Authors: Marvin Tom Teichmann, Manasi Datar, Lisa Kratzke, Fernando Vega, Florin C. Ghesu

Published on: September 27, 2024

Impact Score: 7.8

Arxiv code: Arxiv:2409.18628

Summary

• What is new: Integration of epistemic uncertainty estimation within OAR contouring workflow for OOD detection.
• Why this is important: Ensuring reliability of deep learning models for contouring in radiotherapy, especially with OOD scenarios.
• What the research proposes: Advanced statistical method for OOD detection integrated into clinical workflows.
• Results: Achieved AUC-ROC of 0.95, with 0.95 specificity and 0.92 sensitivity for OOD detection, indicating high reliability.

Technical Details

Technological frameworks used: Deep learning and statistical methods for uncertainty estimation.

Models used: Epistemic uncertainty estimation models.

Data used: Clinical data specifically compiled for OAR contouring with OOD scenarios.

Potential Impact

Radiotherapy equipment manufacturers, radiotherapy clinics, especially those using Varian (a Siemens Healthineers company) products.

Want to implement this idea in a business?

We have generated a startup concept here: UncerGuard.