System and method for computer aided polyp detection
In the present methods, the automatic detection of polyps is converted into a 2D pattern recognition problem using conformal mapping and direct volume rendering. The colon surface is first segmented and extracted from the CT data set of the abdomen, which is then mapped to a 2D plane using conformal mapping. Ray casting is used to determine sub-surface density values and the flattened image is rendered using a volume rendering technique with a translucent electronic biopsy transfer function. Polyp candidates are detected by a clustering method which identifies regions of elevated sub-surface density. The potential for false positives is reduced by analyzing the volumetric shape and texture features of the polyp candidate regions.
1. A computer-based method of detecting polyps comprising:
acquiring a 2D image dataset of a region of interest;
transforming the 2D image dataset into a plurality of voxels forming a volumetric representation of the region of interest, wherein the voxels each have a value corresponding to a density of an object represented by the voxel;
identifying a surface of the region of interest;
determining a density of the object beneath the surface of the region of interest;
identifying suspicious regions by identifying internal sub-surface regions exhibiting a density higher than that of a surrounding region; and
evaluating the suspicious regions to remove false-positive candidates by analyzing at least one of a shape or a texture of the identified internal sub-surface regions.
2. The method of claim 1 wherein the step of evaluating the suspicious region is determined in part by a shape index.
3. The method of claim 1 wherein the step of evaluating the suspicious region is determined in part by a curvedness index.
4. The method of claim 1 wherein the step of evaluating the suspicious region is determined in part by a gradient of texture.
5. A computer-based method of detecting polyps comprising:
acquiring a 2D image dataset of a region of interest;
transforming the 2D image dataset into a plurality of voxels forming a volumetric representation of the region of interest, wherein the voxels each have a value corresponding to a density of an object represented by the voxel;
extracting a surface of the object;
transforming the extracted surface into a 2D planar representation of the surface;
determining sub-surface density values of the object;
clustering the sub-surface density values of internal sub-surface regions to identify suspicious regions on the object; and
evaluating the suspicious regions to remove false-positive candidates by analyzing at least one of a shape or a texture of the internal sub-surface regions.
6. The computer based method of detecting polyps of claim 5 , wherein the object is a colon.
7. The computer based method of detecting polyps of claim 6 , wherein the step of transforming the extracted surface further comprises the use of an angle preserving conformal flattening process.
8. The computer based method of detecting polyps of claim 7 , wherein the conformal flattening process maps the colon surface to a substantially rectangular planar view.
9. The computer based method of detecting polyps of claim 8 , wherein the step of determining sub-surface density values further comprises volumetric ray casting to integrate sub-surface density values.
10. The computer based method of detecting polyps of claim 9 , wherein the step of determining sub-surface density values further comprises a step of applying a transfer function to the integrated sub-surface density values to map density values into color values indicative of object density.
11. The method of claim 5 wherein the step of evaluating the suspicious region is determined in part by a shape index.
12. The method of claim 5 wherein the step of evaluating the suspicious region is determined in part by a curvedness index.
13. The method of claim 5 wherein the step of evaluating the suspicious region is determined in part by a gradient of texture.
14. A graphical user-interface for a system for computer aided detection of polyps, the graphical user interface comprising a computer display, said display having a plurality of display windows comprising:
an endoscopic view of a colon lumen;
at least one cross-sectional view of the colon;
an electronic biopsy image of the colon;
a view of the colon identifying a presence of internal sub-surface regions of elevated sub-surface density as a location of suspicious regions; and
evaluating the suspicious regions to remove false-positive candidates by analyzing at least one of a shape or a texture of the identified internal sub-surface regions.
15. The graphical user-interface for a system for computer aided detection of polyps of claim 14 , wherein the electronic biopsy image displays the presence of regions of elevated sub-surface density in a color that is distinguishable from the surrounding region.
16. The graphical user interface for a system for computer aided detection of polyps of claim 14 , wherein the plurality of display windows are synchronized.
17. The graphical user interface for a system for computer aided detection of polyps of claim 14 , wherein the plurality of display windows further includes a flattened 2D conformal mapping of the colon.