Introduction: Spinal cord injuries (SCI) are seen commonly in Southern Africa and can completely change the course of the affected's life. Lower limb disability is a common complication from this injury, but a patient can be rehabilitated in some cases. Research and clinical observations suggest that early mobilisation and rehabilitation leads to shorter hospital stays and better clinical outcomes. Relieving the time dedication placed onto the rehabilitation team could mean that patients receive a higher standard of care. Methods: A cyclic movement device has been designed to mimic the gait cycle that a patient is attempting to recover. The device was intended towards providing a ground reaction force simulation at the correct points of the gait cycle. The device was tested in-silico with validated skeletal models to determine joint torques and angles. In-silico testing was also utilised to determine the loads placed onto the patient by the device through its use. The force data could then be used to predict possible ground reaction forces. Results: The device allows for a gait similar trace path of the ankle, comparable to that found in the literature. The ankle has a range of motion of 3 1° as the device completes a full cycle in which the crank rotates 360 °. The hip has a range of motion of 28° and the knee 35° in this same movement. The shape of the displacements of the joints of the lower limb is comparable to that seen in researched gait patterns. However, the timing of the knee and hip joints' movements are not synchronous with that of the gait patterns. The device is validated to be sufficiently stable to use, and the motor and power components can provide the 7259N.mm of torque needed to move the model. Conclusion: The results suggest that the device has potentia l as an adjunct to rehabilitation schemes. In-silico testing showed that the device is able to simulate some of the kinetic and kinematic parameters seen in normal gait. Further work is needed to prototype the device to physically and clinically validate the device.

No description available.

High disease morbidity coupled with limited healthcare personnel places the health sector in Kenya under strain, leaving parts of the population with limited access to health services. Electronic health (eHealth), the utilisation of information and communication technologies in healthcare, is an innovation with the potential to improve access to health services. Several examples exist of eHealth projects being undertaken in Kenya. However, eHealth solutions have been poorly adopted in the public healthcare sector, which has partly been blamed on lack of knowledge amongst healthcare providers and patients. The aim of this study was to examine how knowledge is exchanged between the stakeholders currently active within the eHealth implementation space in the Kenyan public sector. The results of the study would aid in identifying communication breakdowns and ways of increasing information flow with regard to eHealth, and ultimately would aid strategies to help improve the uptake of eHealth within the public sector. A mixed methods study was undertaken that combined quantitative social network analysis and qualitative analysis of semi-structured interviews conducted with stakeholders involved in implementation of eHealth projects in Kenya. Publications on implementation of eHealth projects in Kenya from 2001 to 2018 were used to obtain data on relevant organisations. Social network analysis was used to identify prominent actors. Individuals working within such organisations were invited to participate in semi-structured interviews. Further social network analysis was applied to data gathered through the interviews. Foreign universities and foreign not-for-profit organisations were the most commonly identified organisations in the networks generated. The tacit nature of knowledge within networks, low research capacity and output, information guarding, geographical distance between collaborating organisations, and low cohesion were some of the factors found to inhibit knowledge diffusion within the eHealth implementation space in Kenya. The search for capacity and funding were found to contribute to network structure. eHealth knowledge management strategies should be given attention, for enhanced exchange of knowledge within the public health sector in Kenya.

Chronic Kidney Disease (CKD) is a disease that causes kidney damage, often leading to the patient requiring haemodialysis treatment. Haemodialysis treatment requires a vascular access method, commonly Arteriovenous (AV) fistulae and grafts. These access methods must be regularly assessed to ensure the access remains unblocked and the flow rate is normal. Phase Contrast MRA (PC-MRA) is a versatile Magnetic Resonance Imaging (MRI) modality which is capable of imaging and quantifying blood flow in vivo. It is for this reason that this imaging technique was used to image blood flow in the vasculature of the upper arm of volunteers and haemodialysis patients with either an AV fistula or graft. This imaging technique is capable of producing temporally resolved Three-dimensional (3D) datasets (known as "Four-dimensional (4D)" flow) of blood flow in major vessels. Velocities are phase encoded between -π and π based on the chosen Velocity Encoding Constant (venc). To successfully characterise all velocities in the volume it is necessary to set the venc to be approximately equal to the highest velocity found in the vessel. Any lower venc value will cause phase wrapping, an imaging artefact causing all higher velocities to be wrapped by a multiple of 2 π. However, the increase in sensitivity to high velocities reduces the overall specificity of the velocities, especially for low velocities. Due to the pulsatile nature of blood flow in arterial vessels, a large range of velocities are encountered, while venous flow is more constant but lower than the peak arterial flow value. For this reason and due to the length of the 4D flow scans, 20-30 minutes, it would be preferable to perform one scan at a relatively low venc and correct any phase wrapping during post-processing. In this study, we performed both Two-dimensional (2D) PC-MRA scans at various locations in the upper arm and 4D PC-MRA scaans with similar venc settings. The purpose of the study was to implement and test several methods of phase unwrapping to remove phase wrapping artefacts from affected areas within the PC-MRA datasets.

Background: Stroke is the leading cause of adult disability with 70 to 85% of initial strokes resulting in hemiparesis. Physical imparity as a result of stroke tends to be severe and majority of impairments are upper limb-related. Impairment is usually accompanied by long term functional loss which requires dedicated post-stroke rehabilitation to regain motor function. The incidence of stroke is increasing rapidly while there remains a shortage of therapists to provide sufficient rehabilitation. There is therefore a high demand for therapists to attend to the rising number of stroke survivors. Robot-aided therapy has emerged as a beneficial tool for providing continuous rehabilitation of the upper limb and is widely being implemented. With this technology, there is great potential to reduce the ill-effects brought about by the low therapist-patient ratio which has hindered sufficient rehabilitation and consequently the effective recovery of motor function among stroke survivors. Hypothesis: The use of a portable robotic rehabilitation system, as a complementary tool, in hand therapy, would promote continuous rehabilitation by encouraging repetition of task oriented exercises which would enhance motor function of an impaired hand. Task-oriented writing practice would potentially improve hand coordination and result in better accuracy while repetitive training would potentially increase hand motor strength. Objectives: 1.To design and manufacture a portable robotic rehabilitation system. 2. To test the performance and usability of the system. Methods: The system was manufactured and its performance tested in a pilot pre-clinical trial involving three participants. The system's ease of use was assessed using a standardised usability scale. Writing accuracy and hand motor strength were also assessed and the results analysed at the end of the study. Results: The average overall score of usability for the rehabilitation system was a few points higher than the average score. The users of the system also experienced increased motivation whilst performing the repetitive and task oriented exercises. There was an improvement in the completion time of the writing accuracy test and the tasks of the trace sample test. The variation in grip strength of the non-dominant hand during the rehabilitation period was small for each of the participants. Conclusion: The rehabilitation system motivated its users to repetitively perform rehabilitative training which may have improved writing accuracy.

This thesis investigated a multi-class auditory P300 BCI as a step towards FES applicability. A multi-class P300 paradigm approach provides degrees-of-freedom in operating an FES device over the traditional P300 paradigm. Accuracy in classification of target P300s contributes to the paradigm's applicability in a 'real' environment. The computational effectiveness of the paradigm can be enhanced through signal processing prior to classification. A combination of principal component analysis (PCA) and independent component analysis (ICA), together with a method of enhancing the P300 properties through temporal and spatial manipulation are investigated as a means of improving classification accuracy. The combination of these techniques and the use of a multi-class P300 paradigm presents a different approach as a step towards FES applicability in an auditory BCI.

This study describes the development of a filtration technique applied to the Lodox Statscan linear slotscanning digital X-ray system to reduce radiation dose to paediatric patients whilst preserving diagnostic image quality. The Statscan is an FDA approved, commercially available digital X-ray system commonly used for trauma and emergency patients. The Statscan provides significantly lower radiation dose to patients than conventional Xray systems for comparable studies without loss of image quality. This is particularly beneficial in paediatric radiology, where the risks associated with ionizing radiation are much higher. A static dose prediction model for the Statscan which was previously developed at the University of Cape Town has been adapted to create a dynamic dose prediction model which allows the user to adjust the system scanning parameters. The model calculates the patient entrance dose from an energy spectrum generated using the input parameters. The effective dose for a paediatric sized patient is then calculated using a Monte Carlo simulation. The dynamic model allows for variation of the scan parameters and direct observation of the expected dose levels for specific examinations. Filtration is a well-known technique for reducing radiation dose, where a filter material is placed in the path of the X-ray beam to reduce patient exposure to radiation. The dynamic model was used to design a new filtration technique for the paediatric settings on the Statscan.

The need for economic treatment of coronary heart disease in South Africa and the popularity of coronary stent implantation as an effective treatment increases. Stent implants have had to be imported to date with significant cost involved. The project in hand therefore deals with the evaluation of the possibilities of economical local manufacture of a coronary stent. The design of several coronary stents currently in use was carefully analysed and the results of clinical application described in the literature were reviewed. After evaluating the patent situation, a suitable design concept was chosen, based primarily on the availability of manufacturing equipment. The design concept of choice is the balloon expandable slotted tube variety. Stents of this most popular slotted tube design are usually laser cut out of a tube with good results. Since there was no suitable laser cutting equipment in use in South Africa, it was undertaken to manufacture stent prototypes by spark erosion to establish the general suitability of the method for stent manufacturing. The initial design was adjusted several times to the manufacturing process and suitable machine parameters were evaluated. For comparison a stent of a similar design was laser cut by an established European manufacturer. The geometry of the most favourable prototypes was analysed by computer simulation using finite element analysis, followed by simple physical tests of deployment and radial strength. The quality of manufacturing was evaluated with electron microscopy. The computational and mechanical test results of the design analysis yielded largely satisfactory results, but it was found that the manufacturing process of spark erosion is not suitable to manufacture products that would have to compete on the current sophisticated market. With the spark erosion equipment, neither a satisfying surface quality nor acceptable geometrical tolerances could be achieved. This was particularly obvious after comparison with the laser cut stent sample of the same design. An improvement in quality of the spark erosion process would require extensive fine-tuning and would increase the already very long manufacturing time and the cost significantly. Subcontracted laser cutting in Europe was found to be more reliable and economical.

No description available.

Includes bibliographical references (leaves 120-128). / Present-day obstetric decision-making is based on measuring the umbilical arterial blood flow velocity waveforms from one site of the cord. There is an ongoing debate on the predictive value of Doppler measurements in the evaluation of the foetal condition. The aim of this thesis is to investigate the use ofa neural network to recognise blood flow waveform shape patterns associated with placental insufficiency. Eleven backpropagation neural networks have been developed and trained based on the waveforms that are generated from the foetal mathematical model (developed in previous research) at both ends of the cord. Only two networks trained successfully. These two networks are the Levenberg-Marquardt algorithm (Trainlm) and the resilient backpropagation algorithm (Trainrp).