Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Image texture analysis of transvaginal ultrasound in the diagnosis of ovarian lesions

Aldahlawi, Rana 2016. Image texture analysis of transvaginal ultrasound in the diagnosis of ovarian lesions. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of 2017ALdahlawiRHPhD.pdf]
Preview
PDF - Accepted Post-Print Version
Download (4MB) | Preview
[thumbnail of AldahlawiRH.pdf] PDF - Supplemental Material
Restricted to Repository staff only

Download (171kB)

Abstract

Ovarian cancer has the highest mortality rate of all gynaecological cancers and is the fifth most common cancer to occur in women in the UK. Amongst various imaging modalities, ultrasound is considered the main modality for ovarian cancer triage. As with other imaging modalities, the main issue is that the interpretation of ultrasound images is subjective and observer dependent. Texture analysis has been shown to have potential in the objective assessment of ovarian cancer in a preliminary study. Another form of texture analysis is Acoustic Structure Quantification (ASQ), which has been documented to have a number of successful uses in liver diseases. However, it has not been applied to ovarian lesions. Therefore, the aim of this study was to assess prospectively the diagnostic performance of texture analysis methods such as GLCM, Wavelet, and ASQ in discriminating between benign and malignant adnexal masses and between different types of benign masses and compare it to widely used scoring models. Prior to applying ASQ to ovarian images, its reliability and repeatability were first evaluated. This includes random variation caused by the ultrasound system and the operator during image acquisition. A tissue-equivalent phantom was used in these tests. It was found that the ASQ feature demonstrated excellent repeatability for ASQ software, with all transducers showing less than 0.4% variance from the mean: thus, ASQ software is able to produce reliable ASQ output measures. When testing the factors that may influence the performance of the ASQ analysis, the results revealed that three factors do not influence the mean of the output curve: the ROI size, depth and gain setting. However, focal position has a significant effect on the mean of the output curve. Transducer frequency does not affect the output curve except when using high frequencies such as 8 MHz. Other tests were done to determine the appropriate parameters in the software to be used on images of ovarian masses. Firstly, ASQ was applied to 45 pelvic masses. The preliminary results showed no significant difference between benign and malignant masses using the ASQ technique: therefore, the study was terminated due to failure to discriminate the benign from the malignant masses using ASQ. iv Secondly, two types of textural features were investigated in this study: grey-level co-occurrence matrix (GLCM) and wavelet, as recommended by a preliminary study. A sample of 169 masses was collected from participants, of which 140 were benign and 29 were malignant by histology. In addition to texture features, other widely used scoring models were applied on the same images for comparison, namely RMI, PMI and ADNEX. The results revealed excellent discriminatory ability in both GLCM and wavelet between malignant and cystic masses and between benign and cystic masses, with AUC of .994 and .895 for GLCM and .894 and .814 for wavelet respectively, as well as between normal and malignant tissue, with p >.05 and p=.004 in both GLCM and wavelet respectively. Results also showed that GLCM outperformed RMI and ADNEX in distinguishing between benign and malignant masses, even when dividing the study population into pre- and postmenopausal groups. In addition, GLCM has the advantage of being objective and not operator dependent. Receiver operating characteristic (ROC) curve analysis was carried out to determine the discriminatory ability of textural features, which was found to be satisfactory. The principal conclusion was that GLCM and wavelet features can potentially be used as computer aided diagnosis (CAD) tools to help clinicians in the diagnosis of ovarian cancer.

Item Type: Thesis (PhD)
Status: Unpublished
Schools: Engineering
Subjects: R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer)
T Technology > TA Engineering (General). Civil engineering (General)
Uncontrolled Keywords: ovarian cancer; pelvic masses; texture analysis; Grey level co-occurrence matrix (GLCM); wavelet; ultrasound of pelvis.
Date of First Compliant Deposit: 1 March 2017
Last Modified: 04 Jun 2017 09:43
URI: https://orca.cardiff.ac.uk/id/eprint/98631

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics