Virtual colonoscopy via wavelets
Inventors
Summers, Ronald M. • Li, Jiang • Greenblum, Sharon
Assignees
US Department of Health and Human Services
Publication Number
US-8023710-B2
Publication Date
2011-09-20
Expiration Date
2027-03-12
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Various techniques can be used to improve classification of colon polyps candidates found via computed tomographic colonography computer aided detection (CTCCAD). A polyp candidate can be classified as a true positive or a false positive. For example, a two-dimensional projection image of the polyp can be generated from a three-dimensional representation and classified based on features of the projection image. An optimal viewpoint for the projection image can be found via techniques such as maximizing viewpoint entropy. Wavelet processing can be used to extract features from the two-dimensional projection image. Feature extraction can use a piecewise linear orthonormal floating search for locating most predictive neighbors for wavelet coefficients, and support vector machines can be employed for classification. The techniques can be useful for improving accuracy of CTCCAD techniques.
Core Innovation
The invention provides techniques for improving classification accuracy of colon polyp candidates identified via computed tomographic colonography computer aided detection (CTCCAD). After a polyp candidate is identified in a three-dimensional digital representation of the colon, a two-dimensional projection image of the polyp candidate can be generated. This two-dimensional image is processed to determine whether the polyp candidate is a true polyp or a false positive using wavelet-based technologies and classification methods.
The approach involves selecting an optimal viewpoint for generating the two-dimensional projection image using a maximum viewpoint entropy technique. The selected viewpoint is typically a point along the virtual colon centerline near the polyp candidate, and background information is excluded to focus processing on the candidate itself. Wavelet processing extracts features from the two-dimensional projection image, where a piecewise linear orthonormal floating search is used to find the most predictive neighboring wavelet coefficients. A committee of support vector machines (SVMs) then classify the polyp candidate using these features.
The problem being addressed arises from virtual colonoscopy's tendency to generate false positives in polyp detection. False positives increase the workload of human classifiers and reduce the clinical usefulness of CTCCAD systems. The invention aims to increase specificity by effectively filtering false positives through improved image feature extraction and classification techniques, thereby enhancing virtual colonoscopy technologies.
Claims Coverage
The patent contains multiple independent claims encompassing methods and apparatuses for improving polyp candidate classification in virtual colonoscopy using wavelet-based image processing and viewpoint entropy techniques. The main inventive features involve digital representation processing, viewpoint selection, feature extraction, and classification.
Optimized virtual camera viewpoint selection via maximum viewpoint entropy
Choosing a virtual camera viewpoint for generating a two-dimensional projection image of a polyp candidate as a point along a centerline in a virtual colon, where the viewpoint is selected by maximizing viewpoint entropy to capture the most informative view of the candidate.
Wavelet-based feature extraction from two-dimensional projection images
Extracting wavelet features including coefficients in vertical, horizontal, diagonal directions, energy, entropy, and predictive errors based on most predictive neighbor wavelet coefficients from the two-dimensional digital projection image of the polyp candidate.
Application of a committee of support vector machines for classification
Determining whether a polyp candidate is a true polyp or a false positive by applying the extracted wavelet features to a committee classifier composed of multiple SVM members, each trained with different wavelet feature subsets, and combining their outputs to improve classification accuracy.
Computing viewpoint entropy directly from mesh faces to exclude background
Calculating viewpoint entropy directly from faces of a mesh representing the polyp candidate, ignoring background information, to efficiently and accurately select the optimal viewpoint for two-dimensional image generation.
Post-processing of polyp candidates identified by computer-aided detection
Filtering initial polyp candidate lists generated by CTCCAD systems by applying the improved wavelet feature extraction and classification methodologies to reduce false positives before presenting candidates for human evaluation.
The claims collectively define a method and apparatus for enhancing virtual colonoscopy polyp candidate classification by selecting a viewpoint using maximum viewpoint entropy, generating a two-dimensional image, extracting wavelet-based features including predictive neighbor coefficients, and classifying candidates via an SVM committee to effectively reduce false positives and improve diagnostic accuracy.
Stated Advantages
Improved specificity in virtual colonoscopy by filtering out false positives, reducing the human classifier workload.
Enhanced accuracy of polyp candidate classification by utilizing wavelet-based texture features and committee SVM classification.
Automated, fully automatic filtering technique requiring no user intervention.
Ability to exclude approximately 69% of false positives for medium-sized polyps (6-9 mm) while maintaining high sensitivity.
Documented Applications
Processing polyp candidates identified by CT colonography computer-aided detection systems for virtual colonoscopy.
Filtering false positives from initial polyp detections to aid radiologists in reviewing CT colonography results.
Applying wavelet feature extraction to two-dimensional projection images of anatomical structures, particularly colon polyps.
Use in medical imaging analysis workflows to improve early detection and treatment of premalignant and malignant growths in the colon.
Interested in licensing this patent?