Skip to content

The application of Machine Learning within Gynaecology Diagnostics

Ovarian masses are very common, affecting 14% of women. Although the incidence of ovarian cancer is low, it remains the leading cause of death from a gynaecological malignancy in the UK (Cancer Research UK). Ultrasound is recognised as first line imaging for ovarian masses globally. The differentiation  between a benign and a malignant mass can be challenging. Diagnostic uncertainty in ovarian mass classification can drive the need for additional investigations such as MRI and surgical intervention, due to potential concern about malignant potential.

Our current project, in collaboration with Professor Dirk Timmerman and his team at Leuven, is focusing on applying ML clinically to improve the level of diagnostic confidence and provide practitioners with an artificial second reader that encodes representations from thousands of examinations, i.e. often more than a radiologist sees during their career. Furthermore, a ML model, capable of extracting specific features from within an ultrasound image could support the training of basic/intermediate ultrasound scanners.

Led by Mr Srdjan Saso.

Our research