Publication of the week: S. Khazendar, H. Al-Assam, H. Du, S. Jassim et al.

Khazendar, S., A. Sayasneh, H. Al-Assam, H. Du, J. Kaijser, L. Ferrara, D. Timmerman, S. Jassim & T. Bourne, “Automated characterisation of ultrasound images of ovarian tumours: the diagnostic accuracy of a support vector machine and image processing with a local binary pattern operator”, Facts, Views and Visions in ObGyn 7.1 (March 2015), 7-15

Ovarian cancer is the fourth most serious cancer in women. Preoperative characterisation of ovarian masses into benign or malignant is of paramount importance to optimise patient care and management in preventing missed opportunities of early detection and unnecessary biopsy operations. In this study, we developed and validated a computerised model to characterise ovarian masses as benign or malignant based on transvaginal 2D static ultrasound images of ovary. Using image processing techniques to enhance and de-noise the images first followed by extraction of local binary pattern (LBP) textures from the images, we train a Support Vector Machine (SVM) with the extracted features from a training set. We tested our classification method with images of 187 ovarian masses, and the test results show an encouraging level of average accuracy of 0.77 (95% CI: 0.75-0.79), demonstrating the promise of SVM with LBP texture features.

This research is a collaborative project with the Department of Cancer and Surgery, Queen Charlotte’s and Chelsea Hospital, Imperial College London (UK) and the Department of Development and Regeneration: Obstetrics and Gynaecology, University Hospital KU Leuven, Leuven (Belgium).

Read the full article on the FVVO website.

Four of the authors are members of the Department of Applied Computing at Buckingham: Shan Khazendar is a research student, Hisham Al-Assam is Lecturer, Hongbo Du is Senior Lecturer and Sabah Jassim is Professor of Mathematics and Computation.