UCHEATU, MICHAEL IFEANYICHUKWU (2026) AUTOMATED DETECTION OF KIDNEY STONES IN COMPUTER TOMOGRAPHY (CT) IMAGES USING CONVOLUTIONAL NEURAL NETWORKS (CNN). Other thesis, Godfrey Okoye University, Enugu Nigeria.
|
Text
UCHEATU KIDNEY STONE 1.pdf Download (2MB) | Preview |
Abstract
Renal calculi are one of the most frequently encountered urological diseases that have to be properly diagnosed in order to prevent complications. It is widely believed that Computer Tomography (CT) imaging represents an efficient way to diagnose renal calculi but the interpretation of CT scans may be a time-consuming process and is subject to certain human errors. The main purpose of this work is to create an automated system of diagnosing kidney stones using Computed Tomography (CT) images and the technique of Convolutional Neural Networks (CNN). A set of kidney CT images that are available in public domain was divided into two groups, namely "Stone" and "Non-Stone"Image pre-processing steps, like resizing, normalization, grayscale conversion, and data augmentation, have been applied to enhance the model performance. An automatic framework of CNN is proposed to extract important features from CT images and classify them accordingly. The results obtained by the model include an accuracy of 52.6%, a recall of 87.6%, a precision of 17.4%, and F1 score of 0.29, based on the confusion matrix with 241 true positives, 1,069 true negatives, 34 false negatives, and 1,144 false positives from a test set of 2,488 CT images. The model has been designed to be applicable in practice and implemented through web-based application with Flask technology and role- based access, where radiographers are able to upload images, get their prediction and patient's information immediately. The application of deep learning is discussed in the paper, with the particular focus on the role that can be played by CNN-based deep learning approach in terms of improving early diagnosis, simplifying the process of diagnosis and providing health care services in resource-limited settings.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | Q Science > Q Science (General) |
| Divisions: | Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science |
| Depositing User: | Cynthia Ugwuoti |
| Date Deposited: | 22 Jul 2026 10:52 |
| Last Modified: | 22 Jul 2026 10:52 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/5988 |
Actions (login required)
 |
View Item |
