Use this URL to cite or link to this record in EThOS:
Title: Automatic monitoring of physical activity related affective states for chronic pain rehabilitation
Author: Olugbade, Temitayo A.
ISNI:       0000 0004 7231 0577
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Chronic pain is a prevalent disorder that affects engagement in valued activities. This is a consequence of cognitive and affective barriers, particularly low self-efficacy and emotional distress (i.e. fear/anxiety and depressed mood), to physical functioning. Although clinicians intervene to reduce these barriers, their support is limited to clinical settings and its effects do not easily transfer to everyday functioning which is key to self-management for the person with pain. Analysis carried out in parallel with this thesis points to untapped opportunities for technology to support pain self-management or improved function in everyday activity settings. With this long-term goal for technology in mind, this thesis investigates the possibility of building systems that can automatically detect relevant psychological states from movement behaviour, making three main contributions. First, extension of the annotation of an existing dataset of participants with and without chronic pain performing physical exercises is used to develop a new model of chronic disabling pain where anxiety acts as mediator between pain and self-efficacy, emotional distress, and movement behaviour. Unlike previous models, which are largely theoretical and draw from broad measures of these variables, the proposed model uses event-specific data that better characterise the influence of pain and related states on engagement in physical activities. The model further shows that the relationship between these states and guarding during movement (the behaviour specified in the pain behaviour literature) is complex and behaviour descriptions of a lower level of granularity are needed for automatic classification of the states. The model also suggests that some of the states may be expressed via other movement behaviour types. Second, addressing this using the aforementioned dataset with the additional labels, and through an in-depth analysis of movement, this thesis provides an extended taxonomy of bodily cues for the automatic classification of pain, self-efficacy and emotional distress. In particular, the thesis provides understanding of novel cues of these states and deeper understanding of known cues of pain and emotional distress. Using machine learning algorithms, average F1 scores (mean across movement types) of 0.90, 0.87, and 0.86 were obtained for automatic detection of three levels of pain and self-efficacy and of two levels of emotional distress respectively, based on the bodily cues described and thus supporting the discriminative value of the proposed taxonomy. Third, based on this, the thesis acquired a new dataset of both functional and exercise movements of people with chronic pain based on low-cost wearable sensors designed for this thesis and informed by the previous studies. The modelling results of average F1 score of 0.78 for two-level detection of both pain and self-efficacy point to the possibility of automatic monitoring of these states in everyday functioning. With these contributions, the thesis provides understanding and tools necessary to advance the area of pain-related affective computing and groundbreaking insight that is critical to the understanding of chronic pain. Finally, the contributions lay the groundwork for physical rehabilitation technology to facilitate everyday functioning of people with chronic pain.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available