Comparative performance of a primary-reader and second-reader paradigm of computer-aided detection for CT colonography in a low-prevalence screening population.

OBJECTIVE: To compare the efficacy of computer-aided detection (CAD) for computed tomographic colonography (CTC) when employed as either primary-reader or second-reader paradigms in a low-prevalence screening population. METHODS: Ninety screening patients underwent same-day CTC and colonoscopy. Four readers prospectively interpreted all CTC data sets using a second-reader paradigm (unassisted interpretation followed immediately by CAD assistance). Three months later, randomized anonymous data sets were re-interpreted by all readers using a primary-reader paradigm (only CAD prompts evaluated). RESULTS: Compared with the average per-patient sensitivity for unassisted interpretation (0.57), both CAD paradigms significantly increased sensitivity: 0.78 (p < 0.001) for the second-reader paradigm and 0.83 (p < 0.001) for the primary-reader paradigm. There was no significant difference between CAD paradigms (p = 0.25). The average per-patient specificity for polyps ≥6 mm was significantly higher using the primary-reader paradigm than the second-reader paradigm (0.90 vs. 0.83, respectively, p = 0.006), with ROC AUCs of 0.83 and 0.68, respectively. Reading time using CAD as a primary-reader paradigm (median 1.4 min) was significantly shorter than both unassisted (median 4.0 min, p < 0.001) and second-reader paradigms (median 5.5 min, p < 0.001). CONCLUSION: CAD improves radiologist sensitivity in screening patients when used as either a second- or primary-reader paradigm, although the latter may improve specificity and efficiency more.

Incremental benefit of computer-aided detection when used as a second and concurrent reader of CT colonographic data: multiobserver study.

PURPOSE: To quantify the changes in reader performance levels, if any, during interpretation of computed tomographic (CT) colonographic data when a computer-aided detection (CAD) system is used as a second or concurrent reader. MATERIALS AND METHODS: After institutional review board approval was obtained, 16 experienced radiologists searched for polyps in 112 patients, 56 of whom had 132 polyps. Each case was interpreted on three separate occasions by using an unassisted (without CAD), second-read CAD, or concurrent CAD reading paradigm. The reading paradigm and case order were randomized, with a minimal interval of 1 month between consecutive interpretations. The readers' findings were compared with the reference-truth interpretation. The mean per-patient sensitivity and mean per-patient specificity with CAD were compared with those achieved with unassisted reading. An increase in per-patient sensitivity was considered to be clinically more important than an equivalent decrease in specificity. RESULTS: The mean per-patient sensitivity for identification of patients with polyps of any size increased significantly with use of second-read CAD (mean increase, 7.0%; 95% confidence interval [CI]: 4.0%, 9.8%) and concurrent CAD (mean increase, 4.5%; 95% CI: 0.8%, 8.2%). The mean per-patient specificity did not decrease significantly with use of second-read CAD (mean decrease, -2.5%; 95% CI: -5.2%, 0.1%) or concurrent CAD (mean decrease, -2.2%; 95% CI: -4.6%, 0.2%). With analysis restricted to patients with polyps 6 mm or larger, the benefit in sensitivity with second-read CAD remained (mean increase, 7.1%; 95% CI: 3.0%, 11.1%), whereas the increase with concurrent CAD was not significant (mean increase, 4.2%; 95% CI: -0.5%, 8.9%). Use of second-read CAD significantly increased the per-polyp sensitivity for polyps 6 mm or larger (mean increase, 9.0%; 98.3% CI: 4.9%, 12.8%) and polyps 5 mm or smaller (mean increase, 5.9%; 98.3% CI: 3.2%, 9.1%), but use of concurrent CAD increased the per-polyp sensitivity for only those polyps 5 mm or smaller (mean increase, 4.8%; 98.3% CI: 2.2%, 7.9%). CONCLUSION: Use of second-read CAD significantly improves readers' per-patient and per-polyp detection. Concurrent CAD is less effective. SUPPLEMENTAL MATERIAL: http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.10100354/-/DC1.

Automatic Graph Cut Segmentation of Lesions in CT Using Mean Shift Superpixels.

This paper presents a new, automatic method of accurately extracting lesions from CT data. It first determines, at each voxel, a five-dimensional (5D) feature vector that contains intensity, shape index, and 3D spatial location. Then, nonparametric mean shift clustering forms superpixels from these 5D features, resulting in an oversegmentation of the image. Finally, a graph cut algorithm groups the superpixels using a novel energy formulation that incorporates shape, intensity, and spatial features. The mean shift superpixels increase the robustness of the result while reducing the computation time. We assume that the lesion is part spherical, resulting in high shape index values in a part of the lesion. From these spherical subregions, foreground and background seeds for the graph cut segmentation can be automatically obtained. The proposed method has been evaluated on a clinical CT dataset. Visual inspection on different types of lesions (lung nodules and colonic polyps), as well as a quantitative evaluation on 101 solid and 80 GGO nodules, both demonstrate the potential of the proposed method. The joint spatial-intensity-shape features provide a powerful cue for successful segmentation of lesions adjacent to structures of similar intensity but different shape, as well as lesions exhibiting partial volume effect.

Shape-based computer-aided detection of lung nodules in thoracic CT images.

In this paper, a new computer tomography (CT) lung nodule computer-aided detection (CAD) method is proposed for detecting both solid nodules and ground-glass opacity (GGO) nodules (part solid and nonsolid). This method consists of several steps. First, the lung region is segmented from the CT data using a fuzzy thresholding method. Then, the volumetric shape index map, which is based on local Gaussian and mean curvatures, and the "dot" map, which is based on the eigenvalues of a Hessian matrix, are calculated for each voxel within the lungs to enhance objects of a specific shape with high spherical elements (such as nodule objects). The combination of the shape index (local shape information) and "dot" features (local intensity dispersion information) provides a good structure descriptor for the initial nodule candidates generation. Antigeometric diffusion, which diffuses across the image edges, is used as a preprocessing step. The smoothness of image edges enables the accurate calculation of voxel-based geometric features. Adaptive thresholding and modified expectation-maximization methods are employed to segment potential nodule objects. Rule-based filtering is first used to remove easily dismissible nonnodule objects. This is followed by a weighted support vector machine (SVM) classification to further reduce the number of false positive (FP) objects. The proposed method has been trained and validated on a clinical dataset of 108 thoracic CT scans using a wide range of tube dose levels that contain 220 nodules (185 solid nodules and 35 GGO nodules) determined by a ground truth reading process. The data were randomly split into training and testing datasets. The experimental results using the independent dataset indicate an average detection rate of 90.2%, with approximately 8.2 FP/scan. Some challenging nodules such as nonspherical nodules and low-contrast part-solid and nonsolid nodules were identified, while most tissues such as blood vessels were excluded. The method's high detection rate, fast computation, and applicability to different imaging conditions and nodule types shows much promise for clinical applications.

Flat neoplasia of the colon: CT colonography with CAD.

Attention is focusing on testing and developing computer aided detection (CAD) systems to reliably highlight flat polyps and cancers to the reporting radiologist during CT colonography. This review will discuss the clinical relevance of flat colonic neoplasia, describe some of the challenges facing CAD detection algorithms, and review the current CAD literature on this topic.

Computer-aided detection in computed tomography colonography: current status and problems with detection of early colorectal cancer.

PURPOSE: The aim of this study was to evaluate the usefulness of computer-aided detection (CAD) in diagnosing early colorectal cancer using computed tomography colonography (CTC). MATERIALS AND METHODS: A total of 30 CTC data sets for 30 early colorectal cancers in 30 patients were retrospectively reviewed by three radiologists. After primary evaluation, a second reading was performed using CAD findings. The readers evaluated each colorectal segment for the presence or absence of colorectal cancer using five confidence rating levels. To compare the assessment results, the sensitivity and specificity with and without CAD were calculated on the basis of the confidence rating, and differences in these variables were analyzed by receiver operating characteristic (ROC) analysis. RESULTS: The average sensitivities for the detection without and with CAD for the three readers were 81.6% and 75.6%, respectively. Among the three readers, only one reader improved sensitivity with CAD compared to that without. CAD decreased specificity in all three readers. CAD detected 100% of protruding lesions but only 69.2% of flat lesions. On ROC analysis, the diagnostic performance of all three readers was decreased by use of CAD. CONCLUSION: Currently available CAD with CTC does not improve diagnostic performance for detecting early colorectal cancer. An improved CAD algorithm is required for detecting flat lesions and reducing the false-positive rate.

Reduction of false positives in polyp detection using weighted support vector machines.

Colorectal cancer is the third highest cause of cancer deaths in US (2007). Early detection and treatment of colon cancer can significantly improve patient prognosis. Manual identification of polyps by radiologists using CT Colonography can be labour intensive due to the increasing size of datasets and is error prone due to the complexity of the anatomical structures. There has been increasing interest in computer aided detection (CAD) systems for detecting polyps using CT Colonography. For a typical CAD system two major steps can be identified. In the first step image processing techniques are used to detect potential polyp candidates. Many non-polyps are inevitably found in this process. The second step attempts to discount the non-polyp candidates while maintaining true polyps. In practice this is a challenging task as training data is heavily imbalanced, that is, non-polyps dominate the data. This paper describes how the weighted support vector machine (weighted-SVM) can be used to tackle the problem effectively. The weighted-SVM generalizes the traditional SVM by applying different penalties to different classes. This trains the classifier to give favour to the most weighted class (in this case true polyps). In this paper the method was applied to data obtained from the intermediate results from a CAD system, originally applied to 209 cases. The results show that the weighted-SVM can play an important role in CAD algorithms for colorectal polyps.

Efficient computer-aided detection of ground-glass opacity nodules in thoracic CT images.

In this paper, an efficient compute-aided detection method is proposed for detecting Ground-Glass Opacity (GGO) nodules in thoracic CT images. GGOs represent a clinically important type of lung nodule which are ignored by many existing CAD systems. Anti-geometric diffusion is used as preprocessing to remove image noise. Geometric shape features (such as shape index and dot enhancement), are calculated for each voxel within the lung area to extract potential nodule concentrations. Rule based filtering is then applied to remove False Positive regions. The proposed method has been validated on a clinical dataset of 50 thoracic CT scans that contains 52 GGO nodules. A total of 48 nodules were correctly detected and resulted in an average detection rate of 92.3%, with the number of false positives at approximately 12.7/scan (0.07/slice). The high detection performance of the method suggested promising potential for clinical applications.