Architectural distortion is an important sign of breast cancer, but because of its subtlety, it is a common cause of false-negative findings on screening mammograms. This paper presents methods for the detection of architectural distortion in mammograms of interval cancer cases taken prior to the detection of breast cancer using Gabor filters, phase portrait analysis, fractal analysis, and texture analysis. The methods were used to detect initial candidates for sites of architectural distortion in prior mammograms of interval cancer and also normal control cases. A total of 4,224 regions of interest (ROIs) were automatically obtained from 106 prior mammograms of 56 interval cancer cases, including 301 ROIs related to architectural distortion, and from 52 prior mammograms of 13 normal cases. For each ROI, the fractal dimension and Haralick’s texture features were computed. Feature selection was performed separately using stepwise logistic regression and stepwise regression. The best results achieved, in terms of the area under the receiver operating characteristics curve, with the features selected by stepwise logistic regression are 0.76 with the Bayesian classifier, 0.73 with Fisher linear discriminant analysis, 0.77 with an artificial neural network based on radial basis functions, and 0.77 with a support vector machine. Analysis of the performance of the methods with free-response receiver operating characteristics indicated a sensitivity of 0.80 at 7.6 false positives per image. The methods have good potential in detecting architectural distortion in mammograms of interval cancer cases.
Breast diseases; computer-assisted detection; computer-aided diagnosis (CAD); digital image processing; image analysis; mammography CAD; pattern recognition; ROC-based analysis
The effect of pixel resolution on texture features computed using the gray-level co-occurrence matrix (GLCM) was analyzed in the task of discriminating mammographic breast lesions as benign masses or malignant tumors. Regions in mammograms related to 111 breast masses, including 65 benign masses and 46 malignant tumors, were analyzed at pixel sizes of 50, 100, 200, 400, 600, 800, and 1,000 μm. Classification experiments using each texture feature individually provided accuracy, in terms of the area under the receiver operating characteristics curve (AUC), of up to 0.72. Using the Bayesian classifier and the leave-one-out method, the AUC obtained was in the range 0.73 to 0.75 for the pixel resolutions of 200 to 800 μm, with 14 GLCM-based texture features using adaptive ribbons of pixels around the boundaries of the masses. Texture features computed using the ribbons resulted in higher classification accuracy than the same features computed using the corresponding regions within the mass boundaries. The t test was applied to AUC values obtained using 100 repetitions of random splitting of the texture features from the ribbons of masses into the training and testing sets. The texture features computed with the pixel size of 200 μm provided the highest average AUC with statistically highly significant differences as compared to all of the other pixel sizes tested, except 100 μm.
Breast cancer; breast masses; Haralick's texture features; mammography; margins of masses; pixel size; pixel resolution; ribbon around a mass; texture analysis; texture features; tumor classification; digital image processing; image analysis; mammography
We propose a method using Gabor filters and phase portraits to automatically locate the optic nerve head (ONH) in fundus images of the retina. Because the center of the ONH is at or near the focal point of convergence of the retinal vessels, the method includes detection of the vessels using Gabor filters, detection of peaks in the node map obtained via phase portrait analysis, and an intensity-based condition. The method was tested on 40 images from the Digital Retinal Images for Vessel Extraction (DRIVE) database and 81 images from the Structured Analysis of the Retina (STARE) database. An ophthalmologist independently marked the center of the ONH for evaluation of the results. The evaluation of the results includes free-response receiver operating characteristics (FROC) and a measure of distance between the manually marked and detected centers. With the DRIVE database, the centers of the ONH were detected with an average distance of 0.36 mm (18 pixels) to the corresponding centers marked by the ophthalmologist. FROC analysis indicated a sensitivity of 100% at 2.7 false positives per image. With the STARE database, FROC analysis indicated a sensitivity of 88.9% at 4.6 false positives per image.
Retinal images; segmentation; optic nerve head; optic disk; Gabor filters; phase portraits; FROC analysis
Abnormal thinning, thickening, or variation in the thickness of the glomerular basement membrane (GBM) are caused by familial hematuria, diabetes mellitus, and Alport syndrome, respectively. We propose a semi-automated procedure for the segmentation and analysis of the thickness of the GBM in images of renal biopsy samples obtained by using a transmission electron microscope (TEM). The procedure includes the split-and-merge algorithm, morphological image processing, skeletonization, and statistical analysis of the width of the GBM. The procedure was tested with 34 TEM images of six patients. The mean and standard deviation of the GBM width for a patient with normal GBM were estimated to be 368 ± 177 nm, those for a patient with thin GBM associated with familial hematuria were 216 ± 95 nm, and those for a patient with thick GBM due to diabetic nephropathy were 1,094 ± 361 nm. Comparative analysis of the results of image processing with manual measurements by an experienced renal pathologist indicated low error in the range of 12 ± 9 nm.
Glomerular basement membrane; split-and-merge algorithm; skeletonization; segmentation; morphological image processing
Detection of the optic nerve head (ONH) is a key preprocessing component in algorithms for the automatic extraction of the anatomical structures of the retina. We propose a method to automatically locate the ONH in fundus images of the retina. The method includes edge detection using the Sobel operators and detection of circles using the Hough transform. The Hough transform assists in the detection of the center and radius of a circle that approximates the margin of the ONH. Forty images of the retina from the Digital Retinal Images for Vessel Extraction (DRIVE) dataset were used to test the performance of the proposed method. The center and boundary of the ONH were independently marked by an ophthalmologist for evaluation. Free-response receiver operating characteristics (FROC) analysis as well as measures of distance and overlap were used to evaluate the performance of the proposed method. The centers of the ONH were detected with an average distance of 0.36 mm to the corresponding centers marked by the ophthalmologist; the detected circles had an average overlap of 0.73 with the boundaries of the ONH drawn by the ophthalmologist. FROC analysis indicated a sensitivity of detection of 92.5% at 8.9 false-positives per image. With an intensity-based criterion for the selection of the circle and a limit of 40 pixels (0.8 mm) on the distance between the center of the detected circle and the manually identified center of the ONH, a successful detection rate of 90% was obtained with the DRIVE dataset.
Retinal images; fundus images; optic nerve head (ONH); Hough transform; Sobel operators; Canny method; FROC analysis
We propose methods to perform automatic identification of the rib structure, the vertebral column, and the spinal canal in computed tomographic (CT) images of pediatric patients. The segmentation processes for the rib structure and the vertebral column are initiated using multilevel thresholding and the results are refined using morphological image processing techniques with features based on radiological and anatomical prior knowledge. The Hough transform for the detection of circles is applied to a cropped edge map that includes the thoracic vertebral structure. The centers of the detected circles are used to derive the information required for the opening-by-reconstruction algorithm used to segment the spinal canal. The methods were tested on 39 CT exams of 13 patients; the results of segmentation of the vertebral column and the spinal canal were assessed quantitatively and qualitatively by comparing with segmentation performed independently by a radiologist. Using 13 CT exams of six patients, including a total of 458 slices with the vertebra from different sections of the vertebral column, the average Hausdorff distance was determined to be 3.2 mm with a standard deviation (SD) of 2.4 mm; the average mean distance to the closest point (MDCP) was 0.7 mm with SD = 0.6 mm. Quantitative analysis was also performed for the segmented spinal canal with three CT exams of three patients, including 21 slices with the spinal canal from different sections of the vertebral column; the average Hausdorff distance was 1.6 mm with SD = 0.5 mm, and the average MDCP was 0.6 mm with SD = 0.1 mm.
Computed tomographic (CT) images; vertebral column; rib structure; spinal canal; morphological image processing; opening-by-reconstruction; image segmentation; Hough transform; convex hull
Some renal diseases cause changes in the structure of the glomerular basement membranes (GBM). Measurement of the thickness of the GBM can be performed on transmission electron microscopy (TEM) images of renal biopsy samples. Increased thickness of the GBM is observed in patients with diabetic nephropathy. Abnormally thin GBMs are associated with hematuria. We propose image processing methods for the detection and measurement of the GBM. The methods include edge detection, morphological image processing, active contour modeling, skeletonization, and statistical analysis of the width of the GBM. In the present pilot study, the methods were tested with 34 TEM images of six patients. The estimated mean and standard deviation of the GBM width for a patient with normal GBM were 348 ± 135 nm; those for a patient with thin GBMs due to hematuria were 227 ± 94 nm; and those for a patient with diabetic nephropathy were 1,152 ± 411 nm. Comparison with manual measurements by an experienced renal pathologist indicated low error in the range of 36 ± 11 nm.
Glomerular basement membrane; active contours; skeletonization; segmentation
We propose the design of a teaching system named Interpretation and Diagnosis of Mammograms (INDIAM) for training students in the interpretation of mammograms and diagnosis of breast cancer. The proposed system integrates an illustrated tutorial on radiology of the breast, that is, mammography, which uses education techniques to guide the user (doctors, students, or researchers) through various concepts related to the diagnosis of breast cancer. The user can obtain informative text about specific subjects, access a library of bibliographic references, and retrieve cases from a mammographic database that are similar to a query case on hand. The information of each case stored in the mammographic database includes the radiological findings, the clinical history, the lifestyle of the patient, and complementary exams. The breast cancer tutorial is linked to a module that simulates the analysis and diagnosis of a mammogram. The tutorial incorporates tools for helping the user to evaluate his or her knowledge about a specific subject by using the education system or by simulating a diagnosis with appropriate feedback in case of error. The system also makes available digital image processing tools that allow the user to draw the contour of a lesion, the contour of the breast, or identify a cluster of calcifications in a given mammogram. The contours provided by the user are submitted to the system for evaluation. The teaching system is integrated with AMDI—An Indexed Atlas of Digital Mammograms—that includes case studies, e-learning, and research systems. All the resources are accessible via the Web.
Education; medical; computer-assisted instruction; computers in medicine; digital mammography
The last decade witnessed a growing interest in research on content-based image retrieval (CBIR) and related areas. Several systems for managing and retrieving images have been proposed, each one tailored to a specific application. Functionalities commonly available in CBIR systems include: storage and management of complex data, development of feature extractors to support similarity queries, development of index structures to speed up image retrieval, and design and implementation of an intuitive graphical user interface tailored to each application. To facilitate the development of new CBIR systems, we propose an image-handling extension to the relational database management system (RDBMS) PostgreSQL. This extension, called PostgreSQL-IE, is independent of the application and provides the advantage of being open source and portable. The proposed system extends the functionalities of the structured query language SQL with new functions that are able to create new feature extraction procedures, new feature vectors as combinations of previously defined features, and new access methods, as well as to compose similarity queries. PostgreSQL-IE makes available a new image data type, which permits the association of various images with a given unique image attribute. This resource makes it possible to combine visual features of different images in the same feature vector. To validate the concepts and resources available in the proposed extended RDBMS, we propose a CBIR system applied to the analysis of mammograms using PostgreSQL-IE.
Database management systems; digital mammography; digital image management; image database; image retrieval; information storage and retrieval; information system
Segmentation of the internal organs in medical images is a difficult task. By incorporating a priori information regarding specific organs of interest, results of segmentation may be improved. Landmarking (i.e., identifying stable structures to aid in gaining more knowledge concerning contiguous structures) is a promising segmentation method. Specifically, segmentation of the diaphragm may help in limiting the scope of segmentation methods to the abdominal cavity; the diaphragm may also serve as a stable landmark for identifying internal organs, such as the liver, the spleen, and the heart. A method to delineate the diaphragm is proposed in the present work. The method is based upon segmentation of the lungs, identification of the lower surface of the lungs as an initial representation of the diaphragm, and the application of least-squares modeling and deformable contour models to obtain the final segmentation of the diaphragm. The proposed procedure was applied to nine X-ray computed tomographic (CT) exams of four pediatric patients with neuroblastoma. The results were evaluated against the boundaries of the diaphragm as identified independently by a radiologist. Good agreement was observed between the results of segmentation and the reference contours drawn by the radiologist, with an average mean distance to the closest point of 5.85 mm over a total of 73 CT slices including the diaphragm.
Computer-assisted image interpretation; X-ray computed tomography; landmark-based segmentation; diaphragm; deformable contour model; linear least-squares model; active contours
Malignant breast tumors and benign masses appear in mammograms with different shape characteristics: the former usually have rough, spiculated, or microlobulated contours, whereas the latter commonly have smooth, round, oval, or macrolobulated contours. Features that characterize shape roughness and complexity can assist in distinguishing between malignant tumors and benign masses. Signatures of contours may be used to analyze their shapes. We propose to use a signature based on the turning angle function of contours of breast masses to derive features that capture the characteristics of shape roughness as described above. We propose methods to derive an index of the presence of convex regions (XRTA), an index of the presence of concave regions (VRTA), an index of convexity (CXTA), and two measures of fractal dimension (FDTA and FDdTA) from the turning angle function. The methods were tested with a set of 111 contours of 65 benign masses and 46 malignant tumors with different parameters. The best classification accuracies in discriminating between benign masses and malignant tumors, obtained for XRTA, VRTA, CXTA, FDTA, and FDdTA in terms of the area under the receiver operating characteristics curve, were 0.92, 0.92, 0.93, 0.93, and, 0.92, respectively.
Breast cancer; tumor classification; fractal dimension; index of convexity; index of concavity; shape features; turning angle function
Breast masses due to benign disease and malignant tumors related to breast cancer differ in terms of shape, edge-sharpness, and texture characteristics. In this study, we evaluate a set of 22 features including 5 shape factors, 3 edge-sharpness measures, and 14 texture features computed from 111 regions in mammograms, with 46 regions related to malignant tumors and 65 to benign masses. Feature selection is performed by a genetic algorithm based on several criteria, such as alignment of the kernel with the target function, class separability, and normalized distance. Fisher’s linear discriminant analysis, the support vector machine (SVM), and our strict two-surface proximal (S2SP) classifier, as well as their corresponding kernel-based nonlinear versions, are used in the classification task with the selected features. The nonlinear classification performance of kernel Fisher’s discriminant analysis, SVM, and S2SP, with the Gaussian kernel, reached 0.95 in terms of the area under the receiver operating characteristics curve. The results indicate that improvement in classification accuracy may be gained by using selected combinations of shape, edge-sharpness, and texture features.
Breast masses; breast tumors; mammography; computer-aided diagnosis; feature selection; pattern classification; kernel-based classifiers; shape analysis; edge-sharpness analysis; texture analysis
Fractal analysis has been shown to be useful in image processing for characterizing shape and gray-scale complexity. Breast masses present shape and gray-scale characteristics that vary between benign masses and malignant tumors in mammograms. Limited studies have been conducted on the application of fractal analysis specifically for classifying breast masses based on shape. The fractal dimension of the contour of a mass may be computed either directly from the 2-dimensional (2D) contour or from a 1-dimensional (1D) signature derived from the contour. We present a study of four methods to compute the fractal dimension of the contours of breast masses, including the ruler method and the box counting method applied to 1D and 2D representations of the contours. The methods were applied to a data set of 111 contours of breast masses. Receiver operating characteristics (ROC) analysis was performed to assess and compare the performance of fractal dimension and four previously developed shape factors in the classification of breast masses as benign or malignant. Fractal dimension was observed to complement the other shape factors, in particular fractional concavity, in the representation of the complexity of the contours. The combination of fractal dimension with fractional concavity yielded the highest area (Az) under the ROC curve of 0.93; the two measures, on their own, resulted in Az values of 0.89 and 0.88, respectively.
Box counting method; breast cancer; breast masses; breast tumors; contour analysis; fractal analysis; fractal dimension; ruler method; shape analysis; signatures of contours
Segmentation of the tumor in neuroblastoma is complicated by the fact that the mass is almost always heterogeneous in nature; furthermore, viable tumor, necrosis, and normal tissue are often intermixed. Tumor definition and diagnosis require the analysis of the spatial distribution and Hounsfield unit (HU) values of voxels in computed tomography (CT) images, coupled with a knowledge of normal anatomy. Segmentation and analysis of the tissue composition of the tumor can assist in quantitative assessment of the response to therapy and in the planning of delayed surgery for resection of the tumor. We propose methods to achieve 3-dimensional segmentation of the neuroblastic tumor. In our scheme, some of the normal structures expected in abdominal CT images are delineated and removed from further consideration; the remaining parts of the image volume are then examined for the tumor mass. Mathematical morphology, fuzzy connectivity, and other image processing tools are deployed for this purpose. Expert knowledge provided by a radiologist in the form of the expected structures and their shapes, HU values, and radiological characteristics are incorporated into the segmentation algorithm. In this preliminary study, the methods were tested with 10 CT exams of four cases from the Alberta Children's Hospital. False-negative error rates of less than 12% were obtained in eight of the 10 exams; however, seven of the exams had false-positive error rates of more than 20% with respect to manual segmentation of the tumor by a radiologist.
3D image segmentation; neuroblastoma; computed tomography; fuzzy connectivity; tumor segmentation