Doppler Flow Imaging in support of ovarian tumor diagnosis: automated denoising and evaluation of color score in accordance with IOTA guidelines

Although ovarian cancer ranks quite low for what concerns tumor incidence among women, it ranks fifth in cancer deaths among women and it is accountable for the largest number of deaths due to tumors of the reproductive system. These unfavorable statistics are due to the challenging nature of early diagnosis, as this type of cancer is often diagnosed in the more advanced stages of malignancy. It has been demonstrated that early recognition is crucial for the prognosis, having the potential of reducing mortality rates by approximately 40%. Transvaginal ultrasound, coupled with Doppler Flow Imaging, is the method of choice for assessing ovarian adnexal masses, through subjective assessment. However, not all ultrasound operators have a sufficient level of experience in adnexal pathology in order to correctly use subjective assessment, which is why agreement is quite poor. For this reason, a group of researchers came together as the International Ovarian Tumor Analysis (IOTA) group and they have been working to develop widely used, standardized mathematical models and score systems, that use combinations of parameters to make the assessment of adnexal mass malignancy as objective as possible. Color score is a measure of the amount of blood flow in an adnexal mass and it ranges from 1 (no blood flow) to 4 (highly vascular mass). It has been asserted as a valuable predictor of malignancy, as it is used as a variable for several predictive models developed by IOTA but, as of today, it is only deductible through subjective assessment of Doppler Flow images of the ovary. Nonetheless, when it comes to assigning color scores, agreement between clinicians with different levels of adnexal pathology knowledge is not satisfactory. Therefore, the scope of this study is to develop an automatic tool for the evaluation of color score in ovarian adnexal masses. In order to automatically assess Doppler Flow Imaging, considering the moltitude of doppler artifacts, developing a tool for automated denoising of Doppler Flow images is crucial, thus it is another scope of this study.