Active spline model: A shape based model-interactive segmentation
This addresses the need for better editability in medical image segmentation, though it is incremental as it builds on existing shape models.
The paper tackles the problem of post-segmentation editing by proposing an interactive spline-based model that allows users to adjust unsatisfactory segmentation results, achieving an average overlap improvement from 0.879 to 0.945 in lung segmentation on chest radiographs.
Rarely in literature a method of segmentation cares for the edit after the algorithm delivers. They provide no solution when segmentation goes wrong. We propose to formulate point distribution model in terms of centripetal-parameterized Catmull-Rom spline. Such fusion brings interactivity to model-based segmentation, so that edit is better handled. When the delivered segment is unsatisfactory, user simply shifts points to vary the curve. We ran the method on three disparate imaging modalities and achieved an average overlap of 0.879 for automated lung segmentation on chest radiographs. The edit afterward improved the average overlap to 0.945, with a minimum of 0.925. The source code and the demo video are available at http://wp.me/p3vCKy-2S