Lutz Guendel

Automated Polyp Measurement with CT Colonography: Preliminary Observations in a Phantom Colon Model

OBJECTIVE The purpose of this study was to evaluate the accuracy and precision of polyp measurements obtained with an automated tool in a colon phantom containing polyps of multiple sizes, morphologic types, and locations. MATERIALS AND METHODS A colon phantom was scanned at 12, 25, 50, and 100 mA with standard CT colonographic acquisition parameters. Four reviewers using manual 2D methods and an automated polyp measurement tool measured 24 polyps of varying sizes and morphologic types, some at a haustral fold tip and some not at a fold tip. The accuracy (difference from true value) of manual and automated methods was compared across polyp sizes, morphologic types, locations, and doses. Precision (closeness of different measures) was compared for intraobserver and interobserver measurements. RESULTS The accuracy of automated polyp measurement was dependent on morphologic type (p < or = 0.02), size (for three of four reviewers, p < or = 0.05), and location of polyps with respect to haustral folds (two of four reviewers, p < or = 0.01). For two of four reviewers, automated measures were less accurate for 5-mm polyps, flat polyps, and polyps at the tips of folds (p < or = 0.04). Intraobserver precision was high, two automated measurements being within 0.1 mm of each other 82-93% of the time. Interobserver precision values for automated measures were more similar 85% of the time (82/96; p < 0.001). CONCLUSION Accuracy of automated polyp measurements depends on polyp size, morphologic type, and location. When using an automated tool, radiologists should visually inspect automated polyp measurements, particularly for small and flat polyps and those located on folds, because manual measurements may be more accurate in this setting. Automated polyp measurements are more precise than manual measurements.

Comparative Performance of Two Polyp Detection Systems on CT Colonography

OBJECTIVE The purpose of our study was to evaluate two current automatic polyp detection systems to determine their sensitivity and false-positive rate in patients who have undergone CT colonography and subsequent endoscopy. MATERIALS AND METHODS We evaluated two polyp detection systems--Polyp Enhanced Viewing (PEV) and the Summers computer-aided detection (CAD) system (National Institutes of Health [NIH]) using a unique cohort of CT colonography examinations: 31 examinations with true-positive lesions identified by radiologists and 34 examinations with false-positive lesions incorrectly identified by radiologists. All patients had reference-standard colonoscopy within 7 days of CT. Candidate lesions were compared with the endoscopic reference standard and prospective radiologist interpretation. The sensitivity and false-positive rates were calculated for each system. RESULTS The NIH system had a higher sensitivity than the PEV tool for polyps > or = 1 cm (22/23, 96%; 78-99%, 95% CI vs 14/23, 61%; 38-81%, 95% CI; p = 0.008, respectively). There was no significant difference in the detection of medium-sized polyps 6-9 mm in size (8/13 vs 6/13, p = 0.68, respectively). The PEV tool had an average of 1.18 false-positive detections per patient, whereas the NIH tool had an average of 5.20 false-positive detections per patient, with the PEV tool having significantly fewer false-positive detections in both patient groups (p < 0.001). CONCLUSION One polyp detection system tended to operate with a higher sensitivity, whereas the other tended to operate with a lower false-positive rate. Prospective trials using polyp detection systems as a primary or secondary means of CT colonography interpretation appear warranted.

Novel method and applications for labeling and identifying lymph nodes

The lymphatic system comprises a series of interconnected lymph nodes that are commonly distributed along branching or linearly oriented anatomic structures. Physicians must evaluate lymph nodes when staging cancer and planning optimal paths for nodal biopsy. This process requires accurately determining the lymph nodes position with respect to major anatomical landmarks. In an effort to standardize lung cancer staging, The American Joint Committee on Cancer (AJCC) has classified lymph nodes within the chest into 4 groups and 14 sub groups. We present a method for automatically labeling lymph nodes according to this classification scheme, in order to improve the speed and accuracy of staging and biopsy planning. Lymph nodes within the chest are clustered around the major blood vessels and the airways. Our fully automatic labeling method determines the nodal group and sub-group in chest CT data by use of computed airway and aorta centerlines to produce features relative to a given node location. A classifier then determines the label based upon these features. We evaluate the efficacy of the method on 10 chest CT datasets containing 86 labeled lymph nodes. The results are promising with 100% of the nodes assigned to the correct group and 76% to the correct sub-group. We anticipate that additional features and training data will further improve the results. In addition to labeling, other applications include automated lymph node localization and visualization. Although we focus on chest CT data, the method can be generalized to other regions of the body as well as to different imaging modalities.

Colon Unfolding Via Skeletal Subspace Deformation

We present an efficient method to digitally straighten a colon volume using mesh skinning, a technique well known in computer graphics to deform a polygonal mesh attached to a skeleton hierarchy. In our case, the colon centerline is used as the skeleton structure and the polyhedral model of the lumen as the skin that is to be deformed as the centerline is straightened. Once the colon has been straightened, we use standard rendering techniques to compute the virtual dissection. Our approach is significantly more efficient than previously proposed techniques.

Novel method for digital subtraction of tagged stool in virtual colonoscopy

Colon cancer is one of the most frequent causes of death. CT colonography is a novel method for the detection of polyps and early cancer. The general principle of CT colonography includes a cathartic bowel preparation. The resulting discomfort for patients leads to limited patient acceptance and therefore to limited cancer detection rates. Reduced bowel preparation, techniques for stool tagging, and electronic cleansing, however, improve the acceptance rates. Hereby, the high density of oral contrast material highlights residual stool and can be digitally removed. Known subtraction methods cause artifacts: additional 3D objects are introduced and small bowel folds are perforated. We propose a new algorithm that is based on the 2nd derivative of the image data using the Hessian matrix and the following principal axis transform to detect tiny folds which shall not be subtracted together with tagged stool found by a thresholding method. Since the stool is usually not homogenously tagged with contrast media a detection algorithm for island-like structures is incorporated. The interfaces of air-stool level and colon wall are detected by a 3-dimensional difference of Gaussian module. A 3-dimensional filter smoothes the transitions between removed stool and colon tissue. We evaluated the efficacy of the new algorithm with 10 patient data sets. The results showed no introduced artificial objects and no perforated folds. The artifacts at the air-stool and colon tissue-stool transitions are considerably reduced compared to those known from the literature.