Latent to latent flow boosts efficiency in 3d medical image segmentation
Latent-to-Latent Flow for Volumetric Stochastic Segmentation
Computer Vision and Pattern RecognitionArtificial IntelligenceMachine Learning
Summary
Medical images often have differences in how experts mark important areas, and this uncertainty affects treatment decisions. The authors present a new technique that works with simpler, encoded versions of these images to predict different possible segmentations efficiently. Their method speeds up the processing by up to fourteen times without losing important accuracy, helping with tasks like planning radiation therapy and identifying organs. This approach makes handling complex 3D medical images more practical.
What this means in practice
- •For medical imaging teams: Provide faster and efficient 3D segmentation variants reflecting uncertainty to support radiotherapy planning.
- •For healthcare software developers: Integrate scalable segmentation models that estimate multiple organ structures with reduced computational cost for clinical tools.
Authors
Omar Todd, Sooha Kim, Raghav Mehta, Katherine Mackay, David Bernstein, Alexandra Taylor, Fabio De Sousa Ribeiro, Ben Glocker
Abstract
Uncertainty arising from inter-observer variability in medical image segmentation plays an important role in developing treatment plans. Research in this area is inhibited by the lack of multiple annotations for large-scale medical datasets, especially for volumetric data, which suffers from additional scaling and computational complexity challenges. Flow matching has emerged as a powerful framework for generative modelling and has also been demonstrated to maintain strong performance when working with latent representations of images. In this work, we introduce a latent-to-latent flow technique for stochastic segmentation of medical volumes via encoded representations of both the image and label space. We evaluate our method on two challenging applications covering delineation uncertainty for radiotherapy planning and multiple organ structure segmentation, improving efficiency up to 14x compared with full resolution models while maintaining clinically relevant performance.