Snellere diagnostiek voor endometriose en adenomyose door AI ondersteunde echografie

AI research is growing in gynaecological imagery but rarely implemented in practice. To this day, no large Dutch database of transvaginal ultrasound collected for endometriosis and adenomyosis exists, which greatly limits AI research. To build such AI-ready, large database, our Dutch consortium establishes a data collection protocol and explores various Deep Learning models with this data. We also co-design a point-of-care AI tool to assist experts in diagnosing these conditions using ultrasound scans.

Goal

We aim to build a point-of-care AI tool to assist experts in diagnosing these conditions using ultrasound scans. 

Approach

This project is a national, multi-centre, prospective cohort and interdisciplinary study. It will include the following phases:

• Ethical Approval and Preparation
• High-Quality Data Specifications
• Data Collection and Annotation
• AI Model Development and Testing
• Co-Design of Point-of-Care Tool
• Pre-Clinical Study
• Implementation and Reporting

Overall the project will combine expertise in endometriosis and adenomyosis, ultrasonography, Responsible Artificial Intelligence, co-design, data science, clinical research and gynaecological practice.

Collaboration partners

AI4EA is supported by the first national Dutch consortium dedicated to AI research in endometriosis and adenomyosis. The consortium consists of 10 Dutch Hospitals with key expertise in the field, as well as the national patient organisation Endometriosis Stichting, and is coordinated by Amsterdam UMC. 

This interdisciplinary and intersectoral collaboration is key to ensure standardisation in data collection across centres, as well as the creation of a large database suitable for AI research purposes. It also facilitates the inclusion of a more diverse population in our study. Through a participative approach, we therefore aim to better account for the needs of various key stakeholders in the Dutch healthcare system.

(Expected) results

With AI4EA, we expect to publish the first open-source, large, high-quality, and AI-ready dataset of ultrasound scans specifically collected for endometriosis and adenomyosis investigation. 

We also expect to co-design a point-of-care AI tool to support endometriosis and adenomyosis diagnosis, in close collaboration with our consortium partners and patients.

Lastly, we aim to train one of the first PhD students specialised in AI research and endometriosis, adenomyosis and ultrasonography.

Overall, we will publish peer reviewed articles and disseminate our results further in international conferences, as well as collaborate with patient organisations in the Netherlands and beyond to share our work with patients and the wider public.