Search

Ischaemic stroke, occurs due to an interruption in blood flow to the brain tissue, is the leading cause of disability and death worldwide and in the UK. Choosing a patient with ischaemic stroke for the best treatment is a critical step toward individual treatment planning and a successful outcome, as the effectiveness of treatment depends heavily on the time to treatment. However, it remains challenging to predict treatment outcomes for individual patients due to the complexity of data and visually subtle changes in the brain 3D NCCT scan. The work presented in this thesis investigates these challenges by developing novel deep learning approaches to predict functional outcome of ischaemic stroke treatment from baseline 3D NCCT and clinical information available at hospital admission. First, a multimodal CNN-based method is introduced and trained on baseline 3D NCCT scans with and without clinical information to estimate the functional outcome (mRS scores) of patients with ischaemic stroke. To further improve the model performance, two attention modules based on SE which help a CNN network in encoding the global relationship between features both channel-wise and spatially, are incorporated into the model. The results demonstrate that including clinical information and also using attention modules improves model performance. Furthermore, as the stroke lesion evolves - spreads or suppresses - after treatment, estimating stroke progression before treatment can provide significant information about the success rate of treatment and the condition of patients in the future. To encode this information, two CNN approaches, end-to-end and multi-stage models, are proposed. In the end-to-end method, predicting mRS scores and follow-up scans (24-hour and 1-week) are performed together, whereas the multi-stage approach comprises two stages of training, predicting ischaemic stroke evolution at one week without voxel-wise annotation and predicting functional outcome at 90 days. In these approaches, follow-up scans, which are used only during training, are reconstructed from the baseline scan to encode the evolution of stroke lesion. It is shown that encoding stroke evolution information into models increases the performance of models to predict functional outcome. This thesis also presents a transformer-based multimodal method that predicts mRS scores using baseline information. This method investigates different multimodal fusion strategies and various transformer models including ViT variants and Swin transformer. Transformer models that are trained with NCCT scans and clinical information outperform CNN-based approaches.

In recent years, the functional outcomes of patients with right hemisphere stroke (RHS) received considerable attention due to their impact on disability, independent living, quality of life and economic burden. Hemi-inattention (HI) is a complex condition which often accompanies RHS. It is characterised by reduced alertness, attention and low spatial awareness levels. Past studies reported poor outcomes in patients with HI and inconsistent findings in regard to the relationship of HI with functional outcome. Literature review of 13 relevant studies highlighted poor research methodology which complicated interpretation of previous results. Aims: The aim of this study was to address the clinically important question “What is the relationship between early HI status (HI±) and functional change in the 1st six months after right hemisphere stroke?” by improving on research methodology from past studies. Methods: An all-inclusive stroke severity RHS sample (58 with and 35 without HI) were recruited from two stroke units and assessed on motor and cognitive factors with validated measurement tools on four occasions; baseline, hospital discharge, 6 weeks after discharge, and 6 months after stroke. A multi-level modelling approach was used to analyse change in functional progress over time with potential explanatory motor and cognitive factors. Results: HI status was only statistically significant when modelled alone. Its predictive importance greatly diminished when modelled with other factors e.g. stroke severity, time since stroke and age. Conclusion: On average, HI group membership at baseline is unrelated to functional recovery when other influential factors are also considered. The findings extend current knowledge in stroke recovery research and provide suggestions for optimal therapeutic and rehabilitation outcomes. In contrast with traditional methods of regression analysis, multi-level modelling techniques enabled important relationships to be studied in depth. This resulted in new insights into the data which can be used to inform patient management and future research in the field.

Upper limb prostheses are difficult to control due to their fundamental differences from anatomical limbs, including different mechanical properties, reduced terminal device movements, and a lack of available feedback. In order to improve prosthetic function, there have been increased efforts toward altering device characteristics of existing prostheses, such as introducing multi-articulating hands and surgical interventions that allow acquisition of additional control signals. As these efforts continue, it is important to characterize outcomes in users of both existing and new prostheses with appropriate biomechanical measures and functional assessments. Therefore, the purpose of this dissertation was to quantify functional outcomes with conventional and advanced prostheses through measures of movement control, coordination, and performance of complex activities of daily living (ADLs). The first two aims quantified functional outcomes in users of conventional prostheses. Specifically, in the first aim, I compared movement control of prosthesis users and individuals without amputation by evaluating movement strategy during goal-directed reaching movements with spatial and temporal goals. Prosthesis users were able to effectively complete ballistic reaching with movement strategies and accuracy similar to healthy individuals but were unable to accurately complete movements requiring substantial feedback. In the second aim, I quantified differences in inter-joint coordination between prosthesis users and individuals without amputation during a range of unilateral and bilateral ADLs. Across all ADLs, prosthesis users completed multi-joint movements with reduced level of coordination compared to individuals without amputation. The dominant movement patterns were more dependent on the task than on limb type, however. The next two aims quantified functional outcomes in users of advanced prostheses with multi-articulating hands. Specifically, in the third aim I developed a new functional assessment to quantify performance during grasp transitions in users of multi-articulating prostheses. I validated the outcome measures (number of errors and completion time) of this task with healthy individuals who controlled a multi-articulating prosthetic emulator. This functional assessment (‘Coffee Task’) had easily interpretable outcomes with high inter-rater reliability and sensitivity to prosthetic skill. In the fourth aim, I used this assessment and a series of virtual classification tasks to determine whether the Regenerative Peripheral Nerve Interfaces (RPNIs) and intramuscular electromyography (EMG) electrodes can improve myoelectric grasp selection accuracy. The prosthesis user who participated in this case study was able to complete both virtual and physical tasks with greater accuracy with this new surgical approach compared to the commercial approach that uses surface EMG. Aims of my dissertation add to the existing body of literature that characterize the functional limits of conventional prosthesis users and demonstrate how emerging technology can improve expected function in individuals with limb loss. Findings from my work evaluating how conventional prosthesis users control and coordinate their movements will be useful in designing training regimens to reduce movement deficits in individuals with limb loss. Simultaneously, the novel functional assessment developed as a part of this dissertation may be used in the future to quantify performance associated with emerging advanced multi-articulating prostheses and associated control approaches. Continued evaluation of functional outcomes in both conventional and newly developed prostheses is important to ensure that efforts are appropriately directed to reduce movement challenges experienced by upper limb prosthesis users.

Stroke is a leading cause of disability in the United States (US) and the number of disabled stroke survivors will rise with the aging population. The aging of the US population will also lead to an increase in the prevalence of multiple chronic conditions (MCC) at stroke onset, which impacts post-stroke functional outcome (FO), although current evidence is inconsistent. The goal of this dissertation was to advance the understanding of MCC in post-stroke functional outcome through a systematic literature review and data analyses using a bi-ethnic population-based stroke cohort. Specifically, this dissertation investigated the association between MCC and post-stroke FO by 1) summarizing the findings of previous studies that have investigated the relationship between MCC and FO after ischemic stroke using MCC indices, 2) developing and internally validating a new MCC index to predict post-stroke FO using machine learning techniques, and 3) investigating the contribution of MCC to ethnic disparities in post-stroke FO between Mexican American (MA) and non-Hispanic white (NHW) stroke patients using the new index. The systematic literature review showed that hospital-based prognostic studies for post-stroke FO predominantly used the Charlson Comorbidity Index (CCI) and the Modified CCI (mCCI) to measure MCC burden. The negative association between MCC and FO was statistically significant in the meta-analysis. We identified several novel predictors of post-stroke FO and developed a new MCC index among ischemic stroke patients. The new MCC index was validated to improve the prediction of post-stroke FO at 90 days and outperformed the mCCI. MA stroke patients have significantly greater age-adjusted MCC burden at stroke onset compared to NHWs. MCC measured by the new index was found to be an important contributor to worse FO at 90 days in MAs compared with NHWs, although effect modification of the MCC-FO outcome association by ethnicity was not statistically significant. This dissertation research confirmed that MCC is an important predictor for post-stroke FO. External validation is needed before the application of this index in other stroke populations. Accurate measurement of the MCC burden in stroke patients is important for more precise prognosis in post-stroke FO, which could inform stroke treatment, post-acute care, and intervention efforts to mitigate post-stroke functional impairment, promote functional gain and lessen ethnic disparities in post-stroke FO.

Treatment for rectal cancer can be associated with rectal functional impairment, increased morbidity and may impact quality of life. Restorative surgery is the preferred choice of surgery, but in some circumstances, surgery with a permanent stoma is required to ensure tumour clearance. While the aim of treatment for rectal cancer is to ensure cure where possible the long term impact of the functional impact of this treatment is not known. This thesis investigates rectal functional impairment after treatment for rectal cancer in the long-term and its impact on quality of life in rectal cancer compared with age matched colon cancer patients. It also assesses the impact of radiotherapy on this impairment and its effect on quality of life in rectal cancer. In addition, it compares the quality of life in patients with and without a primary stoma after curative surgery for colorectal cancer and determines the impact of country of birth on the perception of a stoma. This thesis finally examines the long term morbidity, local recurrence and survival associated with radiotherapy and provides a protocol for future studies to better evaluate the effects of rectal cancer treatment. Patients treated for rectal cancer have greater rectal functional impairment compared to patients treated for colon cancer. In patients treated for rectal cancer, those receiving radiotherapy had even greater rectal dysfunction. This high rate of rectal dysfunction did not impact on the global quality of life. Patients with a permanent stoma had more restricted physical activity those without a stoma but there were no differences in mental functioning. There were no differences in the overall quality of life associated with country of birth. In this cohort of patients there was no difference in morbidity between those patients who had surgery alone compared with those who had radiotherapy and surgery. The rates of local and distant recurrence were not different between the groups as was overall survival. This thesis has shown that the functional outcomes after treatment for rectal cancer should be an important factor in decision making and that the role of radiotherapy in the treatment of rectal cancer should be re-evaluated.

Abstract: The objective of this study is to determine variables associated with four to five year functional outcomes in children undergoing complex cardiac surgery early in life. This is a longitudinal follow-up study conducted between 2000 and 2005, consisting of 165 survivors of complex cardiac surgery at the age of 6 weeks or younger at the Stollery Children’s Hospital in Edmonton. When children were four to five years of age, parents completed the Adaptive Behavioral Assessment System (ABAS) II. The study focuses on the following four outcomes: General Adaptive Composite of ABAS, and three broad categories of the children’s assessed adaptation in Conceptual, Social and Practical domains. Multivariate linear regression analysis revealed pre-operative plasma lactate, gender and mother’s education were significantly associated with all of the four outcomes in the study. The number of days ventilated was significantly associated with the Conceptual domain. Single Ventricle was associated with the Practical domain.