Mobile depth sensing technology and algorithms with application to occupational therapy healthcare

Abstract

The UK government is striving to shift its current healthcare delivery model from clini-cian–oriented services, to that of patient and self–care–oriented intervention strategies. It seeks to do so through Information Communication (ICT) and Computer Mediated Re-ality Technologies (CMRT) as a key strategy to overcome the ever–increasing scarcity of healthcare resources and costs. To this end, in the UK the use of paper–based information systems have exhibited their limitations in providing apposite care. At the national level, The Royal College of Occupational Therapists (RCOT) identify home visits and modifica-tions as key levers in a multifactorial health programme to evaluate interventions for older people with a history of falling or are identified as being prone to falling. Prescribing Assistive Equipment (AE) is one such mechanism that seeks to reduce the risk of falling whilst promoting the continued independence of physical dexterity and mobility in older adults at home. In the UK, the yearly cost of falls is estimated at £2.3 billion. Further evidence places a 30% to 60% abandonment rate on prescribed AE by and large due to a ‘poor fit’ and measurement inaccuracies. To remain aligned with the national strategy, and assist in the eradication of measurement inaccuracies, this thesis employs Mobile Depth Sensing and Motion Track-ing Devices (MDSMTDs) to assist OTs in in the process of digitally measuring the extrin-sic fall–risk factors for the provision of AE. The quintessential component in this assess-ment lies in the measurement of fittings and furniture items in the home. To digitise and aid in this process, the artefact presented in this thesis employs stereo computer–vision and camera calibration algorithms to extract edges in 3D space. It modifies the Sobel–Feldman convolution filter by reducing the magnitude response and employs the camera intrinsic parameters as a mechanism to calculate the distortion matrix for interpolation between the edges and the 3D point cloud. Further Augmented Reality User Experience (AR-UX) facets are provided to digitise current state of the art clinical guidance and over-lay its instructions onto the real world (i.e., 3D space). Empirical mixed methods assessment revealed that in terms of accuracy, the arte-fact exhibited enhanced performance gains over current paper–based guidance. In terms of accuracy consistency, the artefact can rectify measurement consistency inaccuracies, but there are still a wide range of factors that can influence the integrity of the point-cloud in respect of the device’s point-of-view, holding positions and measurement speed. To this end, OTs usability, and adoption preferences materialise in favour of the artefact. In conclusion, this thesis demonstrates that MDSMTDs are a promising alterna-tive to existing paper–based measurement practices as OTs appear to prefer the digital–based system and that they can take measurements more efficiently and accurately.