Search:
Computing and Library Services - delivering an inspiring information environment

Investigation of a Computer-Aided Detection Solution for Breast Focal Asymmetry

Zhang, Zhiyong (2011) Investigation of a Computer-Aided Detection Solution for Breast Focal Asymmetry. Doctoral thesis, University of Huddersfield.

[img] PDF - Accepted Version
Restricted to Repository staff only

Download (2691kB)

    Abstract

    Breast focal asymmetry is an important sign of early breast cancer and might be a developing masses or other underlying breast cancer. However, breast focal asymmetry is very difficult to recognise, even for the specialists of radiologists because focal asymmetries lack the boundary and consistent shape of masses and are similar to normal fibroglandular tissues. The computer-aided detection (CADe) for focal asymmetry is a novel research topic in the field of computer-aided detection of breast cancer. The detection of focal asymmetry has the significant value to save women’s live in the early stage of breast cancer.

    At present, there is only one existing dedicated detection solution for focal asymmetry, which was proposed in 2008. The aim of this research is to develop a novel dedicated computer-aided detection solution for focal asymmetry and overcome the weakness of the existing approach. The objectives of this research are to propose a novel solution for detecting focal asymmetry, a novel breast segmentation approach, a novel pectoral muscle detection approach and a dedicated focal asymmetry detection approach. The literature review of previous researches for focal asymmetry, asymmetric breast findings, breast segmentation and pectoral muscle detection have been extensively carried out. Based on the literature review, a range of hypotheses have been proposed for breast segmentation, pectoral muscle detection and focal asymmetry detection. Furthermore, a novel framework and three novel approaches have been proposed and developed in this research.

    Firstly, the automatic breast segmentation (ABS) approach was proposed to segment the skin-line of breast from mammographic images and preserve the nipple at the same time, a task which very few breast segmentation methods claimed. The experimental results showed that the ABS approach can adequately segment the breast skin-line and preserve the nipple if it is in profile. The proposed ABS approach is one of three existing breast segmentation approaches that can segment breast skin-line and preserve the nipple at the same time. Secondly, a novel approach for detecting pectoral muscle has been proposed to detect and remove pectoral muscle from the segmented breast areas. The experimental results showed that the proposed maximum intensity change (MIC) approach can detect pectoral muscle with high quality. Thirdly, the novel approach for detecting focal asymmetry has been proposed in this research and various features have been extracted to classify a suspicious region into the category of focal asymmetry or non-focal asymmetry. The experimental results indicated that the proposed detection approach can detect focal asymmetry with the 81.8% sensitivity and the 0.333 false positive regions per image.

    This research has proposed a novel framework and three novel approaches to build the computer-aided detection solution for focal asymmetry. This research overcomes the weakness of current approach for detecting focal asymmetry. The proposed detection approach is the second dedicated solution for detecting breast focal asymmetry in existence.

    Item Type: Thesis (Doctoral)
    Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
    T Technology > T Technology (General)
    Schools: School of Computing and Engineering
    Depositing User: Carol Doyle
    Date Deposited: 16 Aug 2011 16:32
    Last Modified: 21 Nov 2011 12:50
    URI: http://eprints.hud.ac.uk/id/eprint/11302

    Document Downloads

    Downloader Countries

    More statistics for this item...

    Item control for Repository Staff only:

    View Item

    University of Huddersfield, Queensgate, Huddersfield, HD1 3DH Copyright and Disclaimer All rights reserved ©