School of Computing and Information Science

Permanent URI for this communityhttp://localhost:4000/handle/20.500.12504/206

Browse

Now showing 1 - 10 of 10

Adaptive-size dictionary learning using information theoretic criteria for image reconstruction from undersampled k-space data in low field magnetic resonance imaging
(BMC Medical Imaging, 2020-06) Emmanuel, Ahishakiye; Martin, Bastiaan Van Gijzen; Julius, Tumwiine; Johnes, Obungoloch
Background Magnetic resonance imaging (MRI) is a safe non-invasive and nonionizing medical imaging modality that is used to visualize the structure of human anatomy. Conventional (high-field) MRI scanners are very expensive to purchase, operate and maintain, which limit their use in many developing countries. This study is part of a project that aims at addressing these challenges and is carried out by teams from Mbarara University of Science and Technology (MUST) in Uganda, Leiden University Medical Center (LUMC) in the Netherlands, Delft University of Technology (TU Delft) in the Netherlands and Pennsylvania State University (PSU) in the USA. These are working on developing affordable, portable and low-field MRI scanners to diagnose children in developing countries with hydrocephalus. The challenges faced by the teams are that the low-field MRI scanners currently under development are characterized by low Signal-to-Noise Ratio (SNR), and long scan times. Methods We propose an algorithm called adaptive-size dictionary learning algorithm (AS-DLMRI) that integrates information-theoretic criteria (ITC) and Dictionary learning approaches. The result of the integration is an adaptive-size dictionary that is optimal for any input signal. AS-DLMRI may help to reduce the scan time and improve the SNR of the generated images, thereby improving the image quality. Results We compared our proposed algorithm AS-DLMRI with adaptive patch-based algorithm known as DLMRI and non-adaptive CSMRI technique known as LDP. DLMRI and LDP have been used as the baseline algorithms in other related studies. The results of AS-DLMRI are consistently slightly better in terms of PSNR, SNR and HFEN than for DLMRI, and are significantly better than for LDP. Moreover, AS-DLMRI is faster than DLMRI. Conclusion Using a dictionary size that is appropriate to the input data could reduce the computational complexity, and also the construction quality since only dictionary atoms that are relevant to the task are included in the dictionary and are used during the reconstruction. However, AS-DLMRI did not completely remove noise during the experiments with the noisy phantom. Our next step in our research is to integrate our proposed algorithm with an image denoising function.
Blockchain technology needs for sustainable mineral supply chains: A framework for responsible sourcing of Cobalt.
(Elsevier, 2022) Godfrey, Mugurusi; Emmanuel, Ahishakiye
Blockchain technology has recently become the go-to solution for companies and industries that seek to enhance value chain traceability of their products, and transparency in their supply chains. Because of these benefits, it's been proposed for monitoring environmental, social, and governance (ESG) performance and compliance in industries that have weak regulatory and formal structures. The cobalt mining industry especially in the Democratic Republic of Congo, the world’s biggest producer of cobalt ore used in the manufacturing of lithium-ion batteries, is one such environment that’s characterized by conflict, and serious human rights abuses. The key actors in the cobalt supply chain, therefore, face the tradeoff involving maintaining long-term supply versus reducing the risks associated with cobalt sourced from locations with poor environmental and human rights records. Most of such problems emerge from Artisanal and small-scale mining. This paper presents an attempt to tightly link existing blockchain technology frameworks in the cobalt industry with ESG performance of companies to enable them to audit the chain of custody journeys for their components and ultimately sustainability performance. We present a responsible sourcing framework to connect blockchain source data needs to ESG metrics to help companies build interoperable but understandable blockchain architectures.
Classification of cassava leaf diseases using deep Gaussian transfer learning model
(Engineering Reports, 2023-03) Emmanuel, Ahishakiye; Ronald, Waweru Mwangi; Petronilla, Murithi; Fredrick, Kanobe; Danison, Taremwa
In Sub-Saharan Africa, experts visually examine the plants and look for disease symptoms on the leaves to diagnose cassava diseases, a subjective method. Machine learning algorithms have been employed to quickly identify and classify crop diseases. In this study, we propose a model that integrates a transfer learning approach with a deep Gaussian convolutional neural network model. In this study, two pre-trained transfer learning models were used, that is, Mobile Net V2 and VGG16, together with three different kernels: a hybrid kernel (a product of a squared exponential kernel and a rational quadratic kernel), a squared expo-nential kernel, and a rational quadratic kernel. In experiments using MobileNet V2 and the three kernels, the hybrid kernel performed better, with an accuracy of 90.11%, compared to 86.03% and 85.14% for the squared exponential kernel and a rational quadratic kernel, respectively. Additionally, experiments using VGG16 and the three kernels showed that the hybrid kernel performed better, with an accuracy of 88.63%, compared to the squared exponential kernel’s accuracy of 84.62% and the rational quadratic kernel’s accuracy of 83.95%, respectively. All the experiments were done using a traditional computer with no access to GPU and this was the major limitation of the study.
Comparative Performance of Machine Leaning Algorithms in Prediction of Cervical Cancer
(IEEE xplore, 2021-10) Emmanuel, Ahishakiye; Waweru, Mwangi; Petronilla, Muthoni; Lawrence, Nderu; Ruth, Wario
Cervical cancer is among the most common types of cancer affecting women around the world despite the advances in prevention, screening, diagnosis, and treatment during the past decade. Cervical cancer can be treated if diagnosed in its early stages. Machine learning algorithms like multi-layer perceptron, decision trees, random forest, K-Nearest Neighbor, and Naïve-Bayes have been used for the prediction of cervical cancer to aid in its early diagnoses. In this study, we compare the performance of ensemble methods (AdaBoost, Stochastic Gradient Boosting, Random Forests, and Extra Trees), and classification algorithms (K-Nearest Neighbor and Support Vector Machine) in the prediction of cervical cancer basin g risk factors. Ensemble methods and classification algorithms were used during this study. Ensemble methods were selected because they combine several machine learning techniques into one model to decrease variance, bias, or improve performance while the classification methods were selected because our dataset was generally categorical and therefore could work well with our problem domain. Experimental results revealed that all the algorithms did not perform well on the “imbalanced” dataset. Experiments on balanced revealed an improved performance. The performance metrics used include Fl-score, Area Under Curve (AUC), and Recall. Extra Trees performed better than the rest when using the Fl-score metric, Stochastic Gradient Boosting and Random Forest performed better than the rest when using the AUC metric, K-Nearest Neighbors outperformed the rest using the recall metric, and Extra Trees had the best accuracy 0.96. The application of machine learning methods in the prediction of cervical cancer using risk factors may lead to early detection of the disease which can be treated if diagnosed early. Six algorithms have been considered in this study. The general performance reveals that ensemble methods performed better than classification methods using both imbalanced and balanced datasets.
A comparative study of some pre-trained models on transfer learning approach in detection and classification of cassava leaf diseases
(SSRN, 2022) Emmanuel, Ahishakiye; Waweru, Mwangi; Petronilla, Muthoni; Danison, Taremwa; Fredrick, Kanobe
Cassava diseases affect cassava harvest posing the greatest danger to the food security and livelihoods of more than 200 million people. To identify cassava diseases, government professionals visit various sections of the country and visually score the plants by looking for disease indicators on the leaves. This procedure is notoriously subjective; it is not uncommon for specialists to differ on a plant's diagnosis. Automating the detection and classification of crop diseases could help professionals diagnose diseases more accurately and allow farmers in remote locations to monitor their crops without the help of specialists. Machine learning algorithms have been used in the early detection and classification of crop diseases. Motivated by the current developments and many influential studies in the field of deep learning and transfer learning models in the detection and classification of crop diseases, this study evaluates the performance of VGG16, VGG19, ResNet50, InceptionV3, DenseNet201, and MobileNetV2 in detection and classification of cassava leaf diseases. Fine-tuning of the hyperparameters was done during training to improve the accuracy of the models. Experimental results on the cassava dataset revealed that InceptionV3, DenseNet201, and MobileNetV2 models had high training accuracy but low validation accuracy with various epochs which means that they had issues with over fitting while ResNet50 had issues with underfitting. Moreso, VGG16 and VGG19 models performed well on both training and validation datasets, though VGG16 performed relatively well compared to VGG19.
Consumer segmentation and profiling using demographic data and spending habits obtained through daily mobile conversations
(American Journal of Modeling and Optimization, 2018-08) Samuel, W. Kamande; Evans, A. K. Miriti; Emmanuel, Ahishakiye
Knowledge of customer behaviour helps organizations to continuously re-evaluate their strategies with the consumers and plan to improve and expand their application of the mosteffective strategies. Using expenditure data collected through daily mobile conversations with consumers in Kenya, this study sought to compare various clustering algorithms and establish one that best segments consumers, and subsequently providing profiles that provide a basis for marketing and brand strategy based on existing demographic data – age, gender, region and primary income source. K-Means, Hierarchical and Partitioning around Medoids (PAM) clustering algorithms were compared using internal and stability validation tests. Hierarchical clustering with four clusters had the best Connectivity (0.847) and Silhouette width (0.924) measures. Stability validation compares the results by removing a column, one at a time. Average Proportion of Non-overlap (APN), Average Distance (AD), Average Distance Between Means (AND) and Figure of Merit (FOM) were used to compare the algorithms. Again, Hierarchical clustering with four clusters was found to partition the data best. The study forms a basis for the use of additional profile descriptors once available to provide a firmer understanding of the customer segments built on expenditure data in Kenya.
A dictionary learning approach for joint reconstruction and denoising in low field magnetic resonance imaging
(2021 IST-Africa Conference (IST-Africa), 2021-10) Emmanuel, Ahishakiye; Martin, Bastiaan VAN Gijzen; Xiujie, Shan; Julius, Tumwiine; Johnes, Obungoloch
Currently, many children with hydrocephalus in East Africa and other resource-constrained countries do not have access to Magnetic Resonance Imaging (MRI) scanners, the preferred imaging tool during the disease administration and treatment. Conventional MRI scanners are costly to buy and manage, which limits their utilization in low-income countries. Low-field MRI scanners can offer an affordable, sustainable, and safe imaging alternative to high-field MRI. However, they are associated with a low signal-to-noise ratio (SNR), and therefore the images obtained are noisy. In this study, we propose an algorithm that may help to alleviate the drawbacks of low-field MRI by improving the quality of images obtained. The proposed algorithm combines our previous proposed algorithm known as AS-DLMRI for image reconstruction and a nonlinear diffusion filter for image denoising. The formulation is capable of removing additive zero-mean white and homogeneous Gaussian noise, as well as other noise types that could be present in the original signal. Experiments on visual quality revealed that the proposed algorithm is effective in denoising images during reconstruction. The proposed algorithm effectively denoised a noisy phantom, and a noisy MRI image, and had better performance when compared to DLMRI and AS-DLMRI in terms of Peak Signal to Noise ratio (PSNR) and High-Frequency Error Norm (HFEN). Integrating AS-DLMRI and the nonlinear diffusion filter proved to be effective in improving the quality of the images during the experiments performed. The hybrid algorithm may be of great use in imaging modalities like low-field MRI that are associated with low SNR.
An ensemble model based on learning vector quantization algorithms for early detection of cassava diseases using spectral data
(SpringerLink, 2023-03) Emmanuel, Ahishakiye; Waweru, Mwangi; Petronilla, Murithi; Ruth, Wario; Fredrick, Kanobe; Taremwa, Danison
In Sub-Saharan Africa, cassava is the second most significant food crop after maize. Cassava brown streak disease (CBSD) and cassava mosaic virus disease (CMD) combined account for nearly 90% of productivity losses. Automating the detection and classification of crop diseases could help professionals diagnose diseases more accurately and allow farmers in remote locations to monitor their crops without the help of specialists. Machine learning algorithms have been used in the early detection and classification of crop diseases. Previous research has used plant image data captured with smartphones. However, disease symptoms must be observable to use this strategy (using image data). Unfortunately, once symptoms appear on the aerial part of the plant, the root, which is the edible part of the plant, is destroyed. In this study, we used spectral data in a three-class classification challenge for diagnosing cassava diseases. We propose an ensemble model based on Generalized Learning Vector Quantization (GLVQ), Generalized Matrix LVQ (GMLVQ), and Local Generalized Matrix LVQ (LGMLVQ). Experimental results revealed that the LGMLVQ model had the best overall performance on the precision, recall, and F1-score followed by our proposed ensemble model, the GMLVQ model performed third, and finally GLVQ model. Also, using an accuracy performance metric, LGMLVQ had overfitting issues even though it had the highest accuracy of 100%, followed by our proposed ensemble model with an accuracy of 82%, and then the third in performance was the GMLVQ model with an accuracy of 74% and the least performed model on accuracy was GLVQ model with an accuracy of 56%.
A secure web based records management system for prisons: a case of Kisoro prison in Uganda
(International journal of computer, 2017) Emmanuel, Ahishakiye; Danison, Taremwa; Elisha, Opiyo Omulo
Most Prisons in the developing countries are still using the traditional system – pen and papers, to keep track of their records. This system takes long to finish a single transaction; this has led to loss of information of some cases (crimes files), insecurity and data redundancy. Similarly, some cases have been reported where some prison staff connives with clients (victims) to change and hide some information or files hence leading to compromising the evidence of the matter. This has consequently resulted in time wastage to handle cases, increased corruption and insecurity of important files hence making the whole process costly. Also when reports are needed especially about prisoners, it takes a long time and therefore makes it hard for Prison Management to take urgent decisions. This has created a lot of loopholes in the system because there is no tracking and/or monitoring of the information available in the different Departments and there are no security measures in place to safe guard the available information. This necessitated automating the system to make it more efficient and effective. There was close study of the existing manual file based system that was in use, it was compared to the proposed system. A prototype of a proposed system was developed to ease data access, security and retrieval for instant report production by the prison management. The prototype was developed using MySql database, PHP, CSS, JavaScript and HTML.
A survey on deep learning in medical image reconstruction
(Elsevier, 2021-03) Emmanuel, Ahishakiye; Martin, Bastiaan Van Gijzen; Julius, Tumwiine; Ruth, Wario; Johnes, Obungoloch
Medical image reconstruction aims to acquire high-quality medical images for clinical usage at minimal cost and risk to the patients. Deep learning and its applications in medical imaging, especially in image reconstruction have received considerable attention in the literature in recent years. This study reviews records obtained elec- tronically through the leading scientific databases (Magnetic Resonance Imaging journal, Google Scholar, Scopus, Science Direct, Elsevier, and from other journal publications) searched using three sets of keywords: (1) Deep learning, image reconstruction, medical imaging; (2) Medical imaging, Deep learning, Image reconstruction; (3) Open science, Open imaging data, Open software. The articles reviewed revealed that deep learning-based re- construction methods improve the quality of reconstructed images qualitatively and quantitatively. However, deep learning techniques are generally computationally expensive, require large amounts of training datasets, lack decent theory to explain why the algorithms work, and have issues of generalization and robustness. The challenge of lack of enough training datasets is currently being addressed by using transfer learning techniques.

Browse

Browsing School of Computing and Information Science by Author "Emmanuel, Ahishakiye"