Welcome to WIReDSpace(Wits Institutional Repository on DSpace)

For technical support contact Lewatle Phaladi by email: lewatle.phaladi@wits.ac.za, Tel : +27 11 717 6594.


Recent Submissions

Water purifi
(2022-06) Dept of Water
Water purification
Perceptions surrounding Madikwe Game Reserve’s current management: implications for sustainability and future climate change responses
(2022) Galvao, Amy
Ecotourism Based Protected Areas (ETBPAs) are becoming more vulnerable to climate change impacts, hindering their ability to safeguard ecosystems and ecotourism products. A bottom-up understanding of ETBPA’s social-ecological systems could potentially provide climate change impact management solutions. Using a case study approach, this study gains a bottom-up understanding of Madikwe Game Reserve’s (MGR) social-ecological system through the perceptions of 31 onsite community members (OSCMs) through semi-structured interviews. These perceptions focus on MGR’s current management and the implications for MGR’s, and similar ETBPA’s, sustainability, both currently and in the face of future climate change. Findings indicate governance inefficiency, ecotourism-focused management priorities, and scientific knowledge gaps have caused major implications for MGR’s ecological sustainability, mainly through the cascading impacts of an elephant overpopulation. These management issues also limit MGR’s ability to respond to climate change impacts both currently and in the future.
Knowledge discovery through mining mixed data types for cancer diagnostic and prognostic prediction modelling
(2022) Achilonu, Okechinyere Juliet
Background: Cancer is an increasingly multidimensional global health problem that demands continuous and concerted efforts toward diagnosis, treatment and management. Developing countries, including South Africa, are expected to experience an increasing burden of cancer morbidity and mortality because of a late-stage diagnosis, poor awareness, lack of biomedical screening and limited resources for treatment. Nonetheless, there is a growing trend to improve treatment outcomes by understanding the epidemiology and dynamics of cancer. The multifaceted clinical approach toward cancer has resulted in an unprecedented production of a large amount of complex and heterogeneous clinical data. These include information describing disease symptoms, diagnostics reports, drug usage and clinical outcomes, often presented in a structured, semi-structured and unstructured data format. Physicians’ intuition may not efficiently integrate these mixed data types in this situation for a realistic prognosis and treatment. Therefore, the next generation of cancer treatments may be found not by empirical science but by data scientists developing mathematical and statistical models to describe the nature, epidemiology and dynamics of cancer at various stages of the disease. Hence, cancer research is attracting the application of high-performance computing and big data analytics to enhance diagnosis, prognosis, and treatment. It has been shown that the use of data mining (DM), text mining (TM) and machine learning (ML) has efficiently and effectively uncovered trends from cancer data to support oncologists in decision-making. However, Africa is yet to embrace clinical decision-making in cancer diagnosis and prognosis using these algorithms. Current ML models for cancer detection and prognostic classifications have been chiefly designed for developed countries. Nevertheless, findings in the developed countries may not truly reflect the situation in an African population with higher diversity. Aim: This research aimed to contribute to knowledge discovery using descriptive and predictive analytics, specifically in colorectal, breast and prostate cancers. One of the objectives of this study is to develop and evaluate methods for extracting relevant information to classify cancer pathology reports. This will be achieved by analysing a large pool of historical pathology reports and assessing performance across cancer types. Secondly, there is only limited study on biological and clinical parameters which drives cancer prognosis, especially in Africa. This study aimed to develop a rule-based method to automatically extract important cancer prognostic parameters from free-text pathology reports, transform them into structured data, and uncover the trend of these parameters over the years. In addition, studies have shown that hospital length of stay (LOS) following surgical cancer resection varies across countries. However, cancer studies on hospital length of stay are unknown in an African population. This study aimed to present an avenue to understand the dynamics and prognostic classification of patient risk groups post-surgical recession. Finally, cancer recurrence and patient survival studies are limited in an African population. We explored the feasibility of integrating statistical and machine learning algorithms for the first time to achieve higher predictive performance and interpretability models for cancer recurrence and patient survival. Method: We conducted secondary data analysis using two data sources based on the study aims. The first data set consists of 181,000 breast, colorectal and prostate cancer patients whose diagnostic tests were carried out at the National Health Laboratory Services (NHLS) from 2008 to 2019. Each patient is described with demographic information, pathology reports, SNOMED morphology and topography codes. The data set addressed this study’s first and second objectives, focusing on information extraction from pathology reports, text classification and prediction, and trend analysis. The second data set consisted of 761 colorectal cancer patients diagnosed in Johannesburg. The patients were treated in private and public hospitals from 2015 to 2019 and followed until 2020. Each patient is described with more features detailing their demographic, clinical and histological information. We used this data set to address the third and fourth objectives, which focused on prognostic classification and prediction, with machine learning and statistical algorithms. Results: The incidence of breast, colorectal and prostate cancer diagnostic tests trends is increasing over the year. In recent years, pathology reporting on these cancers has increased in text length, with more malignancy than benign. The result also showed inconsistencies and incompleteness in reporting each year and across the year of study. However, our methodologies could standardise and accurately extract essential parameters with high performance. The second objective showed that the developed rule-based method achieved high accurate annotation for all the parameters extracted, with performance measures ranging from 83% -100%. The trend analysis result showed significant trends in the proportion of molecular subtypes and Ki67, comparable to previous studies. Further, we observed that the median hospital LOS post-surgical resection was higher than those reported in the developed world. There is no significant difference in hospital LOS when comparing private and public facilities. Factors predisposing patients to prolonged hospital LOS include preoperative, perioperative and postoperative parameters. However, surgical complications are still the primary driver of prolonged hospital LOS. Finally, younger patients experienced higher recurrence than older patients but had a comparable survival rate. There was an improved survival for patients treated in a private hospital compared to those treated in public hospitals. Conclusion: We developed reproducible frameworks that can form the basis for future studies in South Africa, using DM, TM and ML algorithms. This study supports a nationally agreed standard in pathology reporting and the use of these algorithms for encoding, classifying, and producing high-quality information abstractions for cancer diagnosis, prognosis, incidence reporting, and research. The association established in this study may enable clinicians to understand the diagnostic and prognostic factors that influence patients’ health status and implement changes in patient care pathways. Clinicians can employ this type of study as a simple and quick test to flag patients at high risk. Such a strategy would improve clinical outcomes, but it is also likely to improve efficiency, favourably impacting the cost of care for cancer patients. The findings of this study can be generalised not only to the population of South African cancer patients but in other Sub-Saharan African countries with similar trends in urbanisation and dynamics in cancer epidemiology.
Assessing the acceptability of biometrics in HIV prevention programme by Hillbrow sex workers
(2019) Nyamhuno, Shepherd
Sex workers face several challenges in Hillbrow which include the risks of HIV and STI infection, stigma and discrimination in public health facilities, arrests and exploitation from the police as well as violence from the clients. Sex workers do not always bring their identity documents to the clinic and they are frequently changing their names thereby creating many multiple accounts for an individual. The biometrics have the potential to solve the problems of unique identifiers for Esselen Street Clinic. This study undertook to find out if the biometrics were acceptable to the sex workers. A mixed-methods approach made up of questionnaires, focus group discussions and in-depth interviews for employees was employed. A sample of 120 questionnaires found out that 64.6% accepted the biometrics when used at Esselen Street Clinic. Moreover, 79% had a positive perception of biometrics. The focus group discussion showed that although the sex workers trusted the clinic staff, they had problems with fear of the unknown including someone hacking the system and the data ending up in the hand of future prospective employees. Sex workers also cited that they were using biometrics in other platforms such as the Department of Home Affairs, banks and SASSA. This gave them the confidence to accept them if they were to be placed at the clinic. In order to understand the acceptance of the biometrics by the sex workers, in-depth interviews with employees were carried out. It emerged from the in-depth interviews with the staff members that the clinic had invested much in tailor-made services, which are very convenient for sex workers. Moreover, the continual capacity building in sex worker sensitisation and the quality improvement programme has yielded better outcomes and quality services that are attractive to sex workers. The non-judgemental attitudes of the employees gave the sex workers confidence to accept the biometrics. It was interesting to note that migrant sex workers were more sceptical and fearful of biometrics than local ones due to illegal immigrant status and experience of harassment at the hands of the police. The study concluded that a sex worker-friendly clinic, like Esselen Street Clinic, could perhaps have a better acceptance rate than a public health facility would. With intense lobbying and education, there are higher chances of getting an even higher acceptance rate of the biometrics.
Urban continuity :the urban regeneration of Fordsburg through integrating the mall (the Oriental Plaza) and the grid (urban fabric)
(2020) Da Silva, Paulo
This research explores urban continuity in a part of Johannesburg. The objective is to broadly investigate the history of Fordsburg, its role as a mixed use precinct and to address the issues caused by 1970s imposed Modernism. Cities reveal the relationship between the individual and the greater whole. They bring people together for the mutual benefit and support of each other. This demonstrates the co-existence of different layers of history and embodies a level of continuity - this is the first reference to the title of this research. Continuity does not signify a linear one after the other system, but rather it is an interpretation of time and of physical space, in which the past continues to exist in the present. Michael Focault (1980), an historian and philosopher, states: “space is history.” We all have relationships with the city that rely on the physical formation of the city and its civic history to determine our identities. It is a complex process that is best understood through urban morphology. Oliveira et al (2015) states that there is more to urban morphology that also reveals the elements of a city. These elements form typologies which have genealogies (Dovey et al, 2016). Continuity also applies to the urban patterns produced in cities, which are revealed from examining the physical environment (Alexander et al, 1977). The physical environment, which is an element of urban morphology, focuses on the act of habitation and spatial principles (von Meiss, 2015). It is an appropriate subject because it forms the basis for urbanity and the background atmosphere for living in possible happiness and comfourt. The physical urban environment has codes and conventions that need to be learned, understood and challenged. Fordsburg, a metropolitan suburb in Johannesburg which embodies decades of history and functions within two extremes of Johannesburg’s physical environment: it possesses some of the best qualities of urbanity and also some of the poorest. The site of these urban extremes is the Oriental Plaza: which is a modernist suburban mall masquerading as an urban market. To the west of the mall exists a bastion of urbanity and to the east of the mall is an environment that has become desolate and lost. Auge (1995) and Trancik (1986) describe poor urban spaces as anti-space or lost space, that doesn’t not provide any real benefit for urban living. The challenge is therefore to understand how to connect the eastern and western urban fabric, regenerate the lost urban spaces and stitch urban typologies together meaningfully. The principles of urban continuity (Dewar et al, 1992) offer an appropriate design tool to promote positive change. The principles of urban continuity also offer to the opportunity to realize the mall as a permeable catalyst for Fordsburg’s connection to Johannesburg’s eastern fringe