Automatic Breast Cancer Detection from Biopsy fine Needle Aspiration Microscopic Images
| dc.contributor.advisor | Assabe, Yaregal (PhD) | |
| dc.contributor.author | Nasir, Sebahadin | |
| dc.date.accessioned | 2019-11-20T09:01:50Z | |
| dc.date.accessioned | 2023-11-09T11:26:01Z | |
| dc.date.available | 2019-11-20T09:01:50Z | |
| dc.date.available | 2023-11-09T11:26:01Z | |
| dc.date.issued | 2018-01-04 | |
| dc.description.abstract | Breast cancer is the top type of cancer in women, it takes 25% of all cancer cases worldwide. In Ethiopia, Hospital archives indicates that more than 200,000 cancer cases register annually where a breast cancers is among the top two types of cancer having a high death rate in the country. An accurate cancer diagnosis is vital for effective treatment. At this time in Ethiopia, human experts or pathologists perform breast cancer diagnostic manually. Conversely, manual diagnostic needs experienced pathologists and much amount of time. Automated technique of detecting breast cancer improves accuracy and saves the required diagnosis time. The main objective of this study is to develop a system that perform breast cancer diagnostic from sample microscopic biopsy images using digital image processing techniques based on the standard for fine needle aspiration cytology categories by the American Cancer Institute guideline. A segmentation algorithm has been proposed to segment the epithelial cells from the background region by considering both overlapping epithelial cells and non-uniform illumination effects in the given microscopic slide image. To represent sample epithelial cell 30 features (16 color, 8 geometric and 6 texture) have been used. A feedforward artificial neural network classification model has been designed with 30 input and 4 output nodes, consistent to the number of features and classes respectively to classify epithelial cell samples. The designed network has been trained using 800 sample fine needle aspiration microscopic epithelial cells images. The data is arbitrarily divided into training (70%) validation (15%) and testing (15%). The overall classification accuracy of the classifier is 97.8%. The accuracy for detecting the class benign, abnormal, suspicious, malignant cells are 95.0%, 100%, 95.9% and 100%, respectively. | en_US |
| dc.identifier.uri | http://10.90.10.223:4000/handle/123456789/20151 | |
| dc.language.iso | en | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | Breast Cancer | en_US |
| dc.subject | Artificial Neural Network | en_US |
| dc.subject | Overlapping Object Detection | en_US |
| dc.subject | Color Image Segmentation | en_US |
| dc.subject | Digital Image Processing | en_US |
| dc.subject | Fine Needle Aspiration | en_US |
| dc.title | Automatic Breast Cancer Detection from Biopsy fine Needle Aspiration Microscopic Images | en_US |
| dc.type | Thesis | en_US |