A Novel Approach to Endometriosis Detection: Machine Learning for a Non-invasive Diagnosis

This project aims to turn our AI software driven algorithmic approach to diagnosing endometriosis into a minimal viable product (MVP).

Endometriosis is a common and painful condition affecting 1 in 10 women, and is very difficult to diagnose taking on average 8-years from a patient first presenting to a GP with symptoms.

In the UK 72.5% of patients report being misdiagnosed with other physical or mental health conditions, with 53% of misdiagnoses by gynaecologists. The current gold standard diagnostic test for endometriosis is an invasive exploratory surgical procedure. Our method will non-invasively detect endometriosis from ultrasound scan data in weeks rather than years.

Building on our initial proof of concept algorithm, this project will obtain appropriate data sets to train and test algorithms for detecting endometriosis, providing a digital alternative to surgery.

When this product has met regulatory approval it will dramatically reduce time to diagnosis, improve accessibility to high quality image based diagnostics across the country and reduce the cost of diagnosis to the health service.