The project aims at enhancing performance metrics and prospectively validating a radiogenomics model based on ovarian US images for predicting germline breast cancer susceptibility gene 1 and/or 2 (BRCA) status in women with healthy ovaries. The project is divided in two operational phases:

Model development phase (ambispective)

AIM 1: To define and implement a proper and fine-tuned image preprocessing pipeline on the existing dataset; AIM 2: To enlarge dataset size with new real images from different centers and apply data augmentation techniques, deep neural network models combined with the aforementioned handcrafted imaging features from radiomics analysis;

Implementation and validation phase (prospective)

AIM 3: To further cross-validate the predictive model on US images acquired prospectively in an observational multicenter study.

Inclusion Criteria:

Availability of gBRCA1/2 test results

Transvaginal pelvic US performed providing at least one picture of one healthy ovary

US images stored in .dicom format.

Exclusion Criteria:

Personal diagnosis of OC

Ovarian abnormal findings (eg. ovarian endometriomas, dermoid cyst, ovarian cystoadenofibroma, ovarian borderline tumour…) with the exception of functional cysts at pelvic US

gBRCA1/2 testing results nor provided by an ISO 1589 accredited laboratory

Refusal to provide written informed consent

We will conduct a retrospective multicenter observational study aimed at collecting US images of healthy ovaries with known BRCA status.