Computer-aided screening system for cervical precancerous cells based on field emission scanning electron microscopy and energy dispersive x-ray / Yessi Jusman

Yessi, Jusman (2016) Computer-aided screening system for cervical precancerous cells based on field emission scanning electron microscopy and energy dispersive x-ray / Yessi Jusman. PhD thesis, University of Malaya.

PDF (Thesis PhD)
Download (735Kb) | Preview


    Cervical cancer has caused many deaths each year. Screening tests, such as Papanicolou (Pap) smears and Liquid based Cytology (LBC) for detection of precancerous change, are able to avoid the occurrence of cervical cancer. However, these tests still have disadvantages such as being inaccurate due to human errors (the false negative rate reaching 50%) and long processing time due to manual method of determining cell abnormality based on morphological signs by the pathologists. Several researchers have developed computer systems for analysis of Pap smear and LBC images. However, the systems are still not satisfactory as they only use morphological features. Currently, Field Emission Scanning Electron Microscopy and Energy Dispersive X-Ray (FE-SEM/EDX) has great capability to extract qualitative and quantitative data from all types of materials simultaneously. Based on this capability, this study proposed FE-SEM/EDX as a data acquisition technique to develop computer-aided screening system for cervical precancerous cells. The developed system consists of data acquisition, features extraction, and classification stages. Before the data acquisition stage, sample preparations were investigated to obtain qualitative and quantitative data from FE-SEM/EDX. Hexamethyldisilazane (HMDS) and Critical-Point Drying (CPD) sample preparation techniques were compared in term of the quality of the FE-SEM image, quantification of elements distribution, and rapidity of processing time. The HMDS sample preparation technique has better results than the CPD technique for the terms. For data acquisition, FE-SEM/EDX was implemented at specific working distance (10 mm) for an average voltage range (5 to 20 kV), performed under low vacuum condition. FE-SEM image and elemental distributions (FE-SEM/EDX data) of the samples were obtained simultaneously. In the feature extraction stage, cervical precancerous features were extracted from the FE-SEM/EDX data. Proposed image processing techniques and signal analyses were applied to the FE-SEM/EDX data for feature extraction. Finally, for the classification stage, the Discriminant Analysis (DA) concept was used to classify the cervical precancerous cells based on the Bethesda system (i.e. normal, Low-grade Squamous Intraepithelial Lesion (LSIL) and High-grade Squamous Intraepithelial Lesion (HSIL)). The results indicated that eliminating the osmium HMDS technique from the sample preparation technique was favorable as it resulted in improved quality for sample preparation in term of quality of FE-SEM image and quantification of elements distribution results in data acquisition, and rapid processing time in sample preparation. In the feature extraction stage, FE-SEM/EDX images provided statistical features information based on pixel intensities, namely contrast, homogeneity, entropy and energy. Furthermore, the FE-SEM/EDX spectra can also provide information on the ratio of the peaks and the ratio of the corrected area under the peaks. The extracted features were combined into datasets for further processing using DA for classification. Our system achieved accuracy, sensitivity, and specificity results of 98.9%, 98.8%, and 99.1%, respectively. The results indicate that the chosen sample preparation method and the usage of FE-SEM/EDX for data acquisition, feature extraction, and classification techniques achieved good performances for developing a computer-aided screening system. Therefore, a computer-aided screening system for cervical precancerous cells has been successfully developed in this study.

    Item Type: Thesis (PhD)
    Additional Information: Thesis (PhD) - Faculty of Engineering, University of Malaya, 2016.
    Uncontrolled Keywords: Computer; Processing; Image data; Pattern recognition
    Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
    T Technology > T Technology (General)
    Divisions: Faculty of Engineering
    Depositing User: Mrs Nur Aqilah Paing
    Date Deposited: 28 Sep 2016 17:11
    Last Modified: 09 Oct 2019 08:48

    Actions (For repository staff only : Login required)

    View Item