Segmenting colon wall via level set techniques
Inventors
Van Uitert, JR., Robert L. • Summers, Ronald M. • Bitter, Ingmar
Assignees
US Department of Health and Human Services
Publication Number
US-8175348-B2
Publication Date
2012-05-08
Expiration Date
2027-06-05
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Various level set techniques can be used to automatically segment the colon wall, including identifying the colon wall outer boundary. A speed image can be used during level set processing. For example, the speed image can be generated via inverting the gradient perpendicular to the segmented inner boundary of the colon wall. The techniques can be useful for determining wall thickness, which can be used to classify polyp candidates, diagnose diseases of the colon, and the like.
Core Innovation
The invention presents various level set techniques to automatically segment the colon wall, with a particular focus on identifying the colon wall outer boundary from digital representations such as CT scans. A key component involves generating a speed image via inverting the gradient perpendicular to the segmented inner boundary (lumen) of the colon wall. This speed image is then used in level set processing to guide the evolution of the segmentation surface, enabling accurate delineation of the outer colon wall boundary even with low contrast between the colon wall and surrounding fat tissue.
The problem addressed arises from the difficulty in detecting the colon wall outer boundary using CT scans, due to low contrast between the colon wall and adjacent tissues. While prior efforts commonly focused on segmenting the inner boundary, accurate segmentation of the outer boundary is necessary for determining wall thickness, which in turn is valuable for classifying polyp candidates and diagnosing colon diseases like diverticular disease. The invention solves this by employing level set methods combined with a speed image derived from the inner wall segmentation, enabling fully automatic and subvoxel-accurate segmentation of the colon wall outer boundary.
The segmented outer boundary combined with the inner boundary facilitates determination of colon wall thickness. The thickness measurements can be applied to multiple diagnostic and analytic scenarios, such as polyp candidate identification and classification, colon disease diagnosis including diverticular disease, colon centerline determination, spasm detection, cancer detection, and virtual colonoscopy flythroughs. The technique uses digital representations such as two- or three-dimensional CT images and can provide graphical output for further automated or human evaluation.
Claims Coverage
The patent claims cover several inventive features related to automatic identification and segmentation of the colon wall outer boundary using level set techniques augmented by speed images, as well as further processing based on the segmented boundaries.
Identifying colon wall outer boundary via level set technique using a speed image
A method of receiving a digital representation of a colon and identifying the colon wall outer boundary by generating a speed image and evolving an isosurface based on the speed image in a level set technique, with the lumen segmentation used as an initial boundary.
Generating speed image from directional derivative perpendicular to inner boundary
Generating the speed image by calculating a directional derivative of the digital representation in a direction perpendicular to the colon wall inner boundary, applying sigmoid filtering and inversion to suppress noise and emphasize boundary gradients.
Segmenting colon wall and calculating wall thickness from inner and outer boundaries
Segmenting both inner and outer colon wall boundaries represented as surfaces and calculating colon wall thickness based on the distance between these surfaces.
Using colon wall thickness in polyp candidate classification
Submitting a set of characteristics including colon wall thickness to a polyp candidate classifier to distinguish true positive polyps.
Detecting colonic diseases via colon wall thickness
Using the colon wall thickness determined from segmentations to detect diseases such as colonic diverticular disease, colon spasm, and colon cancer.
Subvoxel accurate segmentation via three-dimensional geodesic active contour level set segmentation
Applying three-dimensional geodesic active contour level set segmentation with curvature terms and advection attracting the level set to high gradient values for precise outer wall detection.
Clustering and classifying diverticular disease candidate detections
Clustering candidate detections based on proximity and computing features such as average thickness and intensity to input to a support-vector machine classifier for detecting diverticular disease.
The claims broadly cover methods and computer-executable media for colon wall segmentation using level set techniques enhanced by speed images derived from inner wall segmentation, subvoxel accurate identification of outer boundaries, calculating wall thickness, and applying these results for classification of polyps and detection of colon diseases.
Stated Advantages
The technique enables fully automatic segmentation of the colon wall outer boundary requiring no user intervention.
Subvoxel accuracy in boundary detection allows precise measurement of colon wall thickness.
Improves detection and classification of polyp candidates, reducing false positives compared to curvature-based methods.
Enables diagnosis of colon diseases such as diverticular disease by accurate thickness computation.
Overcomes difficulty caused by low contrast between colon wall and surrounding tissue in CT images.
Supports various applications including colon spasm detection, colon centerline determination, and virtual colonoscopy flythroughs.
Documented Applications
Determining colon wall thickness for use in polyp candidate identification and classification.
Diagnosing colonic diseases such as diverticular disease using colon wall thickness.
Detecting colon spasm, colon cancer, and presence of polyps via colon wall thickness.
Automatic segmentation of the colon wall outer boundary to improve virtual colonoscopy analysis.
Computing colon centerline and facilitating virtual colonoscopy flythroughs in difficult-to-segment areas.
Using thickness features in support vector machine classifiers for reducing false positives in diverticular disease detection.
Interested in licensing this patent?