Monitoring and Evaluation of Posture in Office Workstations With Artificial Intelligence

Overview

Office workers spend 11.6 hours per day in sedentary activities during the day, leading to increased health risks. Although there is literature on prevalence of musculoskeletal disorders and work station modification for office workers, successful implementation and follow-through of evidence-based recommendations in office environments is extremely challenging. This study aims to validate various techniques for evaluating posture, particularly in the neck, shoulder, and trunk, in order to develop evidence-based feedback for the office desk user. This study will include up to 40 volunteers over the age of 18 who regularly work on computer-based tasks at a desk. Participants will be observed working in a provided work station in a variety of different working postures. Data are collected using electromyography sensors, Kinect camera, and at least two video recording cameras to obtain postural information. Measurement of joint positioning and posture will be completed using an ergonomic screening tool from images collected throughout the testing. Participants will provide information on their level of comfort throughout various body regions in response to working in various positions. This will be an iterative study with multiple positions used to test the positioning of sensors, and the development of suitable algorithms to evaluate posture. The collected data will be used to inform sensing methods for office work posture.

There are more than 81 million office workers in the U.S. These office workers spend 11.6 hours per day in sedentary activities during the day. Several studies looked at workplace adjustments to decrease the health risks of sedentary office work, but there is little conclusive evidence showing that the modifications led to decreased risk of musculoskeletal pain. Ergonomic chair interventions have been found to reduce musculoskeletal symptoms and but evidence is inconclusive in supporting different chairs. Similarly, current literature on different computer adjustments (i.e. keyboard and mouse) has been inconclusive. When evaluating work stations that allow for both sitting and standing posture, prolonged static posture in either position can increase musculoskeletal discomfort. Many ergonomic interventions have become commercially available, but there is a lack of conclusive evidence showing that these interventions reduce risk for musculoskeletal pain. The most successful interventions, meant to promote worker health, include an individualized, behavioral component. Workers who receive individualized attention, have higher overall awareness of ergonomic guidelines, demonstrate significantly better postures, and report higher productivity than their counterparts who receive group education or no intervention. Furthermore, adding real time prompts to an individualized approach can further increase the success of an intervention, with one study indicating that prompts may have up to four-fold effect on outcomes. Finally, it has been shown that the giving users the choice in how they implement ergonomic recommendations, input into methods and timing of cuing, and flexibility in adjusting their work environment demonstrate significant and sustained improvements in postures, productivity and overall well-being. This study will provide valuable input into the sensing mechanism for an intelligent workstation that will adapt the workspace in a way that will promote productivity, health and well-being through consistent behaviors to increase activity and improve posture. This study will contribute to existing knowledge by: 1. Develop sensing methods that accurately measure postural parameters. 2. Develop guidelines for postural sensing (i.e. type of sensor, suitable algorithms for analysis, placement of sensor) of the trunk, neck, and upper extremities of office workers. 3. Understand individual position changes and the related/associated level of comfort. If successful, the results of this study can be used to determine posture in order to provide real-time evidence-based ergonomic feedback for the workplace to decrease risk of musculoskeletal pain for office workers. Method and Procedures The primary objective of this study is to validate various techniques for evaluating posture at an office workstation and relate these postures to physical discomfort. The data will be collected by taking baseline descriptive information on age, gender, and upper body discomfort using the Visual Analog Scale. Participants will be asked to independently arrange the given workstation in any way to work comfortably, and perform office work-like tasks. Then, the participants will be asked to work in a series of alterations to the workstation to test various different postures of the neck, shoulder, arms, and trunk. These will include neck rotation (0-45 degrees), neck lateral flexion (0-45 degrees), neck flex/ext (0-60 degrees), shoulder elevation and scapular protraction, shoulder flex (0-120 degrees), shoulder abduction (0-90 degrees), shoulder horizontal rotation (0-90 degrees), elbow flex/ext (0-145 degrees), wrist flex/ext (0-80 degrees), wrist ulnar/radial deviation (0-20 degrees), trunk rotation, trunk kyphosis/lordosis, and trunk lateral flexion. Participants will rotate between the postures when they become uncomfortable with their current posture and will be observed for a minimum of 2 hours and a maximum of 4 hours per session. Participants may be asked to return for up to two observation sessions. Participants will report any pain, discomfort, and overall comfort in each region of the body every 10-15 minutes. Participants will be informed to notify the study personnel of any increase in discomfort and testing in that posture will be terminated. There will be a termination protocol for clinically significant pain increase (more than 2 points on the Visual Analog Scale). The study may include up to 40 participants. The sample size was selected to allow for multiple rounds of data collection, as this will be an iterative study to test various methods to sense posture. The Kinect/electromyography will be coded to track joint angles and posture in real-time. The video recording will be analyzed separately after the observation session using the Rapid Upper Limb Assessment and automated computer algorithms. This study will be an iterative process and camera positions and postures may be adjusted to validate various techniques for posture evaluation. The data from the Kinect/electromyography, video recording, original intended posture, and participant self report will be compared during data analysis.