Surface normal overlap: a computer-aided detection algorithm with application to colonic polyps and lung nodules in helical CT

Reliable evaluation of performance level for computer-aided diagnostic scheme

RATIONALE AND OBJECTIVES

Computer-aided diagnostic (CAD) schemes have been developed for assisting radiologists in the detection of various lesions in medical images. The reliable evaluation of CAD schemes is an important task in the field of CAD research.

MATERIALS AND METHODS

Many evaluation approaches have been proposed for evaluating the performance of various CAD schemes in the past. However, some important issues in the evaluation of CAD schemes have not been systematically analyzed. The first important issue is the analysis and comparison of various evaluation methods in terms of certain characteristics. The second includes the analysis of pitfalls in the incorrect use of various evaluation methods and the effective approaches to the reduction of the bias and variance caused by these pitfalls. We attempt to address the first important issue in details in this article by conducting Monte Carlo simulation experiments, and to discuss the second issue in the Discussion section.

RESULTS

No single evaluation method is universally superior to the others; different situations of CAD applications require different evaluation methods, as recommended in this article. Bias and variance in the estimated performance levels caused by various pitfalls can be reduced considerably by the correct use of good evaluation methods.

CONCLUSIONS

This article would be useful to researchers in the field of CAD research for selecting appropriate evaluation methods and for improving the reliability of the estimated performance of their CAD schemes.

Computer-aided detection of colonic polyps at CT colonography using a Hessian matrix-based algorithm: preliminary study

OBJECTIVE

The purpose of our study was to develop a Hessian matrix-based computer-aided detection (CAD) algorithm for polyp detection on CT colonography (CTC) and to analyze its performance in a high-risk population.

SUBJECTS AND METHODS

The CTC data sets of 35 patients with at least one colonoscopically proven polyp were interpreted with a Hessian matrix-based CAD algorithm, which was designed to depict bloblike structures protruding into the lumen. Our gold standard was a combination of segmental unblinded optical colonoscopy and retrospective unblinded consensus review by two radiologists. Sensitivity of CAD for polyp detection was evaluated on both per-polyp and per-patient bases. The average number of false-positive detections was calculated, and the causes of false-positives and false-negatives were analyzed.

RESULTS

Ninety-four polyps were identified on colonoscopy. Forty-six polyps were smaller than 6 mm and 48 were 6 mm or larger. Seventy-five (79.8%) of these 94 polyps were identified by radiologists in a retrospective review. When colonoscopy was used as a standard of reference, the sensitivity of CAD was 77.1% for polyps 6 mm or larger. For large polyps (> or = 6 mm) that could be identified on retrospective review, the CAD algorithm achieved sensitivities of 92.5% (37/40) and 91.7% (22/24), respectively, on per-polyp and per-patient bases. There were an average of 5.5 false-positive detections per patient and 3.1 false-positive detections per data set for CAD. The two most frequent causes of false-positives on CAD were prominent or converging fold (78/191) and feces (50/191). Of the three polyps 6 mm or larger that were missed by CAD, two had a flat appearance on colonoscopy and the remaining one was located in the narrow area between the rectal tube and the rectal wall.

CONCLUSION

A Hessian matrix-based CAD algorithm for CTC has the potential to depict polyps larger than or equal to 6 mm with high sensitivity and an acceptable false-positive rate.

Automated detection of small pulmonary nodules in whole lung CT scans

RATIONALE AND OBJECTIVES

The objective of this work was to develop and evaluate a robust algorithm that automatically detects small solid pulmonary nodules in whole lung helical CT scans from a lung cancer screening study.

MATERIALS AND METHODS

We developed a three-stage detection algorithm for both isolated and attached nodules. The algorithm consisted of nodule search space demarcation, nodule candidates' generation, and a sequential elimination of false positives. Isolated nodules are nodules that are surrounded by lung parenchyma, whereas attached nodules are connected to large, dense structures such as pleural and/or mediastinal surface. Two large well-documented whole lung CT scan databases (Databases A and B) were created to train and test the detection algorithm. Database A contains 250 sequentially selected scans with 2.5-mm slice thickness that were obtained at Weill Medical College of Cornell University. With equipment upgrade at this college, a second database, Database B, was created containing 250 scans with a 1.25-mm slice thickness. A total of 395 and 482 nodules were identified in Databases A and B, respectively. In both databases, the majority of the nodules were isolated, comprising 72.1% and 82.3% of nodules in Databases A and B, respectively.

RESULTS

The detection algorithm was trained and tested on both Databases A and B. For isolated nodules with sizes 4 mm or larger, the algorithm achieved 94.0% sensitivity and 7.1 false positives per case (FPPC) for Database A (2.5 mm). Similarly, the algorithm achieved 91% sensitivity and 6.9 FPPC for Database B (1.25 mm). The algorithm achieved 92% sensitivity with 17.4 FPPC and 89% sensitivity with 5.5 FFPC for attached nodules with sizes 3 mm or larger in the Database A (2.5 mm) and Database B (1.25 mm), respectively.

CONCLUSION

The developed algorithm achieved practical performance for automated detection of both isolated and the more challenging attached nodules. The automated system will be a useful tool to assist radiologists in identifying nodules from whole lung CT scans in a clinical setting.

Part-based local shape models for colon polyp detection

This paper presents a model-based technique for lesion detection in colon CT scans that uses analytical shape models to map the local shape curvature at individual voxels to anatomical labels. Local intensity profiles and curvature information have been previously used for discriminating between simple geometric shapes such as spherical and cylindrical structures. This paper introduces novel analytical shape models for colon-specific anatomy, viz. folds and polyps, built by combining parts with simpler geometric shapes. The models better approximate the actual shapes of relevant anatomical structures while allowing the application of model-based analysis on the simpler model parts. All parameters are derived from the analytical models, resulting in a simple voxel labeling scheme for classifying individual voxels in a CT volume. The algorithm's performance is evaluated against expert-determined ground truth on a database of 42 scans and performance is quantified by free-response receiver-operator curves.

Recent progress in computer-aided diagnosis of lung nodules on thin-section CT

Computer-aided diagnosis (CAD) provides a computer output as a "second opinion" in order to assist radiologists in the diagnosis of various diseases on medical images. Currently, a significant research effort is being devoted to the detection and characterization of lung nodules in thin-section computed tomography (CT) images, which represents one of the newest directions of CAD development in thoracic imaging. We describe in this article the current status of the development and evaluation of CAD schemes for the detection and characterization of lung nodules in thin-section CT. We also review a number of observer performance studies in which it was attempted to assess the potential clinical usefulness of CAD schemes for nodule detection and characterization in thin-section CT. Whereas current CAD schemes for nodule characterization have achieved high performance levels and would be able to improve radiologists' performance in the characterization of nodules in thin-section CT, current schemes for nodule detection appear to report many false positives, and, therefore, significant efforts are needed in order further to improve the performance levels of current CAD schemes for nodule detection in thin-section CT.

CAD in CT colonography without and with oral contrast agents: progress and challenges

Computed tomographic colonography (CTC), also known as virtual colonoscopy, is an emerging alternative technique for screening of colon cancers. CTC uses CT to provide a series of cross-sectional images of the colon for detection of polyps and masses. Fecal tagging is a means of labeling of residual feces by an oral contrast agent for improving the accuracy in the detection of polyps. Computer-aided diagnosis (CAD) for CTC automatically determines the locations of suspicious polyps and masses in CTC and presents them to radiologists, typically as a second opinion. Despite its relatively short history, CAD has become one of the mainstream techniques that could make CTC prime time for screening of colorectal cancer. Rapid technical developments have advanced CAD substantially during the last several years, and a fundamental scheme for the detection of polyps has been established, in which sophisticated 3D image processing, analysis, and display techniques play a pivotal role. The latest CAD systems indicate a clinically acceptable high sensitivity and a low false-positive rate, and observer studies have demonstrated the benefits of these systems in improving radiologists' detection performance. Some technical and clinical challenges, however, remain unresolved before CAD can become a truly useful tool for clinical practice. Also, new challenges are facing CAD as the methods for bowel preparation and image acquisition, such as tagging of fecal residue with oral contrast agents, and interpretation of CTC images evolve. This article reviews the current status and future challenges in CAD for CTC without and with fecal tagging.

A statistical approach for robust polyp detection in CT colonography

In this paper we describe the development of a computationally efficient computer-aided detection (CAD) algorithm based on the statistical features derived from the local colonic surface that are used for the detection of colonic polyps in computed tomography (CT) colonography. The candidate surface voxels were detected and clustered using the surface normal intersection, convexity test, region growing and Hough Transform. The main objective of this paper is the selection of the statistical features that optimally capture the convexity of the candidate surface and consequently provide a high discrimination between local surfaces defined by polyps and folds. The developed polyp detection scheme is computationally efficient (typically takes 3.9 minute per dataset) and shows 100% sensitivity for phantom polyps greater than 5mm and 87.5% sensitivity for real polyps greater than 5mm with an average of 4.05 false positives per dataset.

Improved Diagnosis and Navigation for CT Colonography

The goal of this research project is to develop a fast, accurate, and patient-friendly computer-aided diagnosis (CAD) component of CT colonography, that improves the robustness and accuracy of current colon wall segmentation and achieves earlier colorectal cancer diagnoses through an improved polyp detection method. Many advanced image processing techniques are applied to clearly outline the colon wall in the CT data set of human abdomen, and subtract the colon portion from the entire data set. After the subtraction, the detailed information and the surface curvature information on the colon wall is analyzed. The active contour model is assisted by presegmentation steps including mathematical morphology filtering, edge detection and other image processing techniques.

Lung Nodule Detection Via Bayesian Voxel Labeling

This paper describes a system for detecting pulmonary nodules in CT images. It aims to label individual image voxels in accordance to one of a number of anatomical (pulmonary vessels or junctions), pathological (nodules), or spurious (noise) events. The approach is orthodoxly Bayesian, with particular care taken in the objective establishment of prior probabilities and the incorporation of relevant medical knowledge. We provide, under explicit modeling assumptions, closed-form expressions for all the probability distributions involved. The technique is applied to real data, and we present a discussion of its performance.

A new method for spherical object detection and its application to computer aided detection of pulmonary nodules in CT images

A novel method called local shape controlled voting has been developed for spherical object detection in 3D voxel images. By introducing local shape properties into the voting procedure of normal overlap, the proposed method improves the capability of differentiating spherical objects from other structures, as the normal overlap technique only measures the 'density' of normal overlapping, while how the normals are distributed in 3D is not discovered. The proposed method was applied to computer aided detection of pulmonary nodules based on helical CT images. Experiments showed that this method attained a better performance compared to the original normal overlap technique.

Massive-training artificial neural network (MTANN) for reduction of false positives in computer-aided detection of polyps: Suppression of rectal tubes

One of the limitations of the current computer-aided detection (CAD) of polyps in CT colonography (CTC) is a relatively large number of false-positive (FP) detections. Rectal tubes (RTs) are one of the typical sources of FPs because a portion of a RT, especially a portion of a bulbous tip, often exhibits a cap-like shape that closely mimics the appearance of a small polyp. Radiologists can easily recognize and dismiss RT-induced FPs; thus, they may lose their confidence in CAD as an effective tool if the CAD scheme generates such "obvious" FPs due to RTs consistently. In addition, RT-induced FPs may distract radiologists from less common true positives in the rectum. Therefore, removal RT-induced FPs as well as other types of FPs is desirable while maintaining a high sensitivity in the detection of polyps. We developed a three-dimensional (3D) massive-training artificial neural network (MTANN) for distinction between polyps and RTs in 3D CTC volumetric data. The 3D MTANN is a supervised volume-processing technique which is trained with input CTC volumes and the corresponding "teaching" volumes. The teaching volume for a polyp contains a 3D Gaussian distribution, and that for a RT contains zeros for enhancement of polyps and suppression of RTs, respectively. For distinction between polyps and nonpolyps including RTs, a 3D scoring method based on a 3D Gaussian weighting function is applied to the output of the trained 3D MTANN. Our database consisted of CTC examinations of 73 patients, scanned in both supine and prone positions (146 CTC data sets in total), with optical colonoscopy as a reference standard for the presence of polyps. Fifteen patients had 28 polyps, 15 of which were 5-9 mm and 13 were 10-25 mm in size. These CTC cases were subjected to our previously reported CAD scheme that included centerline-based segmentation of the colon, shape-based detection of polyps, and reduction of FPs by use of a Bayesian neural network based on geometric and texture features. Application of this CAD scheme yielded 96.4% (27/28) by-polyp sensitivity with 3.1 (224/73) FPs per patient, among which 20 FPs were caused by RTs. To eliminate the FPs due to RTs and possibly other normal structures, we trained a 3D MTANN with ten representative polyps and ten RTs, and applied the trained 3D MTANN to the above CAD true- and false-positive detections. In the output volumes of the 3D MTANN, polyps were represented by distributions of bright voxels, whereas RTs and other normal structures partly similar to RTs appeared as darker voxels, indicating the ability of the 3D MTANN to suppress RTs as well as other normal structures effectively. Application of the 3D MTANN to the CAD detections showed that the 3D MTANN eliminated all RT-induced 20 FPs, as well as 53 FPs due to other causes, without removal of any true positives. Overall, the 3D MTANN was able to reduce the FP rate of the CAD scheme from 3.1 to 2.1 FPs per patient (33% reduction), while the original by-polyp sensitivity of 96.4% was maintained.

The use of 3D surface fitting for robust polyp detection and classification in CT colonography

In this paper we describe the development of a computationally efficient computer-aided detection (CAD) algorithm based on the evaluation of the surface morphology that is employed for the detection of colonic polyps in computed tomography (CT) colonography. Initial polyp candidate voxels were detected using the surface normal intersection values. These candidate voxels were clustered using the normal direction, convexity test, region growing and Gaussian distribution. The local colonic surface was classified as polyp or fold using a feature normalized nearest neighborhood classifier. The main merit of this paper is the methodology applied to select the robust features derived from the colon surface that have a high discriminative power for polyp/fold classification. The devised polyp detection scheme entails a low computational overhead (typically takes 2.20min per dataset) and shows 100% sensitivity for phantom polyps greater than 5mm. It also shows 100% sensitivity for real polyps larger than 10mm and 91.67% sensitivity for polyps between 5 to 10mm with an average of 4.5 false positives per dataset. The experimental data indicates that the proposed CAD polyp detection scheme outperforms other techniques that identify the polyps using features that sample the colon surface curvature especially when applied to low-dose datasets.

Development of a synthetic phantom for the selection of optimal scanning parameters in CAD-CT colonography

The aim of this paper is to present the development of a synthetic phantom that can be used for the selection of optimal scanning parameters in computed tomography (CT) colonography. In this paper we attempt to evaluate the influence of the main scanning parameters including slice thickness, reconstruction interval, field of view, table speed and radiation dose on the overall performance of a computer aided detection (CAD)-CTC system. From these parameters the radiation dose received a special attention, as the major problem associated with CTC is the patient exposure to significant levels of ionising radiation. To examine the influence of the scanning parameters we performed 51 CT scans where the spread of scanning parameters was divided into seven different protocols. A large number of experimental tests were performed and the results analysed. The results show that automatic polyp detection is feasible even in cases when the CAD-CTC system was applied to low dose CT data acquired with the following protocol: 13 mAs/rotation with collimation of 1.5 mm x 16 mm, slice thickness of 3.0mm, reconstruction interval of 1.5 mm, table speed of 30 mm per rotation. The CT phantom data acquired using this protocol was analysed by an automated CAD-CTC system and the experimental results indicate that our system identified all clinically significant polyps (i.e. larger than 5 mm).

CT Colonography with Computer-aided Polyp Detection: Volume and Attenuation Thresholds to Reduce False-Positive Findings Owing to the Ileocecal Valve

PURPOSE

To retrospectively identify volume and average attenuation thresholds for differentiating between ileocecal valve (ICV) and polyp at computed tomographic (CT) colonography with computer-aided detection (CAD).

MATERIALS AND METHODS

Informed consent (with consent for future retrospective research) and institutional review board (IRB) approval were obtained for the original prospective study. This retrospective study had IRB approval, as well, and was HIPAA-compliant. A total of 496 patients were selected from a larger screening population. CT colonographic images from 394 patients (227 men, 167 women; mean age, 58.0 years; range, 40-79 years) were used as a training set, and images from 102 patients (76 men, 26 women; mean age, 59.8 years; range, 46-79 years) were used as a test set. A series of 2742 volume and attenuation thresholds, for which segmented findings both larger in volume and lower in average attenuation were labeled as ICVs and remaining findings were labeled polyps, were applied to the training set to determine settings with 100% sensitivity for polyp detection and the highest specificity for ICV detection. The optimal settings were then applied to the test set. Significance was assessed with the Fisher exact test, and 95% confidence intervals (CIs) were computed for sensitivity and specificity.

RESULTS

A total of 386 ICVs and 67 adenomatous polyps from the training set and 102 ICVs and 138 adenomatous polyps from the test set could be segmented with a three-dimensional segmentation algorithm. When supine and prone images were counted individually, 746 nonunique ICVs from the training set and 191 from the test set were segmentable. In the training set, a volume of 600 mm(3) and an attenuation of 36 HU provided 100% sensitivity (67 polyps; 95% CI: 93%, 100%) and the optimal 83% specificity (618 of 746 ICVs; 95% CI: 80%, 85%). When applied to the test set, this combination provided 97% sensitivity (134 of 138 polyps; 95% CI: 92%, 99%) and 84% specificity (160 of 191 ICVs; 95% CI: 78%, 89%). Differences in sensitivity and specificity in the detection of polyps between the sets were not significant.

CONCLUSION

Volume and average CT attenuation thresholds can help differentiate most ICVs from true polyps.

A pipeline for computer aided polyp detection

We present a novel pipeline for computer-aided detection (CAD) of colonic polyps by integrating texture and shape analysis with volume rendering and conformal colon flattening. Using our automatic method, the 3D polyp detection problem is converted into a 2D pattern recognition problem. The colon surface is first segmented and extracted from the CT data set of the patient's abdomen, which is then mapped to a 2D rectangle using conformal mapping. This flattened image is rendered using a direct volume rendering technique with a translucent electronic biopsy transfer function. The polyps are detected by a 2D clustering method on the flattened image. The false positives are further reduced by analyzing the volumetric shape and texture features. Compared with shape based methods, our method is much more efficient without the need of computing curvature and other shape parameters for the whole colon surface. The final detection results are stored in the 2D image, which can be easily incorporated into a virtual colonoscopy (VC) system to highlight the polyp locations. The extracted colon surface mesh can be used to accelerate the volumetric ray casting algorithm used to generate the VC endoscopic view. The proposed automatic CAD pipeline is incorporated into an interactive VC system, with a goal of helping radiologists detect polyps faster and with higher accuracy.

Model-based analysis of local shape for lesion detection in CT scans

Thin-slice computer tomography provides high-resolution images that facilitate the diagnosis of early-stage lung cancer. However, the sheer size of the CT volumes introduces variability in radiological readings, driving the need for automated detection systems. The main contribution of this paper is a technique for combining geometric and intensity models with the analysis of local curvature for detecting pulmonary lesions in CT. The local shape at each voxel is represented via the principal curvatures of its associated isosurface without explicitly extracting the isosurface. The comparison of these curvatures to values derived from analytical shape models is then used to label the voxel as belonging to particular anatomical structures, e.g., nodules or vessels. The algorithm was evaluated on 242 CT exams with expert-determined ground truth. The performance of the algorithm is quantified by free-response receiver-operator characteristic curves, as well as by its potential for improvement in radiologist sensitivity.

CT colonography: influence of 3D viewing and polyp candidate features on interpretation with computer-aided detection

PURPOSE

To retrospectively determine if three-dimensional (3D) viewing improves radiologists' accuracy in classifying true-positive (TP) and false-positive (FP) polyp candidates identified with computer-aided detection (CAD) and to determine candidate polyp features that are associated with classification accuracy, with known polyps serving as the reference standard.

MATERIALS AND METHODS

Institutional review board approval and informed consent were obtained; this study was HIPAA compliant. Forty-seven computed tomographic (CT) colonography data sets were obtained in 26 men and 10 women (age range, 42-76 years). Four radiologists classified 705 polyp candidates (53 TP candidates, 652 FP candidates) identified with CAD; initially, only two-dimensional images were used, but these were later supplemented with 3D rendering. Another radiologist unblinded to colonoscopy findings characterized the features of each candidate, assessed colon distention and preparation, and defined the true nature of FP candidates. Receiver operating characteristic curves were used to compare readers' performance, and repeated-measures analysis of variance was used to test features that affect interpretation.

RESULTS

Use of 3D viewing improved classification accuracy for three readers and increased the area under the receiver operating characteristic curve to 0.96-0.97 (P<.001). For TP candidates, maximum polyp width (P=.038), polyp height (P=.019), and preparation (P=.004) significantly affected accuracy. For FP candidates, colonic segment (P=.007), attenuation (P<.001), surface smoothness (P<.001), distention (P=.034), preparation (P<.001), and true nature of candidate lesions (P<.001) significantly affected accuracy.

CONCLUSION

Use of 3D viewing increases reader accuracy in the classification of polyp candidates identified with CAD. Polyp size and examination quality are significantly associated with accuracy.

Computer-assisted reader software versus expert reviewers for polyp detection on CT colonography

OBJECTIVE

The purpose of our study was to assess the sensitivity of computer-assisted reader (CAR) software for polyp detection compared with the performance of expert reviewers.

MATERIALS AND METHODS

A library of colonoscopically validated CT colonography cases were collated and separated into training and test sets according to the time of accrual. Training data sets were annotated in consensus by three expert radiologists who were aware of the colonoscopy report. A subset of 45 training cases containing 100 polyps underwent batch analysis using ColonCAR version 1.2 software to determine the optimum polyp enhancement filter settings for polyp detection. Twenty-five consecutive positive test data sets were subsequently interpreted individually by each expert, who was unaware of the endoscopy report, and before generation of the annotated reference via an unblinded consensus interpretation. ColonCAR version 1.2 software was applied to the test cases, at optimized polyp enhancement filter settings, to determine diagnostic performance. False-positive findings were classified according to importance.

RESULTS

The 25 test cases contained 32 nondiminutive polyps ranging from 6 to 35 mm in diameter. The ColonCAR version 1.2 software identified 26 (81%) of 32 polyps compared with an average sensitivity of 70% for the expert reviewers. Eleven (92%) of 12 polyps > or = 10 mm were detected by ColonCAR version 1.2. All polyps missed by experts 1 (n = 4) and 2 (n = 3) and 12 (86%) of 14 polyps missed by expert 3 were detected by ColonCAR version 1.2. The median number of false-positive highlights per case was 13, of which 91% were easily dismissed.

CONCLUSION

ColonCAR version 1.2 is sensitive for polyp detection, with a clinically acceptable false-positive rate. ColonCAR version 1.2 has a synergistic effect to the reviewer alone, and its standalone performance may exceed even that of experts.

Computer analysis of computed tomography scans of the lung: a survey

Current computed tomography (CT) technology allows for near isotropic, submillimeter resolution acquisition of the complete chest in a single breath hold. These thin-slice chest scans have become indispensable in thoracic radiology, but have also substantially increased the data load for radiologists. Automating the analysis of such data is, therefore, a necessity and this has created a rapidly developing research area in medical imaging. This paper presents a review of the literature on computer analysis of the lungs in CT scans and addresses segmentation of various pulmonary structures, registration of chest scans, and applications aimed at detection, classification and quantification of chest abnormalities. In addition, research trends and challenges are identified and directions for future research are discussed.

Reduction of bias and variance for evaluation of computer-aided diagnostic schemes

Computer-aided diagnostic (CAD) schemes have been developed to assist radiologists in detecting various lesions in medical images. In addition to the development, an equally important problem is the reliable evaluation of the performance levels of various CAD schemes. It is good to see that more and more investigators are employing more reliable evaluation methods such as leave-one-out and cross validation, instead of less reliable methods such as resubstitution, for assessing their CAD schemes. However, the common applications of leave-one-out and cross-validation evaluation methods do not necessarily imply that the estimated performance levels are accurate and precise. Pitfalls often occur in the use of leave-one-out and cross-validation evaluation methods, and they lead to unreliable estimation of performance levels. In this study, we first identified a number of typical pitfalls for the evaluation of CAD schemes, and conducted a Monte Carlo simulation experiment for each of the pitfalls to demonstrate quantitatively the extent of bias and/or variance caused by the pitfall. Our experimental results indicate that considerable bias and variance may exist in the estimated performance levels of CAD schemes if one employs various flawed leave-one-out and cross-validation evaluation methods. In addition, for promoting and utilizing a high standard for reliable evaluation of CAD schemes, we attempt to make recommendations, whenever possible, for overcoming these pitfalls. We believe that, with the recommended evaluation methods, we can considerably reduce the bias and variance in the estimated performance levels of CAD schemes.

Computed tomographic virtual colonoscopy computer-aided polyp detection in a screening population

BACKGROUND & AIMS

The sensitivity of computed tomographic (CT) virtual colonoscopy (CT colonography) for detecting polyps varies widely in recently reported large clinical trials. Our objective was to determine whether a computer program is as sensitive as optical colonoscopy for the detection of adenomatous colonic polyps on CT virtual colonoscopy.

METHODS

The data set was a cohort of 1186 screening patients at 3 medical centers. All patients underwent same-day virtual and optical colonoscopy. Our enhanced gold standard combined segmental unblinded optical colonoscopy and retrospective identification of precise polyp locations. The data were randomized into separate training (n = 394) and test (n = 792) sets for analysis by a computer-aided polyp detection (CAD) program.

RESULTS

For the test set, per-polyp and per-patient sensitivities for CAD were both 89.3% (25/28; 95% confidence interval, 71.8%-97.7%) for detecting retrospectively identifiable adenomatous polyps at least 1 cm in size. The false-positive rate was 2.1 (95% confidence interval, 2.0-2.2) false polyps per patient. Both carcinomas were detected by CAD at a false-positive rate of 0.7 per patient; only 1 of 2 was detected by optical colonoscopy before segmental unblinding. At both 8-mm and 10-mm adenoma size thresholds, the per-patient sensitivities of CAD were not significantly different from those of optical colonoscopy before segmental unblinding.

CONCLUSIONS

The per-patient sensitivity of CT virtual colonoscopy CAD in an asymptomatic screening population is comparable to that of optical colonoscopy for adenomas > or = 8 mm and is generalizable to new CT virtual colonoscopy data.

Computer-aided detection for virtual colonoscopy

Colon cancer is one of the leading causes of cancer deaths in the developed countries. Most colon cancers can be prevented if precursor colon polyps are detected and removed. Virtual colonoscopy, or CT colonography, has shown promise to be the future screening tool for polyp detection, with a number of studies performed at academic institutions showing high sensitivity and specificity. Two main factors limiting CT colonography in general use are its excessive interpretation time and the variable sensitivity among readers. This article discusses the potential of computer-aided detection to address these problems. We also review the current state of research in this field and the future roles and challenges of CAD for CT colonography.

CAD techniques, challenges, and controversies in computed tomographic colonography

Computer-aided diagnosis (CAD) for computed tomographic colonography (CTC) automatically detects the locations of suspicious polyps and masses on CTC and provides radiologists with a second opinion. CAD has the potential to increase radiologists' diagnostic performance in the detection of polyps and masses and to decrease variability of the diagnostic accuracy among readers without significantly increasing the reading time. Technical developments have advanced CAD substantially during the past several years, and a fundamental scheme for the detection of polyps has been established. The most recent CAD systems based on this scheme produce a clinically acceptable high sensitivity and a low false-positive rate. However, CAD for CTC is still under active development, and the technology needs to be improved further. This report describes the expected benefits, the current fundamental scheme, the key techniques used for detection of polyps and masses on CTC, the current detection performance, as well as the pitfalls, challenges, controversies, and the future of CAD.

Computer-Aided diagnosis for CT colonography

CT colonography, or virtual colonoscopy, is a promising alternative screening tool for colon cancer. Computer-aided diagnosis (CAD) for CT colonography has the potential to increase radiologists' diagnostic performance in the detection of polyps and to reduce variability of the diagnostic accuracy among readers. Technical developments have advanced CAD for CT colonography substantially during the last several years. This paper describes the key techniques used for CAD for detection of polyps and masses in CT colonography, the current detection performance, and challenges and the future of CAD.

Registration of central paths and colonic polyps between supine and prone scans in computed tomography colonography: Pilot study

Computed tomography colonography (CTC) is a minimally invasive method that allows the evaluation of the colon wall from CT sections of the abdomen/pelvis. The primary goal of CTC is to detect colonic polyps, precursors to colorectal cancer. Because imperfect cleansing and distension can cause portions of the colon wall to be collapsed, covered with water, and/or covered with retained stool, patients are scanned in both prone and supine positions. We believe that both reading efficiency and computer aided detection (CAD) of CTC images can be improved by accurate registration of data from the supine and prone positions. We developed a two-stage approach that first registers the colonic central paths using a heuristic and automated algorithm and then matches polyps or polyp candidates (CAD hits) by a statistical approach. We evaluated the registration algorithm on 24 patient cases. After path registration, the mean misalignment distance between prone and supine identical anatomic landmarks was reduced from 47.08 to 12.66 mm, a 73% improvement. The polyp registration algorithm was specifically evaluated using eight patient cases for which radiologists identified polyps separately for both supine and prone data sets, and then manually registered corresponding pairs. The algorithm correctly matched 78% of these pairs without user input. The algorithm was also applied to the 30 highest-scoring CAD hits in the prone and supine scans and showed a success rate of 50% in automatically registering corresponding polyp pairs. Finally, we computed the average number of CAD hits that need to be manually compared in order to find the correct matches among the top 30 CAD hits. With polyp registration, the average number of comparisons was 1.78 per polyp, as opposed to 4.28 comparisons without polyp registration.