Assesment of physical activity in amputees: A systematic review of the literature

Understanding LSTM Network Behaviour of IMU-Based Locomotion Mode Recognition for Applications in Prostheses and Wearables

Human Locomotion Mode Recognition (LMR) has the potential to be used as a control mechanism for lower-limb active prostheses. Active prostheses can assist and restore a more natural gait for amputees, but as a medical device it must minimize user risks, such as falls and trips. As such, any control system must have high accuracy and robustness, with a detailed understanding of its internal operation. Long Short-Term Memory (LSTM) machine-learning networks can perform LMR with high accuracy levels. However, the internal behavior during classification is unknown, and they struggle to generalize when presented with novel users. The target problem addressed in this paper is understanding the LSTM classification behavior for LMR. A dataset of six locomotive activities (walking, stopped, stairs and ramps) from 22 non-amputee subjects is collected, capturing both steady-state and transitions between activities in natural environments. Non-amputees are used as a substitute for amputees to provide a larger dataset. The dataset is used to analyze the internal behavior of a reduced complexity LSTM network. This analysis identifies that the model primarily classifies activity type based on data around early stance. Evaluation of generalization for unseen subjects reveals low sensitivity to hyper-parameters and over-fitting to individuals’ gait traits. Investigating the differences between individual subjects showed that gait variations between users primarily occur in early stance, potentially explaining the poor generalization. Adjustment of hyper-parameters alone could not solve this, demonstrating the need for individual personalization of models. The main achievements of the paper are (i) the better understanding of LSTM for LMR, (ii) demonstration of its low sensitivity to learning hyper-parameters when evaluating novel user generalization, and (iii) demonstration of the need for personalization of ML models to achieve acceptable accuracy.

Reported physical activity and quality of life in people with lower limb amputation using two types of prosthetic suspension systems

BACKGROUND

Vacuum-assisted suspension systems provide better suspension than non-vacuum systems, but data are limited on whether they improve physical activity levels and quality of life for people with amputation.

OBJECTIVES

To compare the physical activity and quality of life levels of people with transtibial amputation using PIN/LOCK suspension system or vacuum-assisted suspension systems with those of able-bodied controls and to investigate parameters associated with physical activity levels.

STUDY DESIGN

A cross-sectional observational study.

METHODS

Fifty-one people with amputation and 51 controls participated. The International Physical Activity Questionnaire Short Form and Short Form 36 were used to measure the physical activity and quality of life, respectively.

RESULTS

The total physical activity and Short Form 36 scores were significantly lower in the participants with amputation than the controls. There were no significant differences between the two types of suspension systems in terms of physical activity levels and quality of life. The vacuum-assisted suspension system users reported significantly more bodily pain on the Short Form 36 questionnaire than the controls (p = 0.003). The only parameter that correlated significantly with the total physical activity was the Short Form 36 physical functioning subscale (r = 0.302, p = 0.031).

CONCLUSION

Contrary to our expectations, vacuum-assisted suspension system users compared to PIN/LOCK users did not report greater levels of physical activity or improved quality of life or levels closer to comparable controls.

CLINICAL RELEVANCE

A better understanding of the effects of different prosthetic suspension systems on physical activity and quality of life may help clinicians when prescribing prostheses, as well as setting appropriate prosthetic expectations. This study suggests that vacuum-assisted suspension systems and PIN/LOCK suspension systems provide equal benefit to users with regards to physical activity and quality of life.

Predicting ambulatory energy expenditure in lower limb amputees using multi-sensor methods

Purpose

To assess the validity of a derived algorithm, combining tri-axial accelerometry and heart rate (HR) data, compared to a research-grade multi-sensor physical activity device, for the estimation of ambulatory physical activity energy expenditure (PAEE) in individuals with traumatic lower-limb amputation.

Methods

Twenty-eight participants [unilateral (n = 9), bilateral (n = 10) with lower-limb amputations, and non-injured controls (n = 9)] completed eight activities; rest, ambulating at 5 progressive treadmill velocities (0.48, 0.67, 0.89, 1.12, 1.34m.s^-1) and 2 gradients (3 and 5%) at 0.89m.s^-1. During each task, expired gases were collected for the determination of V˙O2 and subsequent calculation of PAEE. An Actigraph GT3X+ accelerometer was worn on the hip of the shortest residual limb and, a HR monitor and an Actiheart (AHR) device were worn on the chest. Multiple linear regressions were employed to derive population-specific PAEE estimated algorithms using Actigraph GT3X+ outputs and HR signals (GT3X+HR). Mean bias±95% Limits of Agreement (LoA) and error statistics were calculated between criterion PAEE (indirect calorimetry) and PAEE predicted using GT3X+HR and AHR.

Results

Both measurement approaches used to predict PAEE were significantly related (P<0.01) with criterion PAEE. GT3X+HR revealed the strongest association, smallest LoA and least error. Predicted PAEE (GT3X+HR; unilateral; r = 0.92, bilateral; r = 0.93, and control; r = 0.91, and AHR; unilateral; r = 0.86, bilateral; r = 0.81, and control; r = 0.67). Mean±SD percent error across all activities were 18±14%, 15±12% and 15±14% for the GT3X+HR and 45±20%, 39±23% and 34±28% in the AHR model, for unilateral, bilateral and control groups, respectively.

Conclusions

Statistically derived algorithms (GT3X+HR) provide a more valid estimate of PAEE in individuals with traumatic lower-limb amputation, compared to a proprietary group calibration algorithm (AHR). Outputs from AHR displayed considerable random error when tested in a laboratory setting in individuals with lower-limb amputation.