Automatic Malaria Detection from Images of Microscopic Thin Blood Films
| dc.contributor.advisor | Assabe, Yaregal (PHD) | |
| dc.contributor.author | Zemene, Daniel | |
| dc.date.accessioned | 2018-06-18T11:16:18Z | |
| dc.date.accessioned | 2023-11-04T12:23:35Z | |
| dc.date.available | 2018-06-18T11:16:18Z | |
| dc.date.available | 2023-11-04T12:23:35Z | |
| dc.date.issued | 2016-03 | |
| dc.description.abstract | Malaria is one of the most common infectious diseases, causing wide spread sufferings and deaths in various parts of the world. The accurate and timely diagnosis of malaria infection is an essential requirement to control and cure the disease. Automated method of detecting malaria parasite improves accuracy, saves time, reduces the required human resource and minimizes human errors. In line with this, determination of parasitemia is also a crucial step to measure the amount of Plasmodium parasites in the patient's blood, an indicator for the degree of infection. Dealing with this, many researches have proposed different algorithms to solve the problem. However, accuracy of detecting the presence of parasites and estimating parasitemia under the occurrence of cells with complex structural arrangement is still regarded as a challenging task. Hence, in this research, a better technique to address Plasmodium parasite detection and computing parasitemia on the basis of counting red blood cells (RBCs), also called erythrocytes, that are extracted from microscopic thin blood films is looked for. After images of microscopic thin blood film are acquired, we apply color median filter as a preprocessing activity to enhance the quality of images. A novel segmentation technique is proposed to isolate background from foreground (RBCs and parasites) and also to separate Plasmodium parasites from the rest of blood cell components. A hybrid of adaptive threshold and color structure tensor threshold is applied to come up with a better segmentation result. We use a total of 11 features derived from geometric, color structure tensor and color characteristics to detect RBCs and parasites. A novel red blood cells arrangement identification technique is proposed to deal with overlapping RBCs, which enables us to compute parasitemia even with the presence of high degree of RBCs overlap. The proposed algorithms are tested using sample image data collected from different sources. A diagnostic accuracy of 93 % is achieved to detect Plasmodium parasites with sensitivity and specificity values of 94% and 91%, respectively. This was achieved even in the context where poorly illuminated microscopic images and complex red blood cells’ arrangement existed. The accuracy of the proposed algorithm to compute parasitemia is found to be better than a system developed based on watershed segmentation technique, which is very popular and frequently used in many research works. Keywords: Plasmodium parasites, Color structure tensor, Parasitemia, Watershed segmentation, Red blood cell arrangement | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/123456789/1303 | |
| dc.language.iso | en | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | Plasmodium Parasites;Color Structure Tensor; Parasitemia, Watershed Segmentation; Red Blood Cell Arrangement | en_US |
| dc.title | Automatic Malaria Detection from Images of Microscopic Thin Blood Films | en_US |
| dc.type | Thesis | en_US |