surface.

The role of the cerebral cortex in maintaining human standing balance remains unclear. Beta corticomuscular coherence (CMC) provides a measure of communication between the sensory-motor cortex and muscle, but past literature has not demonstrated significant beta CMC during human stance. This study evaluated the effects of stance width, vision, and surface compliance on beta CMC during human stance using methods to enhance sensitivity to CMC. Ten healthy, young adults stood for three 60-s trials in each of a wide or narrow stance width while on a firm surface and in narrow stance on a foam surface, each with eyes open or closed. Beta CMC was calculated between contralateral electroencephalographic and electromyographic recordings. Electromyography was recorded from bilateral tibialis anterior and gastrocnemius lateralis muscles. CMC magnitude was defined as the average integrated area of coherence spectrum above a significance threshold. Measures of center-of-pressure (COP) sway were derived from force plates under the subjects' feet. Results of CMC from four muscles across six stance conditions (a total of 24 combinations) demonstrated significant average CMC magnitude from every subject in 20 combinations and significant average CMC magnitude in nine of 10 subjects in the remaining four combinations. The CMC magnitude was significantly larger in the wide-stance condition than in the narrow-stance condition with eyes open. No significant differences were detected when comparing eyes-open to eyes-closed conditions or when comparing firm- to foam-surface conditions. Correlations between CMC magnitude and COP sway elicited some significant relationships, but there was no consistent direction or pattern of correlation based on muscle or stance condition. Results demonstrate that significant beta CMC is evident during human standing balance, and that beta CMC is responsive to changes in mechanical, but not visual or surface, conditions.

The supplementary motor area contributes to the timing of the anticipatory postural adjustment during step initiation in participants with and without Parkinson's disease.

The supplementary motor area (SMA) is thought to contribute to the generation of anticipatory postural adjustments (APAs, which act to stabilize supporting body segments prior to movement), but its precise role remains unclear. In addition, participants with Parkinson's disease (PD) exhibit impaired function of the SMA as well as decreased amplitudes and altered timing of the APA during step initiation, but the contribution of the SMA to these impairments also remains unclear. To determine how the SMA contributes to generating the APA and to the impaired APAs of participants with PD, we examined the voluntary steps of eight participants with PD and eight participants without PD, before and after disrupting the SMA and dorsolateral premotor cortex (dlPMC), in separate sessions, with 1-Hz repetitive transcranial magnetic stimulation (rTMS). Both groups exhibited decreased durations of their APAs after rTMS over the SMA but not over the dlPMC. Peak amplitudes of the APAs were unaffected by rTMS to either site. The symptom severity of the participants with PD positively correlated with the extent that rTMS over the SMA affected the durations of their APAs. The results suggest that the SMA contributes to the timing of the APA and that participants with PD exhibit impaired timing of their APAs, in part, due to progressive dysfunction of circuits associated with the SMA.

Cortical control of postural responses.

This article reviews the evidence for cortical involvement in shaping postural responses evoked by external postural perturbations. Although responses to postural perturbations occur more quickly than the fastest voluntary movements, they have longer latencies than spinal stretch reflexes, suggesting greater potential for modification by the cortex. Postural responses include short, medium and long latency components of muscle activation with increasing involvement of the cerebral cortex as latencies increase. Evidence suggests that the cortex is also involved in changing postural responses with alterations in cognitive state, initial sensory-motor conditions, prior experience, and prior warning of a perturbation, all representing changes in "central set." Studies suggest that the cerebellar-cortical loop is responsible for adapting postural responses based on prior experience and the basal ganglia-cortical loop is responsible for pre-selecting and optimizing postural responses based on current context. Thus, the cerebral cortex likely influences longer latency postural responses both directly via corticospinal loops and shorter latency postural responses indirectly via communication with the brainstem centers that harbor the synergies for postural responses, thereby providing both speed and flexibility for pre-selecting and modifying environmentally appropriate responses to a loss of balance.

Abnormal proprioceptive-motor integration contributes to hypometric postural responses of subjects with Parkinson's disease.

Subjects with Parkinson's disease exhibit abnormally short compensatory steps in response to external postural perturbations. We examined whether: (1) Parkinson's disease subjects exhibit short compensatory steps due to abnormal central proprioceptive-motor integration, (2) this proprioceptive-motor deficit can be overcome by visual-motor neural circuits using visual targets, (3) the proprioceptive-motor deficit relates to the severity of Parkinson's disease, and (4) the dysfunction of central dopaminergic circuits contributes to the Parkinson's disease subjects' proprioceptive-motor deficit. Ten Parkinson's disease subjects and 10 matched control subjects performed compensatory steps in response to backward surface translations in five conditions: with eyes closed, with eyes open, to a remembered visual target, to a target without seeing their legs, and to a target while seeing their legs. Parkinson's disease subjects were separated into a moderate group and a severe group based on scores from the Unified Parkinson's Disease Rating Scale and were tested off and on their dopamine medication. Parkinson's disease subjects exhibited shorter compensatory steps than did the control subjects, but all subjects increased their step length when stepping to targets. Compared with the other subject groups, the severe Parkinson's disease subjects made larger accuracy errors when stepping to targets, and the severe Parkinson's disease subjects' step accuracy worsened the most when they were unable to see their legs. Thus, Parkinson's disease subjects exhibited short compensatory steps due to abnormal proprioceptive-motor integration and used visual input to take longer compensatory steps when a target was provided. In severe Parkinson's disease subjects, however, visual input does not fully compensate because, even with a target and unobstructed vision, they still exhibited poor step accuracy. Medication did not consistently improve the length and accuracy of the Parkinson's disease subjects' compensatory steps, suggesting that degeneration of dopamine circuits within the basal ganglia is not responsible for the proprioceptive-motor deficit that degrades compensatory steps in Parkinson's disease subjects.

Multiple balance tests improve the assessment of postural stability in subjects with Parkinson's disease.

OBJECTIVES: Clinicians often base the implementation of therapies on the presence of postural instability in subjects with Parkinson's disease (PD). These decisions are frequently based on the pull test from the Unified Parkinson's Disease Rating Scale (UPDRS). We sought to determine whether combining the pull test, the one-leg stance test, the functional reach test, and UPDRS items 27-29 (arise from chair, posture, and gait) predicts balance confidence and falling better than any test alone. METHODS: The study included 67 subjects with PD. Subjects performed the one-leg stance test, the functional reach test, and the UPDRS motor exam. Subjects also responded to the Activities-specific Balance Confidence (ABC) scale and reported how many times they fell during the previous year. Regression models determined the combination of tests that optimally predicted mean ABC scores or categorised fall frequency. RESULTS: When all tests were included in a stepwise linear regression, only gait (UPDRS item 29), the pull test (UPDRS item 30), and the one-leg stance test, in combination, represented significant predictor variables for mean ABC scores (r2 = 0.51). A multinomial logistic regression model including the one-leg stance test and gait represented the model with the fewest significant predictor variables that correctly identified the most subjects as fallers or non-fallers (85% of subjects were correctly identified). CONCLUSIONS: Multiple balance tests (including the one-leg stance test, and the gait and pull test items of the UPDRS) that assess different types of postural stress provide an optimal assessment of postural stability in subjects with PD.