Mathematical models integrating an ultrasound-based technology improve the diagnosis of ovarian cancer

(eng) In the European countries, age standardized incidence rates (European standard) for ovarian cancer vary between 7.2 and 19.3/100,000 while mortality rates are ranging between 2.8 and 12.2/100,000.1 In Belgium, ovarian cancer is not as frequent as breast cancer since breast cancer presents with very high incidence rates (similar for other European countries). However, age standardized breast cancer mortality rates in 2008 were less than one fourth of the age standardized incidence rates whereas for ovarian cancer, mortality rates were two thirds of the incidence rates (see figure 1). And, unlike for breast cancer, mortality rates for ovarian cancer were not decreasing over the past years.2 Indeed, ovarian cancer is one of the leading causes of death from gynaecological malignancies.3 This is explained by the fact that in general, ovarian cancer is detected at too advanced stages. Early diagnosis of ovarian cancer is thus the key for improving outcomes for the disease. Medical imaging techniques have revolutionised medicine during the last decades. Ultrasound (US) in particular gives access to vital data in a non-invasive way and is effective for imaging soft tissues of the body. Compared to other medical imaging modalities, US has the following positive attributes: • US is a real-time, easy operation medical imaging technique • US has a non-invasive and radiation free nature • US is relatively low-priced, Hence US has become widely used as a diagnostic technique in general clinical practice. In gynaecology, US is one of the most important and primary diagnostic tools. Its use continues to increase and it is now an essential part of the diagnostic procedure in examining the female pelvis. Indeed, the use of US morphology to characterize adnexal masses and thus diagnose ovarian cancer is well established.4 As a part of patient management, gynaecologists use US morphology to differentiate between malignant and non-malignant ovarian masses that come to their attention. In addition, a large number of indexes and mathematical prediction models that assess the likelihood of malignancy for an ovarian mass have been developed and they all incorporate US. A novel computer-aided technology, HistoScanningTM, makes use of US data for characterising ovarian tissue suspicious of being malignant. In part I of this research work, we investigated whether ovarian HistoScanning could improve the performance of existing prediction models for the differential diagnosis of ovarian masses and we also explored what place could be granted to this new technology in clinical practise. Part I is organised as follows: Chapter 1 concerns the epidemiology of ovarian cancer and discusses the importance of differential diagnosis of ovarian masses. Detailed aims are described in chapter 2. Chapter 3 introduces the general methods used for this work. In chapter 4, two publications that present the results of this work are presented. Finally, a discussion, regarding the results presented in part I, concludes this work in chapter 5.