Show simple item record

dc.contributor.advisor Mokwena, S. N.
dc.contributor.author Baloyi, Vekani Reviet
dc.date.accessioned 2025-10-08T09:25:08Z
dc.date.available 2025-10-08T09:25:08Z
dc.date.issued 2025
dc.identifier.uri http://hdl.handle.net/10386/5090
dc.description Thesis (Ph. D. (Computer Science)) -- University of Limpopo, 2025 en_US
dc.description.abstract Recent developments in artificial intelligence have allowed significant gains in computer vision, cybersecurity, e-Commerce, and healthcare through the application of machine learning and deep learning models. A plethora of applications have offered effective ways to support physicians and radiologists in their medical imaging analysis, which continues to be the fundamental component of the visual representation used to formulate the final observation and diagnostic. Expert knowledge in data science and computer engineering has recently been combined with medical studies in oncology. In this regard, automatic assistance, sometimes referred to as computer-aided diagnosis (CAD) systems, has gained popularity as a field for study and development in recent decades. As a result, experience and multidisciplinary knowledge were used in the creation of CAD systems, which are now used to analyse patient data and assist physicians and physicians in making decisions. Every day, radiologists and oncologists tackle the vital task of diagnosing, evaluating, and treating cancer, in addition to preventing it. To provide decision support for many applications in cancer patient care operations, such as lesion detection, cancer staging, characterisation, recurrence, tumour evaluation, and prognosis forecasting, a computer-aided design (CAD) system may be developed. Throughout the world, breast cancer is believed to be one of the most prevalent cancers that affect women. It was also thought to be the main cause of death for women, and the number of deaths from it has been rising yearly. Early detection and diagnosis of breast cancer through regular screenings is the most effective way to significantly improve the chances of successful treatment by identifying breast cancer at its early stages. Cancer arises from uncontrolled cell division that allows these cells to invade surrounding tissues. This uncontrolled growth is often triggered by mutations within genes that are the blueprints for cellular function. Breast cancer is a highly heterogeneous disease that exhibits variations in tumour biology, aggressiveness, and response to treatment. In South Africa, breast cancer is the most common cancer among women, posing a significant health threat. Early detection is crucial, as it greatly improves treatment outcomes. While experienced physicians play a vital role in diagnosis, machine learning systems are emerging as promising tools with high accuracy in cancer identification. The primary goal of this research was to create a model that can assist radiologists in identifying and categorising breast cancer. The Mankweng Hospital repository provided the mammogram images that were used to create the deep learning algorithm. In transfer learning, a pre-trained model, VGG19, InceptionV3 and MobilenetV2, was utilised for fine-tuning the model. A convolutional neural network (CNN) model was developed and optimised using techniques to determine the best batch size, learning rate, epoch, and optimise parameters. During training, the InceptionV3 model achieved the best accuracy of 88%. The models generated are capable of classifying breast cancer cases. However, for some classes, there was not enough data available. This study applied augmentation to address the over-fitting of data. Thus, the next steps of this dissertation involve collecting a lot of data for every class and creating a more reliable categorisation model. This dissertation therefore suggests a new model for breast cancer diagnostics that is based on the latest advances in computer vision and machine learning technologies. The best method to identify breast lesions early on and to reduce the risk of death is mammography screening. It helps to expose breast anomalies such as microcalcification, architectural distortion, and mass lesions. Having an additional reading tool or help system could improve the breast cancer diagnostic procedure, as the number of patients examined every day is always increasing. Several modalities, including an X-ray scanner and a full-field digital mammography (FFDM) system, can be used to obtain mammograms. The ultimate diagnosis may depend on the quality of the mammograms, the attributes (such as density and size) or the attributes (such as location, size, and form). Consequently, radiologists run the risk of overlooking the lesions, which could lead to incorrect diagnosis and detection. Therefore, the objective of this effort was to improve mammography reading to increase the accuracy of difficult assignments. Incoming research may incorporate novel approaches to techniques by merging several mammography data sets and enhancing the extended training of deep learning models. Motives could also improve the model by including more breast cancer lesions, such as calcification and architectural distortion, using annotated information. In this dissertation, a model was presented to help medical professionals and specialists determine the probability of the presence of breast cancer. All things considered, the suggested model approach combines the latest developments in deep learning, image processing, and picture-to-image translation for biomedical use. en_US
dc.format.extent xvii, 167 leaves en_US
dc.language.iso en en_US
dc.relation.requires PDF en_US
dc.subject Breast cancer en_US
dc.subject Deep learning en_US
dc.subject Computer-aided diagnosis en_US
dc.subject Convolutional neural network en_US
dc.subject Detection en_US
dc.subject Classification en_US
dc.subject Multiple classification en_US
dc.subject.lcsh Breast -- Cancer en_US
dc.subject.lcsh Machine learning en_US
dc.subject.lcsh Cancer -- Early detection en_US
dc.subject.lcsh Artificial intelligence -- Computer programs en_US
dc.title Early detection of breast cancer using machine learning en_US
dc.type Thesis en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search ULSpace


Browse

My Account