Lubomir M. Hadjiiski

Lung nodule detection on thoracic computed tomography images: preliminary evaluation of a computer-aided diagnosis system.

We are developing a computer-aided diagnosis (CAD) system for lung nodule detection on thoracic helical computed tomography (CT) images. In the first stage of this CAD system, lung regions are identified by a k-means clustering technique. Each lung slice is classified as belonging to the upper, middle, or the lower part of the lung volume. Within each lung region, structures are segmented again using weighted k-means clustering. These structures may include true lung nodules and normal structures consisting mainly of blood vessels. Rule-based classifiers are designed to distinguish nodules and normal structures using 2D and 3D features. After rule-based classification, linear discriminant analysis (LDA) is used to further reduce the number of false positive (FP) objects. We performed a preliminary study using 1454 CT slices from 34 patients with 63 lung nodules. When only LDA classification was applied to the segmented objects, the sensitivity was 84% (53/63) with 5.48 (7961/1454) FP objects per slice. When rule-based classification was used before LDA, the free response receiver operating characteristic (FROC) curve improved over the entire sensitivity and specificity ranges of interest. In particular, the FP rate decreased to 1.74 (2530/1454) objects per slice at the same sensitivity. Thus, compared to FP reduction with LDA alone, the inclusion of rule-based classification lead to an improvement in detection accuracy for the CAD system. These preliminary results demonstrate the feasibility of our approach to lung nodule detection and FP reduction on CT images.

A comparative study of limited-angle cone-beam reconstruction methods for breast tomosynthesis.

Digital tomosynthesis mammography (DTM) is a promising new modality for breast cancer detection. In DTM, projection-view images are acquired at a limited number of angles over a limited angular range and the imaged volume is reconstructed from the two-dimensional projections, thus providing three-dimensional structural information of the breast tissue. In this work, we investigated three representative reconstruction methods for this limited-angle cone-beam tomographic problem, including the backprojection (BP) method, the simultaneous algebraic reconstruction technique (SART) and the maximum likelihood method with the convex algorithm (ML-convex). The SART and ML-convex methods were both initialized with BP results to achieve efficient reconstruction. A second generation GE prototype tomosynthesis mammography system with a stationary digital detector was used for image acquisition. Projection-view images were acquired from 21 angles in 3 degrees increments over a +/- 30 degrees angular range. We used an American College of Radiology phantom and designed three additional phantoms to evaluate the image quality and reconstruction artifacts. In addition to visual comparison of the reconstructed images of different phantom sets, we employed the contrast-to-noise ratio (CNR), a line profile of features, an artifact spread function (ASF), a relative noise power spectrum (NPS), and a line object spread function (LOSF) to quantitatively evaluate the reconstruction results. It was found that for the phantoms with homogeneous background, the BP method resulted in less noisy tomosynthesized images and higher CNR values for masses than the SART and ML-convex methods. However, the two iterative methods provided greater contrast enhancement for both masses and calcification, sharper LOSF, and reduced interplane blurring and artifacts with better ASF behaviors for masses. For a contrast-detail phantom with heterogeneous tissue-mimicking background, the BP method had strong blurring artifacts along the x-ray source motion direction that obscured the contrast-detail objects, while the other two methods can remove the superimposed breast structures and significantly improve object conspicuity. With a properly selected relaxation parameter, the SART method with one iteration can provide tomosynthesized images comparable to those obtained from the ML-convex method with seven iterations, when BP results were used as initialization for both methods.

Improvement of mammographic mass characterization using spiculation measures and morphological features

We are developing new computer vision techniques for characterization of breast masses on mammograms. We had previously developed a characterization method based on texture features. The goal of the present work was to improve our characterization method by making use of morphological features. Toward this goal, we have developed a fully automated, three-stage segmentation method that includes clustering, active contour, and spiculation detection stages. After segmentation, morphological features describing the shape of the mass were extracted. Texture features were also extracted from a band of pixels surrounding the mass. Stepwise feature selection and linear discriminant analysis were employed in the morphological, texture, and combined feature spaces for classifier design. The classification accuracy was evaluated using the area Az under the receiver operating characteristic curve. A data set containing 249 films from 102 patients was used. When the leave-one-case-out method was applied to partition the data set into trainers and testers, the average test Az for the task of classifying the mass on a single mammographic view was 0.83 +/- 0.02, 0.84 +/- 0.02, and 0.87 +/- 0.02 in the morphological, texture, and combined feature spaces, respectively. The improvement obtained by supplementing texture features with morphological features in classification was statistically significant (p = 0.04). For classifying a mass as malignant or benign, we combined the leave-one-case-out discriminant scores from different views of a mass to obtain a summary score. In this task, the test Az value using the combined feature space was 0.91 +/- 0.02. Our results indicate that combining texture features with morphological features extracted from automatically segmented mass boundaries will be an effective approach for computer-aided characterization of mammographic masses.

Computer-aided characterization of mammographic masses: accuracy of mass segmentation and its effects on characterization

Mass segmentation is used as the first step in many computer-aided diagnosis (CAD) systems for classification of breast masses as malignant or benign. The goal of this paper was to study the accuracy of an automated mass segmentation method developed in our laboratory, and to investigate the effect of the segmentation stage on the overall classification accuracy. The automated segmentation method was quantitatively compared with manual segmentation by two expert radiologists (R1 and R2) using three similarity or distance measures on a data set of 100 masses. The area overlap measures between R1 and R2, the computer and R1, and the computer and R2 were 0.76/spl plusmn/0.13,0.74 /spl plusmn/0.11, and 0.74/spl plusmn/0.13, respectively. The interobserver difference in these measures between the two radiologists was compared with the corresponding differences between the computer and the radiologists. Using three similarity measures and data from two radiologists, a total of six statistical tests were performed. The difference between the computer and the radiologist segmentation was significantly larger than the interobserver variability in only one test. Two sets of texture, morphological, and spiculation features, one based on the computer segmentation, and the other based on radiologist segmentation, were extracted from a data set of 249 films from 102 patients. A classifier based on stepwise feature selection and linear discriminant analysis was trained and tested using the two feature sets. The leave-one-case-out method was used for data sampling. For case-based classification, the area A/sub z/ under the receiver operating characteristic (ROC) curve was 0.89 and 0.88 for the feature sets based on the radiologist segmentation and computer segmentation, respectively. The difference between the two ROC curves was not statistically significant.

Computer-aided diagnosis of pulmonary nodules on CT scans: Segmentation and classification using 3D active contours

We are developing a computer-aided diagnosis (CAD) system to classify malignant and benign lung nodules found on CT scans. A fully automated system was designed to segment the nodule from its surrounding structured background in a local volume of interest (VOI) and to extract image features for classification. Image segmentation was performed with a three-dimensional (3D) active contour (AC) method. A data set of 96 lung nodules (44 malignant, 52 benign) from 58 patients was used in this study. The 3D AC model is based on two-dimensional AC with the addition of three new energy components to take advantage of 3D information: (1) 3D gradient, which guides the active contour to seek the object surface, (2) 3D curvature, which imposes a smoothness constraint in the z direction, and (3) mask energy, which penalizes contours that grow beyond the pleura or thoracic wall. The search for the best energy weights in the 3D AC model was guided by a simplex optimization method. Morphological and gray-level features were extracted from the segmented nodule. The rubber band straightening transform (RBST) was applied to the shell of voxels surrounding the nodule. Texture features based on run-length statistics were extracted from the RBST image. A linear discriminant analysis classifier with stepwise feature selection was designed using a second simplex optimization to select the most effective features. Leave-one-case-out resampling was used to train and test the CAD system. The system achieved a test area under the receiver operating characteristic curve (A(z)) of 0.83 +/- 0.04. Our preliminary results indicate that use of the 3D AC model and the 3D texture features surrounding the nodule is a promising approach to the segmentation and classification of lung nodules with CAD. The segmentation performance of the 3D AC model trained with our data set was evaluated with 23 nodules available in the Lung Image Database Consortium (LIDC). The lung nodule volumes segmented by the 3D AC model for best classification were generally larger than those outlined by the LIDC radiologists using visual judgment of nodule boundaries.

Correlation between mammographic density and volumetric fibroglandular tissue estimated on breast MR images.

Previous studies have found that mammographic breast density is highly correlated with breast cancer risk. Therefore, mammographic breast density may be considered as an important risk factor in studies of breast cancer treatments. In this paper, we evaluated the accuracy of using mammograms for estimating breast density by analyzing the correlation between the percent mammographic dense area and the percent glandular tissue volume as estimated from MR images. A dataset of 67 cases having MR images (coronal 3-D SPGR T1-weighted pre-contrast) and corresponding 4-view mammograms was used in this study. Mammographic breast density was estimated by an experienced radiologist and an automated image analysis tool, Mammography Density ESTimator (MDEST) developed previously in our laboratory. For the estimation of the percent volume of fibroglandular tissue in breast MR images, a semiautomatic method was developed to segment the fibroglandular tissue from each slice. The tissue volume was calculated by integration over all slices containing the breast. Interobserver variation was measured for 3 different readers. It was found that the correlation between every two of the three readers for segmentation of MR volumetric fibroglandular tissue was 0.99. The correlations between the percent volumetric fibroglandular tissue on MR images and the percent dense area of the CC and MLO views segmented by an experienced radiologist were both 0.91. The correlation between the percent volumetric fibroglandular tissue on MR images and the percent dense area of the CC and MLO views segmented by MDEST was 0.91 and 0.89, respectively. The root-mean-square (rms) residual ranged from 5.4% to 6.3%. The mean bias ranged from 3% to 6%. The high correlation indicates that changes in mammographic density may be a useful indicator of changes in fibroglandular tissue volume in the breast.

Feature selection and classifier performance in computer-aided diagnosis: The effect of finite sample size

In computer-aided diagnosis (CAD), a frequently used approach for distinguishing normal and abnormal cases is first to extract potentially useful features for the classification task. Effective features are then selected from this entire pool of available features. Finally, a classifier is designed using the selected features. In this study, we investigated the effect of finite sample size on classification accuracy when classifier design involves stepwise feature selection in linear discriminant analysis, which is the most commonly used feature selection algorithm for linear classifiers. The feature selection and the classifier coefficient estimation steps were considered to be cascading stages in the classifier design process. We compared the performance of the classifier when feature selection was performed on the design samples alone and on the entire set of available samples, which consisted of design and test samples. The area Az under the receiver operating characteristic curve was used as our performance measure. After linear classifier coefficient estimation using the design samples, we studied the hold-out and resubstitution performance estimates. The two classes were assumed to have multidimensional Gaussian distributions, with a large number of features available for feature selection. We investigated the dependence of feature selection performance on the covariance matrices and means for the two classes, and examined the effects of sample size, number of available features, and parameters of stepwise feature selection on classifier bias. Our results indicated that the resubstitution estimate was always optimistically biased, except in cases where the parameters of stepwise feature selection were chosen such that too few features were selected by the stepwise procedure. When feature selection was performed using only the design samples, the hold-out estimate was always pessimistically biased. When feature selection was performed using the entire finite sample space, the hold-out estimates could be pessimistically or optimistically biased, depending on the number of features available for selection, the number of available samples, and their statistical distribution. For our simulation conditions, these estimates were always pessimistically (conservatively) biased if the ratio of the total number of available samples per class to the number of available features was greater than five.

Digital breast tomosynthesis is comparable to mammographic spot views for mass characterization.

PURPOSE To determine if digital breast tomosynthesis (DBT) performs comparably to mammographic spot views (MSVs) in characterizing breast masses as benign or malignant. MATERIALS AND METHODS This IRB-approved, HIPAA-compliant reader study obtained informed consent from all subjects. Four blinded Mammography Quality Standards Act-certified academic radiologists individually evaluated DBT images and MSVs of 67 masses (30 malignant, 37 benign) in 67 women (age range, 34-88 years). Images were viewed in random order at separate counterbalanced sessions and were rated for visibility (10-point scale), likelihood of malignancy (12-point scale), and Breast Imaging Reporting and Data System (BI-RADS) classification. Differences in mass visibility were analyzed by using the Wilcoxon matched-pairs signed-ranks test. Reader performance was measured by calculating the area under the receiver operating characteristic curve (A(z)) and partial area index above a sensitivity threshold of 0.90 (A(z)(0.90)) by using likelihood of malignancy ratings. Masses categorized as BI-RADS 4 or 5 were compared with histopathologic analysis to determine true-positive results for each modality. RESULTS Mean mass visibility ratings were slightly better with DBT (range, 3.2-4.4) than with MSV (range, 3.8-4.8) for all four readers, with one readers improvement achieving statistical significance (P = .001). The A(z) ranged 0.89-0.93 for DBT and 0.88-0.93 for MSV (P ≥ .23). The A(z)((0.90)) ranged 0.36-0.52 for DBT and 0.25-0.40 for MSV (P ≥ .20). The readers characterized seven additional malignant masses as BI-RADS 4 or 5 with DBT than with MSV, at a cost of five false-positive biopsy recommendations, with a mean of 1.8 true-positive (range, 0-3) and 1.3 false-positive (range, -1 to 4) assessments per reader. CONCLUSION In this small study, mass characterization in terms of visibility ratings, reader performance, and BI-RADS assessment with DBT was similar to that with MSVs. Preliminary findings suggest that MSV might not be necessary for mass characterization when performing DBT.

Computer-aided detection of breast masses on full field digital mammograms

We are developing a computer-aided detection (CAD) system for breast masses on full field digital mammographic (FFDM) images. To develop a CAD system that is independent of the FFDM manufacturers proprietary preprocessing methods, we used the raw FFDM image as input and developed a multiresolution preprocessing scheme for image enhancement. A two-stage prescreening method that combines gradient field analysis with gray level information was developed to identify mass candidates on the processed images. The suspicious structure in each identified region was extracted by clustering-based region growing. Morphological and spatial gray-level dependence texture features were extracted for each suspicious object. Stepwise linear discriminant analysis (LDA) with simplex optimization was used to select the most useful features. Finally, rule-based and LDA classifiers were designed to differentiate masses from normal tissues. Two data sets were collected: a mass data set containing 110 cases of two-view mammograms with a total of 220 images, and a no-mass data set containing 90 cases of two-view mammograms with a total of 180 images. All cases were acquired with a GE Senographe 2000D FFDM system. The true locations of the masses were identified by an experienced radiologist. Free-response receiver operating characteristic analysis was used to evaluate the performance of the CAD system. It was found that our CAD system achieved a case-based sensitivity of 70%, 80%, and 90% at 0.72, 1.08, and 1.82 false positive (FP) marks/image on the mass data set. The FP rates on the no-mass data set were 0.85, 1.31, and 2.14 FP marks/image, respectively, at the corresponding sensitivities. This study demonstrated the usefulness of our CAD techniques for automated detection of masses on FFDM images.

Computer-aided detection of lung nodules: False positive reduction using a 3D gradient field method and 3D ellipsoid fitting

We are developing a computer-aided detection system to assist radiologists in the detection of lung nodules on thoracic computed tomography (CT) images. The purpose of this study was to improve the false-positive (FP) reduction stage of our algorithm by developing features that extract three-dimensional (3D) shape information from volumes of interest identified in the prescreening stage. We formulated 3D gradient field descriptors, and derived 19 gradient field features from their statistics. Six ellipsoid features were obtained by computing the lengths and the length ratios of the principal axes of an ellipsoid fitted to a segmented object. Both the gradient field features and the ellipsoid features were designed to distinguish spherical objects such as lung nodules from elongated objects such as vessels. The FP reduction performance in this new 25-dimensional feature space was compared to the performance in a 19-dimensional space that consisted of features extracted using previously developed methods. The performance in the 44-dimensional combined feature space was also evaluated. Linear discriminant analysis with stepwise feature selection was used for classification. The parameters used for feature selection were optimized using the simplex algorithm. Training and testing were performed using a leave-one-patient-out scheme. The FP reduction performances in different feature spaces were evaluated by using the area Az under the receiver operating characteristic curve and the number of FPs per CT section at a given sensitivity as accuracy measures. Our data set consisted of 82 CT scans (3551 axial sections) from 56 patients with section thickness ranging from 1.0 to 2.5 mm. Our prescreening algorithm detected 111 of the 116 solid nodules (nodule size: 3.0-30.6 mm) marked by experienced thoracic radiologists. The test Az values were 0.95 +/- 0.01, 0.88 +/- 0.02, and 0.94 +/- 0.01 in the new, previous, and combined feature spaces, respectively. The number of FPs per section at 80% sensitivity in these three feature spaces were 0.37, 1.61, and 0.34, respectively. The improvement in the test Az with the 25 new features was statistically significant (p<0.0001) compared to that with the previous 19 features alone.