Ram, M (2022) RISK DETECTION FOR THE PREVENTION OF FALLS ON STAIRS IN OLDER PEOPLE. Doctoral thesis, Liverpool John Moores University.

Preview

Text 2022MalarPhD.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (6MB) | Preview

Abstract

One of the most hazardous and demanding tasks for older people is stair negotiation, often resulting in falls. These falls lead to significant loss of mobility and independence. In addition, after stair falls, quality of life is affected, often leading to serious complications. These stair falls cost National Health Service (NHS) approximately £2.3bn per year. A laboratory-based motion capture system and force plates are currently used to investigate older people's fall risk factors while on the stairs to prevent falls whilst ascending or descending stairs. However, these procedures are costly, needs a dedicated motion laboratory, and use expensive and cumbersome equipment. Laboratory-based measurements help identify key fall risk factors that affect people in a controlled environment. However, implementing effective prevention measures requires regular monitoring of these factors over a long period in the user’s environment (home), where a fall is likely to occur. Therefore, there is an urgent need for low-cost devices to measure these risk parameters in a home environment and provide accurate and repeatable results to detect people who are all at risk. This work aimed to detect the stair fall risk factors using wearable sensors incorporated in shoes and machine learning algorithms to detect risk factors to improve the ability of older adults to negotiate different staircase environments safely. This aim had been achieved through four studies: In the first study, wearable sensors were incorporated into a standard shoe to detect stair-fall risk parameters. These parameters were identified as stair fall risk parameters in a controlled laboratory environment. The fall risk parameters were foot clearance and foot contact length ratio. A sensor insole was designed to detect foot contact length ratio (foot overhang) using force-sensitive resistors. Two distance sensors (VL6180X) were attached to the shoe to detect foot clearance for ascending and descending. In addition, BNO055 IMU was fitted in the shoe to measure foot motion kinematics such as velocity and acceleration. In the second study, the developed wearable sensor shoe was tested in the laboratory for validation against the current motion system’s standard biomechanical risk factors during stair negotiation. Foot clearance was validated with an accuracy of 0.05mm, and the precision was between 4.79mm to -4.67mm. Foot contact length ratio was validated with an accuracy of -2%, and the precision was between 10% to -13.91%. The developed sensor shoe was tested in the third study at different exemplar houses staircases to measure stair fall risk factors. In addition, a comparison was made between lab and houses to measure stair-negotiation behaviour changes. The results showed a significant difference in selected stair fall biomechanical factors among the three exemplar houses and laboratory stairs. In the fourth study, the supervised machine learning algorithm was trained to classify fall risk using collected sensor data along with self-reported falls. The support vector machine algorithm was trained to classify stair fall risk with a precision of 90.4%, a sensitivity of 88.9%, and an F-measure of 90%. So, this trained algorithm can be used in the future to predict stair fall risk at home environments based on data collected from the developed sensors and instrumented shoes. Future research should include more stair fall risk factors and more people in real-life stair negotiation conditions (houses) to improve future stair fall risk prediction.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Stair fall risk detection; Machine learning; wearable sensors to detect stair fall risk
Subjects:	R Medicine > R Medicine (General) R Medicine > RC Internal medicine > RC1200 Sports Medicine
Divisions:	Sport & Exercise Sciences
Date Deposited:	01 Feb 2022 12:11
Last Modified:	01 Feb 2023 00:50
DOI or ID number:	10.24377/LJMU.t.00016198
Supervisors:	Baltzopoulos, B, Andy, S, Maganaris, C, O'Brien, T and Cullen, J
URI:	https://researchonline.ljmu.ac.uk/id/eprint/16198

View Item