Ultrasound RF Data for Discriminating Between Benign and Malignant Ovarian Masses With Solid Ultrasound Morphology, a Feasibility Study

Ultrasound imaging provides useful information for the characterization of ovarian masses as benign or malignant. The most accurate mathematical model to categorize ovarian masses is the IOTA ADNEX model.This model estimates the risk of malignancy and performs similarly to subjective assessment by an experienced ultrasound examiner for discriminating between benign and malignant adnexal masses. The ability of IOTA ADNEX to discriminate between benign and malignant masses is very good (area under the receiver operator characteristic curve 0.937 (95% CI: 0.915-0.954). The ADNEX model maintains its accuracy even in the hands of operators with different experience and training.

According to IOTA terminology, 13% of ovarian masses detected on ultrasound examination are classified as solid. Solid ovarian masses have a risk of malignancy of 60%-75%2 and the discrimination between benign and malignant in this morphological category is challenging. Additionally, it has been estimated that 30% (25/84; 95% CI 18 to 44%) of solid malignant ovarian masses are metastases from non-ovarian tumors. The discrimination between primary ovarian cancer and metastatic tumors in the ovary is also clinically important for planning adequate therapeutic procedures. It is worth exploring the predictive performance of the diagnostic tools in identifying ovarian masses with ultrasound solid morphology.

Preliminary data (unpublished) on radiomics analysis and ovarian masses provided that benign and malignant ovarian masses with solid morphology have different radiomics features in a monocentric retrospective study. However, no statistically significant differences have been observed between primary ovarian cancer and metastases to the ovary.

A new technology is emerging in engineering ultrasound field: the analysis of ultrasound summed RF data- raw data generated by the interface of ultrasound beams with human tissues. To date, raw data are not utilized for conventional imaging and their eventual role in clinical practice is unknown. Indeed, summed RF data could better correlate with biological parameters then parameters identifiable in B-mode images. Summed RF data could also improve radiomic analysis.