Straightening the colon with curved cross sections: an approach to CT colonography

Machine Learning for Automated Polyp Detection in Computed Tomography Colonography

This chapter presents a comprehensive scheme for automated detection of colorectal polyps in computed tomography colonography (CTC) with particular emphasis on robust learning algorithms that differentiate polyps from non-polyp shapes. The authors’ automated CTC scheme introduces two orientation independent features which encode the shape characteristics that aid in classification of polyps and non-polyps with high accuracy, low false positive rate, and low computations making the scheme suitable for colorectal cancer screening initiatives. Experiments using state-of-the-art machine learning algorithms viz., lazy learning, support vector machines, and naïve Bayes classifiers reveal the robustness of the two features in detecting polyps at 100% sensitivity for polyps with diameter greater than 10 mm while attaining total low false positive rates, respectively, of 3.05, 3.47 and 0.71 per CTC dataset at specificities above 99% when tested on 58 CTC datasets. The results were validated using colonoscopy reports provided by expert radiologists.

Normalized distance along the colon centerline: a method for correlating polyp location on CT colonography and optical colonoscopy

OBJECTIVE

The ability to accurately locate a polyp found on CT colonography (CTC) at subsequent optical colonoscopy (OC) is an important part of the successful implementation of CTC for colorectal cancer screening. The purpose of this study was to determine whether a polyp's normalized distance along the colon centerline derived from CTC data can accurately predict its location on OC.

MATERIALS AND METHODS

The polyp population consisted of 152 polyps in 121 patients. CTC polyp findings were verified by same-day segmentally-unblinded OC. Each polyp's normalized distance along the colon centerline was computed by dividing its distance from the anorectal junction measured along the colon centerline by the length of the colon at CTC. The predicted polyp location at OC was computed by multiplying the normalized distance along the colon centerline by the colon length at OC (i.e., the distance to the cecum as determined at full colonoscope insertion). The differences between the true and predicted polyp locations at OC were compared using paired Student's t tests, linear regression, prediction interval assessment, and Bland-Altman analyses.

RESULTS

The differences between the true and predicted polyp locations at OC using the supine and prone CTC-normalized distances along the colon centerline were 2.2 +/- 10.5 cm (mean +/- SD; n = 136) and 1.5 +/- 10.5 cm (n = 135), respectively. The predicted location was within 10 cm of its true location for 71.3% (97/136) to 74.8% (101/135) of polyps and within 20 cm of its true location for 93.3% (126/135) to 93.4% (127/136) of polyps.

CONCLUSION

By computing the normalized distance along the colon centerline of a polyp found at CTC, the location of a polyp at OC can be predicted to within 10 cm (i.e., 1 colonoscope mark) for the majority of polyps.

Digital eversion of a hollow structure: an application in virtual colonography

A new methodology is presented for digital eversion of a hollow structure. The digital eversion is advantageous for better visualization of a larger portion of the inner surface with preservation of geometric relationship and without time-consuming navigation. Together with other techniques, digital eversion may help improve screening, diagnosis, surgical planning, and medical education. Two eversion algorithms are proposed and evaluated in numerical simulation to demonstrate the feasibility of the approach.

Two- versus Three-dimensional Colon Evaluation with Recently Developed Virtual Dissection Software for CT Colonography

This retrospective study was institutional review board approved; the requirement for informed patient consent was waived. The purpose of this study was to retrospectively compare a two-dimensional (2D) data interpretation technique with a three-dimensional (3D) colon dissection technique in terms of interpretation time and sensitivity for colonic polyp detection, with colonoscopy as the reference standard. Ninety-six patients (56 men, 40 women; mean age, 54.8 years) underwent colonoscopy and multidetector computed tomographic (CT) colonography on the same day. Two radiologists independently analyzed the data on a per-polyp and per-patient basis. The sensitivity of both approaches was compared by using the McNemar test. The time required to interpret CT colonographic data with each technique was also assessed. Compared with the conventional 2D colonic polyp detection method, primary 3D interpretation with use of virtual dissection software for CT colonography revealed comparable per-polyp (77% and 69% for two readers) and per-patient (77% and 73% for two readers) sensitivities and comparable per-patient specificity (99% and 89% for two readers) for the detection of polyps 6 mm in diameter or larger and involved a shorter interpretation time.

Three-dimensional virtual dissection at CT colonography: unraveling the colon to search for lesions

Computed tomographic (CT) colonography is a promising noninvasive examination for colorectal cancer screening; however, the optimal interpretation strategy remains undecided. Virtual dissection is an innovative technique whereby the three-dimensional (3D) model of the colon is virtually unrolled, sliced open, and displayed as a flat 3D rendering of the mucosal surface, similar to a gross pathologic specimen. This technique has the potential to reduce evaluation time by providing a more rapid 3D image assessment than is possible with an antegrade and retrograde 3D endoluminal fly-through. It may also ultimately improve accuracy by reducing blind spots present with 3D endoluminal displays and by reducing reader fatigue. A disadvantage of virtual dissection is the potential for distortion of colonic lesions and normal anatomy. To avoid potential pitfalls in image interpretation, the radiologist must be familiar with the unique appearance of the normal colon anatomy and of various pathologic findings when using virtual dissection with two-dimensional axial and 3D endoluminal CT colonographic image data sets.

Virtual Endoscopic Visualization of the Colon by Shape–Scale Signatures

We developed a new visualization method for virtual endoscopic examination of computed tomographic (CT) colonographic data by use of shape-scale analysis. The method provides each colonic structure of interest with a unique color, thereby facilitating rapid diagnosis of the colon. Two shape features, called the local shape index and curvedness, are used for defining the shape-scale spectrum. When we map the shape index and curvedness values within CT colonographic data to the shape-scale spectrum, specific types of colonic structures are represented by unique characteristic signatures in the spectrum. The characteristic signatures of specific types of lesions can be determined by use of computer-simulated lesions or by use of clinical data sets subjected to a computerized detection scheme. The signatures are used for defining a two-dimensional color map by assignment of a unique color to each signature region. The method was evaluated visually by use of computer-simulated lesions and clinical CT colonographic data sets, as well as by an evaluation of the human observer performance in the detection of polyps without and with the use of the color maps. The results indicate that the coloring of the colon yielded by the shape-scale color maps can be used for differentiating among the chosen colonic structures. Moreover, the results indicate that the use of the shape-scale color maps can improve the performance of radiologists in the detection of polyps in CT colonography.

Fast and robust computation of colon centerline in CT colonography

Although several methods for generating the centerline of a colon from CT colonographic scans have been proposed, in general they are time-consuming and do not take into account that the images of the colon may be of nonoptimal quality, with collapsed regions, and stool within the colon. Furthermore, the colonic lumen or wall, which is often used as a basis for computation of a centerline, is not always precisely segmented. In this study, we have developed an algorithm for computation of a colon centerline that is fast compared to the centerline algorithms presented in the reviewed literature, and that relies little on a complete colon segments identification. The proposed algorithm first extracts local maxima in a distance map of a segmented colonic lumen. The maxima are considered to be nodes in a set of graphs, and are iteratively linked together, based on a set of connection criteria, giving a minimum distance spanning tree. The connection criteria are computed from the distance from object boundary, the Euclidean distance between nodes and the voxel values on the pathway between pairs of nodes. After the last iteration, redundant branches are removed and end segments are recovered for each remaining graph. A subset of the initial maxima is used for distinguishing between the colon and noncolonic centerline segments among the set of graphs, giving the final centerline representation. A phantom study showed that, with respect to phantom variations, the algorithm achieved nearly constant computation time (2.3-2.9 s) except for the most extreme setting (20.2 s). The algorithm successfully found all, or most of, the centerline (93% - 100%). Displacement from optimum varied with colon diameter (1.2-6.6 mm). By use of 40 CT colonographic scans, the computer-generated centerlines were compared with the centerlines generated by three radiologists. The similarity was measured based on percent coverage and average displacement. The computer-generated centerlines, when compared with human-generated centerlines, had approximately the same displacement as when the human-generated centerlines were compared among each other (3.8 mm versus 4.0 mm). The coverage of the computer-generated centerlines was slightly less than that of the human-generated centerlines (92% versus 94%). The 40 centerlines were, on average, computed in 10.5 seconds, including computation time for the distance transform, with an Intel Pentium-based 800 MHz computer, as compared with 12-17 seconds or more (excluding computation time for the distance transform needed) per centerline as reported in other studies.

Virtual Colon Dissection with CT Colonography Compared with Axial Interpretation and Conventional Colonoscopy:Preliminary Results

OBJECTIVE

The aim of this study was to determine whether a new virtual colon dissection 3D visualization technique for CT colonography has a shorter analysis time and better sensitivity for detection of colonic polyps than interpretation of axial CT images. SUBJECTS AND METHODS. CT colonography was performed in 22 patients using 4-MDCT followed by conventional colonoscopy on the same day. The CT colonography data sets were analyzed by virtual colon dissection, which virtually bisects and unfolds the colon along its longitudinal axis to inspect the inner colonic surface for polyps. The same CT data sets were independently evaluated using axial interpretation. All data sets were independently interpreted by two radiologists in a blinded manner.

RESULTS

Conventional colonoscopy revealed 31 colonic lesions in 20 patients. Twenty two of the lesions were smaller than 10 mm; nine were 10 mm or larger. Two of the original 22 patients were excluded, one because of residual stool and fluid and the other because of an impassable stenosing rectal wall cancer. For virtual colon dissection, the per-lesion sensitivity was 42% for observer 1 and 68% for observer 2; for axial interpretation, the respective sensitivities were 48% and 61%. For polyps 10 mm or larger, the respective sensitivities were 67% and 89% for virtual colon dissection and 89% and 100% for axial interpretation. The average time for reconstruction and analysis of virtual colon dissection was 36.8 min versus 29.2 min for axial images. Virtual colon dissection was feasible in both the supine and the prone positions in 45.5% of colonic segments, in either the supine or the prone position in 24.5%, and in neither position in 30% of segments.

CONCLUSION

Although virtual colon dissection may facilitate detection of colonic polyps in isolated cases, its detection rate is not superior to axial interpretation, which is mainly attributable to failed rendering of insufficiently distended colonic segments or regions with residual feces. Virtual colon dissection is also the more time-consuming of the two procedures. With further improvement of path-finding and image segmentation, however, virtual colon dissection has the potential to be a useful interpretation tool for CT colonography.

Three-dimensional display modes for CT colonography: conventional 3D virtual colonoscopy versus unfolded cube projection

The authors compared a conventional two-directional three-dimensional (3D) display for computed tomography (CT) colonography with an alternative method they developed on the basis of time efficiency and surface visibility. With the conventional technique, 3D ante- and retrograde cine loops were obtained (hereafter, conventional 3D). With the alternative method, six projections were obtained at 90 degrees viewing angles (unfolded cube display). Mean evaluation time per patient with the conventional 3D display was significantly longer than that with the unfolded cube display. With the conventional 3D method, 93.8% of the colon surface came into view; with the unfolded cube method, 99.5% of the colon surface came into view. Sensitivity and specificity were not significantly different between the two methods. Agreements between observers were kappa = 0.605 for conventional 3D display and kappa = 0.692 for unfolded cube display. Consequently, the latter method enhances the 3D endoluminal display with improved time efficiency and higher surface visibility.

Colorectal cancer screening: a challenge for magnetic resonance colonography

The high incidence of colorectal carcinoma and the fact that colorectal cancer mostly arises from benign adenomas have led to recommendations for screening programs. The introduction of ultrafast three-dimensional datasets acquired by cross-sectional imaging modalities (computed tomography or magnetic resonance imaging) in combination with new postprocessing modes, known as virtual endoscopy, has led to new discussion on the recommendation of screening tests for colorectal cancer. Published results have indicated a high sensitivity for computed tomographic colonography and magnetic resonance-based colonography. Both techniques currently must be combined with colon cleansing. Three-dimensional data acquisition for magnetic resonance-based colonography is less than 1 minute using three-dimensional gradient-echo sequences. The lack of ionizing radiation, the low risk and discomfort to patients, and new techniques of minimized patient preparation make this magnetic resonance technique an attractive diagnostic procedure for colorectal lesions, with many aspects for use as a screening method.

Computerized detection of colonic polyps at CT colonography on the basis of volumetric features: pilot study

PURPOSE

To develop a computer-aided diagnosis (CAD) scheme for automated detection of colonic polyps on the basis of volumetric features and to assess its accuracy on the basis of colonoscopy, the standard.

MATERIALS AND METHODS

Computed tomographic (CT) colonography was performed in patients with use of standard bowel cleansing, air insufflation, and helical scanning in supine and prone positions. The colon was extracted from volumetric data sets generated from transverse CT sections. Volumetric features characterizing polyps were computed at each point in the extracted colon. Polyps were detected by means of hysteresis thresholding and fuzzy clustering followed by a rule-based test on the basis of feature values. Locations of the detected polyps were compared with those detected at conventional colonoscopy.

RESULTS

Forty-one cases were analyzed: nine cases with polyps and 32 without polyps. Each case with polyps had one polyp of clinically important size (six were 5-9 mm; three, 10 mm). Thus, there were 82 volumetric data sets, 18 included polyps. Eighty-nine percent (16 of 18) of the polyps were detected. Each of the two false-negative findings was detected in the other position; thus, 100% of polyp cases were detected, with 2.5 false-positive findings per patient. The false-positive findings were similar to those due to common perceptual errors. Most of the false-positive findings were easily distinguishable from true polyps by experienced radiologists.

CONCLUSION

The CAD scheme has the potential to depict polyps with high sensitivity and an acceptable false-positive rate.

Colon cancer screening with virtual colonoscopy: promise, polyps, politics

Virtual colonoscopy (CT colonography) promises to become a primary method for colorectal cancer screening and return radiologists to a major role in colon cancer prevention. Results from major centers in the United States show accuracy to be comparable to conventional colonoscopy for detection of polyps of significant size--that is, greater than 10 mm--with few false-positives. The advent of virtual colonoscopy has also heightened awareness of the natural history of colonic polyps, particularly in terms of identifying an appropriate target size for detection in colorectal screening programs. Small polyps (<10 mm) are often either hyperplastic on histology or are unlikely to progress to frank cancer in the patient's lifetime and are therefore of little clinical significance for the average adult. Thus, the rationale for detecting and removing each and every colonic polyp regardless of size has come under increasing scrutiny in the context of cost-benefit analysis of various test strategies for colorectal cancer screening. Virtual colonoscopy may allow patients to obtain reliable information about the status of their colonic mucosa noninvasively and thus make a more informed decision as to whether to proceed to conventional colonoscopy for polypectomy.

Colonography using multislice CT

Computed tomography (CT) represents the preferred imaging modality for imaging the large bowel when virtual endoscopic reconstructions are desired. Using the spiral acquisition technique, it has become possible to scan the entire abdomen within a single breathhold, however, slice thicknesses of 5 mm or more are necessary should the breathhold not last longer than 30-40 s. With the advent of multislice CT, contiguous 1-mm slices can be obtained through the entire abdomen while even shortening the breathhold to 25-30 s. The improved speed and spatial resolution of multislice CT results in remarkably sharp virtual reconstructions allowing detection of polyps with sizes less than 3 mm. The disadvantages must still be considered including a dataset consisting of up to 800 images representing a new challenge for postprocessing hard- and software.

Distortion reduction for fast soft straightening of the colon

RATIONALE AND OBJECTIVES

The purpose of this study was to suppress the geometric distortion associated with soft straightening of the colon by moderately adjusting curved cross sections, which is equivalent to appropriately modifying the underlying electrical field.

MATERIALS AND METHODS

A computational mechanism to reduce the geometric distortion associated with soft straightening was developed. Because the cause of distortion is uneven sampling of the colon wall with curved cross sections, the curved cross sections formed according to a numerically simulated electrical field were redistributed in the distortion reduction process. This mechanism can be directly incorporated into a previously published fast soft-straightening algorithm. Simulations were performed to evaluate the effectiveness of the mechanism, and a phantom colon with digitally implanted, spherical polyps was straightened and rendered to visually display the efficacy of the mechanism.

RESULTS

The distortion reduction mechanism was advantageous regarding shape preservation of polyps and alleviated distortion substantially. The mechanism occupied only a small portion of the total processing time.

CONCLUSION

The distortion reduction mechanism can effectively reduce the geometric distortion associated with soft straightening at little computational cost and has potential for use with computed tomographic colonography.

Fast algorithm for soft straightening of the colon

RATIONALE AND OBJECTIVES

In this study, the authors developed a fast algorithm for soft straightening of the colon with computed tomographic data that greatly accelerates the unraveling process based on the interpolation of representative electric force lines.

MATERIALS AND METHODS

Each curved cross section of the colon is defined by electric force lines of a common origin on an electrically charged central path and is constructed by interpolating most of these force lines from a limited number of representative force lines that are traced directly. Both a synthetic colon phantom and a colon in a living patient were used to demonstrate the feasibility of the fast interpolation algorithm compared with direct implementation for soft straightening of the colon.

RESULTS

The interpolation-based soft-straightening algorithm ran approximately 40 times faster than the direct implementation of the electric field-based soft-straightening algorithm.

CONCLUSION

The fast algorithm for soft straightening of the colon has potential for use in computed tomographic colonography.