Impairments in hand function lead to reduced independence and quality of life after cervical spinal cord injury (SCI). In order to develop effective rehabilitation interventions for individuals with cervical SCI, it is important to assess hand function throughout the rehabilitation process. Currently, the efficacy of new treatments is measured by assessments limited to a controlled setting or based on self-report;​ there is currently no viable method to collect quantitative information once the patient has returned to the community. This thesis attempts to solve this gap by developing a computer vision-based wearable camera system for monitoring hand use. Our research involved (1) the collection of egocentric video that represents activities of daily living, (2) the development of an algorithm that captures interactions between the hand and objects in the environment, and (3) the evaluation of the system in both laboratory and home settings.

Four studies were executed, involving three video datasets (20 able-bodied participants and 17 participants with SCI in a home simulation laboratory, as well as 3 participants with SCI in their home). We introduced the concept of hand-object interaction detection, defined as a binary decision about whether or not the hand is manipulating an object for a functional purpose. The datasets were used in the development and evaluation of an algorithmic pipeline consisting of hand detection and segmentation, followed by hand-object interaction detection. For this step, a random forest classifier was trained on hand motion, hand shape and scene colour features. The frame-by-frame binary output data over time was further analysed to extract three functional hand-use metrics:​ (1) the amount of total interaction as a percentage of testing time, (2) the average duration of interactions in seconds, and (3) the number of interactions per hour. The final study investigated the views of participants with SCI on the use of wearable cameras.

With appropriate strategies determined by input from individuals with SCI, this thesis demonstrates the potential of a wearable egocentric camera as a unique tool to allow researchers and clinicians to gauge the user's level of independence at home in activities involving upper limb function.