OBA, Emmanuel Babatunde and Covenant University, Theses Masters (2024) AN AUTOMATED MALARIA DIAGNOSIS SYSTEM FOR DETECTING LIFE CYCLE STAGES OF PLASMODIUM IN THIN BLOOD SMEAR IMAGES. Masters thesis, Covenant University.
![[img]](http://eprints.covenantuniversity.edu.ng/style/images/fileicons/application_pdf.png) |
PDF
Download (217kB) |
Abstract
The worldwide influence of malaria has hastened the creation and execution of innovative diagnostic approaches aimed at combating the disease. Several attempts have been made to develop an automated malaria diagnosis system (AMDS), but most of these systems are trained to perform binary classification to distinguish between infected and uninfected Plasmodium falciparum parasites. This study investigates the application of deep learning techniques in developing an Automated Malaria Diagnosis System (AMDS) to enhance the accuracy and efficiency of malaria life cycle stage detection from thin blood smear images. The main objectives are to curate a novel dataset of thin blood smear images with the lifecycle stages of Plasmodium and to leverage pre-trained convolutional neural networks (CNNs) to identify the life cycle stages of Plasmodium, which are responsible for malaria. The study specifically focuses on evaluating the performance of different CNN architectures, including VGG16, VGG19, ResNet50, and MobileNet. The methodology involved curating a dataset of annotated thin blood smear images representing various life stages of the Plasmodium parasite. This dataset was then used to fine-tune the selected CNN models. The models were evaluated based on metrics such as accuracy, precision, recall, F1-score, and support. The results showed that VGG16 achieved the highest accuracy of 0.72, but its precision (0.47) and F1-score (0.49) indicated room for improvement in classification performance. VGG19, while slightly lower in accuracy at 0.71, demonstrated better precision (0.55) and recall (0.64), resulting in a higher F1-score (0.56). ResNet50, commonly recognized for its robustness in other domains, underperformed with an accuracy of 0.61 and a notably lower recall of 0.26. MobileNet displayed moderate results, with an accuracy of 0.68 and balanced precision and recall values. The findings suggest that VGG19 offers a more balanced performance for malaria stage classification, making it a promising candidate for deployment in a clinical setting. However, further optimization and refinement of these models are necessary to improve diagnostic precision and reliability. The study concludes that deep learning models, particularly VGG19, hold significant potential in supporting malaria diagnosis and contributing to more effective disease management and control strategies.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Automated Malaria Diagnosis System, Convolutional Neural Networks, Deep Learning, Plasmodium, Thin Blood Smear Images, Web Application |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science R Medicine > R Medicine (General) |
| Divisions: | Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science |
| Depositing User: | Patricia Nwokealisi |
| Date Deposited: | 30 Sep 2024 09:03 |
| Last Modified: | 30 Sep 2024 09:03 |
| URI: | http://eprints.covenantuniversity.edu.ng/id/eprint/18463 |
Actions (login required)
 |
View Item |
