Mohd Nizom, Ahmad Nabil (2018) Image Processing Of Digital Mammograms For Breast Cancer Detection And Classification. Project Report. Universiti Sains Malaysia, Pusat Pengajian Kejuruteraan Aeroangkasa. (Submitted)
|
PDF
Download (464kB) | Preview |
Abstract
Due to an increase in number of breast cancer screening worldwide, development of accurate CAD is needed for tumour detection in mammograms. This study aims to develop an image processing algorithm that can produce lesser errors than human operators. The algorithms to be developed will consist of pre-processing, enhancement and image segmentation. This study also aims to develop an algorithm that uses conversion of greyscale image into RGB as an approach to image processing for greyscale image. For the image processing, the pre-processing is done by removal of artefacts and pectorals muscle using image segmentation and selection by region area and region ID respectively. Then, the process begins with the image enhancement using CLAHE to improve the details and contrast in the image. After that, the greyscale image undergo conversion into RGB by changing the colourmap. The image is segmented based on colour then translated into a circle which centroid is same with the cluster and the number of pixel is same to the tumour detected for comparison with the ground truth data. The accuracy of the algorithm developed in detecting tumour is 94.38% showing that it is relevant for use by radiologists. The algorithm may be developed for application in other field that uses greyscale image as well.
| Item Type: | Monograph (Project Report) |
|---|---|
| Subjects: | T Technology T Technology > TL Motor vehicles. Aeronautics. Astronautics |
| Divisions: | Kampus Kejuruteraan (Engineering Campus) > Pusat Pengajian Kejuruteraan Aeroangkasa (School of Aerospace Engineering) > Monograph |
| Depositing User: | Mr Engku Shahidil Engku Ab Rahman |
| Date Deposited: | 09 Jun 2022 06:07 |
| Last Modified: | 09 Jun 2022 06:07 |
| URI: | http://eprints.usm.my/id/eprint/52593 |
Actions (login required)
 |
View Item |
