Degree Name

Doctor of Philosophy


School of Electrical, Computer and Telecommunications Engineering


This thesis proposes a computer aided diagnostic support system (CADSS) for cervical cancer classification using histology images with a spatial resolution of 3072 x 4080 pixels. Histological images are captured using a digital microscope 400 x magnifications with a resolution 0.25mm and the resolution chosen retains enough information for effective classification. Histology images of cervical biopsies captured through the digital microscope consist of three major sections; background, stroma and squamous epithelium. Most diagnostic information is contained within the squamous epithelium region and diagnosis involves inspecting the sample for cellular level abnormalities and determining the spread of the abnormalities. This is a tedious, subjective and time-consuming process with considerable variations in diagnosis between experts. Conventional analysis of a cervical histology image, such as a pap smear or a biopsy sample, is performed manually by an expert pathologist. Cancer is graded based on the spread of these abnormal cells. This thesis presents a computer aided diagnostic support system tool to help pathologists in their examination of cervical cancer biopsies. This thesis explores various components of an effective CADSS: image acquisition, pre-processing, segmentation, feature extraction, classification, grading and disease identification. The main aim of the proposed CADSS system is to identify abnormalities and quantify cancer grading in a systematic and repeatable manner. This thesis proposes three different methods and presents and compares the results using approximately 475 images of cervical biopsies which include normal, three stages of pre cancer, and malignant cases. The number of images that are used for validation is forty images.

The system is divided into a two part analysis, especially for segmentation: global (macro) and local (micro). At the global level the squamous epithelium is segmented from the background and stroma. At the local or cellular level, the nuclei and cytoplasm are segmented for feature]e extraction. Image features that influence the pathologists’ decision during the analysis and classification of a cervical biopsy are the nuclei’s shape and spread, the ratio of the areas of nuclei and cytoplasm as well as the texture and spread of the abnormalities. Similar features are extracted as part of the automated classification process. This thesis investigates various feature extraction methods including, colour, shape and texture using a Gabor wavelet as well as various quantitative metrics. Generated features are used to classify cells or regions into normal and abnormal categories. Following the classification process, the cancer is graded based on the spread of the abnormal cells. The result indicates the grading process using the five stages of cervical cancer is including normal, Cervical Intraepithelial Neoplasia (CIN)1, CIN2, CIN3 and malignant.

The application of hybrid graph cut and colour segmentation in CADSS for cervical cancer classification is a new research that uses the sliding block algorithm for the analysis of the nuclei. The block moves in a horizontal and vertical direction covering the squamous epithelium, analysing the presence of nuclei in fine detail. As a result, this method provides better performance than the K-means clustering and Gabor wavelet in terms of specificity and False Positives (FP). The results are promising, with the sensitivity for detection of normal cases being 97%, CIN1: 88%, CIN2: 94%, CIN3: 80% and malignant cases: 97%. The system tested 475 histology images of various degrees of malignancies and produced a sensitivity of 99% and specificity of 97%.