PlacentaVision was built using cross-modal contrastive learning, aligning visual data (placental images) with textual data (pathological reports). Researchers collected and analyzed a 12-year dataset, simulating various imaging conditions to ensure the model’s robustness. This tool’s versatility was further validated across different populations.
The next steps involve developing a mobile app for healthcare professionals. This app will:
- Require minimal training.
- Provide immediate diagnostic feedback.
- Enhance clinical decision-making, especially in under-resourced areas.
The team aims to expand the tool’s capabilities, integrating additional clinical data and placental features to improve predictions. Testing in diverse settings will ensure reliable performance.
MEDICA-tradefair.com; Source: Northwestern University