Image To Tree with Recursive Prompting
This addresses a key step in automated medical image analysis for anatomical structures like coronary arteries, though it appears incremental as it builds on existing architectures like UNet and Transformer.
The paper tackles the problem of extracting tree-structured geometries from 2D medical images, where overlapping branches in projections pose challenges, by reformulating it as an optimization over recursive steps and achieves compelling results on synthetic datasets, outperforming a shortest-path baseline.
Extracting complex structures from grid-based data is a common key step in automated medical image analysis. The conventional solution to recovering tree-structured geometries typically involves computing the minimal cost path through intermediate representations derived from segmentation masks. However, this methodology has significant limitations in the context of projective imaging of tree-structured 3D anatomical data such as coronary arteries, since there are often overlapping branches in the 2D projection. In this work, we propose a novel approach to predicting tree connectivity structure which reformulates the task as an optimization problem over individual steps of a recursive process. We design and train a two-stage model which leverages the UNet and Transformer architectures and introduces an image-based prompting technique. Our proposed method achieves compelling results on a pair of synthetic datasets, and outperforms a shortest-path baseline.