Browsing by Author "Ofosu Mensah, Samuel"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Analysing retinal fundus images with deep learning models
    (Stellenbosch : Stellenbosch University, 2023-12) Ofosu Mensah, Samuel; Bah, Bubacarr; Brink, Willie; Stellenbosch University. Faculty of Science. Dept. of Mathematical Sciences. Applied Mathematics Division.
    ENGLISH ABSTRACT: Convolutional neural networks (CNNs) have successfully been used to classify diabetic retinopathy but they do not provide immediate explanations for their decisions. Explainability is relevant, especially for clinicians. To make results explainable, we use a post-attention technique called gradient-weighted class activation mapping (Grad- CAM) on the penultimate layer of deep learning models to produce localisation maps on retinal fundus images after using them to classify diabetic retinopathy. Moreover, the models were initialised using pre-trained weights obtained from training models on the ImageNet dataset. The results of this are fewer training epochs and improved performance. Next, we predict cardiovascular risk factors (CVFs) using retinal fundus images. In detail, we use a multi-task learning (MTL) model since there are several CVFs. The impact of using an MTL model is the advantage of simultaneously training for and predicting several CVFs rather than doing so individually. Also, we investigate the performance of the fundus cameras used to capture the retinal fundus images. We notice a superior performance of the desktop fundus cameras to the handheld fundus camera. Finally, we propose a hybrid model that fuses convolutions and Transformer encoders. This is done to harness the benefits of convolutions and Transformer encoders. We compare the performance of the proposed model with other attention-based models and observe on-par performance.
  • Thumbnail Image
    Item
    Transcriptomic profile based cancer disease prediction and patient survival time differentiation
    (Stellenbosch : Stellenbosch University, 2018-12) Ofosu Mensah, Samuel; Mazandu, Gaston Kuzamunu; Utete, Simukai Wanzira; Stellenbosch University. Faculty of Science. Dept. of Mathematical Sciences. Division Applied Mathematics.
    ENGLISH ABSTRACT : Cancer disease is an abnormal growth of cells, which may be caused by mutations in genes which, as a result, alter the way cells function mainly in the way they grow and divide. Cancer cells are regulated by complex interactions mediated by a group of proteins and miRNAs which are expressed and repressed. With the help of transcriptomic technologies such as RNA–sequencing (RNA–seq), it is now possible to profile thousands of genes at once to create a global picture of the functions of cells. Here, the study employs a statistical approach, called Significance Analysis of Microarray (SAM), to identify genes that are differentially expressed in breast cancer patients. Genes with scores greater than a threshold are deemed potentially significant. Genes identified as significantly different are used for twofold reasons. First, the study uses these significantly identified genes to predict breast cancer using three machine learning algorithms. The machine learning algorithms used are random forests, artificial neural networks and support vector machines. Secondly, clinical details of patients and significantly identified genes are combined to build a survival model to predict the probability of survival and risk to the event in breast cancer patients. Using The Cancer Genome Atlas (TCGA) as the primary data for the study, SAM reported 23 genes as significantly different. Further investigations revealed that these 23 significant genes are involved in tumour suppression, angiogenesis, cell growth factor, tumourigenesis, cell proliferation, tumour progression and tumour necrosis activities. In predicting breast cancer, 10 out of the 23 genes contribute significantly to the model. Finally, it was identified that log–logistic distribution best describes the survival time of breast cancer patients. Moreover, the survival model revealed that expression levels of six genes influence the survival probability of a breast cancer patient.