Jayasree Chakraborty

Automatic Detection of Pectoral Muscle Using Average Gradient and Shape Based Feature

In medio-lateral oblique view of mammogram, pectoral muscle may sometimes affect the detection of breast cancer due to their similar characteristics with abnormal tissues. As a result pectoral muscle should be handled separately while detecting the breast cancer. In this paper, a novel approach for the detection of pectoral muscle using average gradient- and shape-based feature is proposed. The process first approximates the pectoral muscle boundary as a straight line using average gradient-, position-, and shape-based features of the pectoral muscle. Straight line is then tuned to a smooth curve which represents the pectoral margin more accurately. Finally, an enclosed region is generated which represents the pectoral muscle as a segmentation mask. The main advantage of the method is its’ simplicity as well as accuracy. The method is applied on 200 mammographic images consisting 80 randomly selected scanned film images from Mammographic Image Analysis Society (mini-MIAS) database, 80 direct radiography (DR) images, and 40 computed radiography (CR) images from local database. The performance is evaluated based upon the false positive (FP), false negative (FN) pixel percentage, and mean distance closest point (MDCP). Taking all the images into consideration, the average FP and FN pixel percentages are 4.22%, 3.93%, 18.81%, and 6.71%, 6.28%, 5.12% for mini-MIAS, DR, and CR images, respectively. Obtained MDCP values for the same set of database are 3.34, 3.33, and 10.41 respectively. The method is also compared with two well-known pectoral muscle detection techniques and in most of the cases, it outperforms the other two approaches.

Statistical measures of orientation of texture for the detection of architectural distortion in prior mammograms of interval-cancer

Abstract. Architectural distortion is an important sign of early breast cancer. Due to its subtlety, it is often missed during screening. We propose a method to detect architectural distortion in prior mammograms of interval-cancer cases based on statistical measures of oriented patterns. Oriented patterns were analyzed in the present work because regions with architectural distortion contain a large number of tissue structures spread over a wide angular range. Two new types of cooccurrence matrices were derived to estimate the joint occurrence of the angles of oriented structures. Statistical features were computed from each of the angle cooccurrence matrices to discriminate sites of architectural distortion from falsely detected regions in normal parts of mammograms. A total of 4,224 regions of interest (ROIs) were automatically obtained from 106 prior mammograms of 56 interval-cancer cases and 52 mammograms of 13 normal cases with the application of Gabor filters and phase portrait analysis. For each ROI, Haralick’s 14 features were computed using the angle cooccurrence matrices. The best result obtained in terms of the area under the receiver operating characteristic (ROC) curve with the leave-one-patient-out method was 0.76; the free-response ROC curve indicated a sensitivity of 80% at 4.2 false positives per patient.

Measures of divergence of oriented patterns for the detection of architectural distortion in prior mammograms

AbstractPurpose We propose a method for the detection of architectural distortion in prior mammograms of interval-cancer cases based on the expected orientation of breast tissue patterns in mammograms. Methods The expected orientation of the breast tissue at each pixel was derived by using automatically detected landmarks including the breast boundary, the nipple, and the pectoral muscle (in mediolateral-oblique views). We hypothesize that the presence of architectural distortion changes the normal expected orientation of breast tissue patterns in a mammographic image. The angular deviation of the oriented structures in a given mammogram as compared to the expected orientation was analyzed to detect potential sites of architectural distortion using a measure of divergence of oriented patterns. Each potential site of architectural distortion was then characterized using measures of spicularity and angular dispersion specifically designed to represent spiculating patterns. The novel features for the characterization of spiculating patterns include an index of divergence of spicules computed from the intensity image and Gabor magnitude response using the Gabor angle response; radially weighted difference and angle-weighted difference (AWD) measures of the intensity, Gabor magnitude, and Gabor angle response; and AWD in the entropy of spicules computed from the intensity, Gabor magnitude, and Gabor angle response. Results Using the newly proposed features with a database of 106 prior mammograms of 56 interval-cancer cases and 52 mammograms of 13 normal cases, through feature selection and pattern classification with an artificial neural network, an area under the receiver operating characteristic curve of 0.75 was obtained. Free-response receiver operating characteristic analysis indicated a sensitivity of 0.80 at 5.3 false positives (FPs) per patient. Combining the features proposed in the present paper with others described in our previous works led to significant improvement with a sensitivity of 0.80 at 3.7 FPs per patient. Conclusion The proposed methods can detect architectural distortion in prior mammograms taken 15 months (on the average) before clinical diagnosis of breast cancer, but the FP rate needs to be reduced.

Automatic characterization of masses in mammograms

The classification of benign and malignant masses in digital mammogram is an important yet challenging step for the early detection of breast cancer. This paper presents statistical measures of the orientation of texture to classify malignant and benign masses. Since the presence of mass in mammogram may change the orientation of normal breast tissues, two types of co-occurrence matrices are derived to estimate the joint occurrence of the angles of oriented structures for characterizing them. Haralicks 14 features are then extracted from each of the matrices derived from different regions related to mass. A total of 444 mass regions from 434 scanned-film images of the DDSM database are selected to evaluate the performance of the proposed features to differentiate the masses. The features are also compared with Haralicks features, obtained from well-known gray-level co-occurrence matrix. The best Az value of 0.77 is achieved with the stepwise logistic regression method for feature selection, an Fisher linear discriminant analysis for classification, and the leave-one-ROI-out approach for cross validation.

Classification of benign and malignant masses in mammograms using multi-resolution analysis of oriented patterns

The paper proposes a novel approach for the classification of breast masses as benign and malignant using multi-resolution analysis of oriented patterns of tissues in mammograms. Since, the oriented structures of normal breast near the mass region may be changed in presence of masses, three regions are defined, first, for the analysis. Statistical features are then extracted using two angle co-occurrence matrices derived at different resolution levels of each region with Haar-wavelet transform to quantify the joint occurrences of different angle pairs of oriented patterns. The experiments show best classification accuracy of 81.23% and area under the receiver operating characteristic curve of 0.86 with 433 images from the DDSM database using artificial neural network and tenfold cross-validation method.

Detection of architectural distortion in prior mammograms using statistical measures of orientation of texture

We present a method using statistical measures of the orientation of texture to characterize and detect architectural distortion in prior mammograms of interval-cancer cases. Based on the orientation field, obtained by the application of a bank of Gabor filters to mammographic images, two types of co-occurrence matrices were derived to estimate the joint occurrence of the angles of oriented structures. For each of the matrices, Haralicks 14 texture features were computed. From a total of 106 prior mammograms of 56 interval-cancer cases and 52 mammograms of 13 normal cases, 4,224 regions of interest (ROIs) were automatically obtained by applying Gabor filters and phase portrait analysis. For each ROI, statistical features were computed using the angle co-occurrence matrices. The performance of the features in the detection of architectural distortion was analyzed and compared with that of Haralicks features computed using the gray-level co-occurrence matrices of the ROIs. Using logistic regression for feature selection, an artificial neural network for classification, and the leave-one-image-out approach for cross-validation, the best result achieved was 0.77 in terms of the area under the receiver operating characteristic (ROC) curve. Analysis of the free-response ROC curve yielded a sensitivity of 80% at 5.4 false positives per image.

Detection of masses in mammograms using region growing controlled by multilevel thresholding

Mammographic masses are important signs of breast cancer. However, due to its irregular and obscured margin, variability in size, and occlusion within dense breast tissue, a mass is often difficult to detect. In this paper, a multilevel thresholding approach controlled by gradient and intensity is proposed, where an image is considered as a 3D topographic map with intensity as the third dimension. A multilevel high-to-low intensity thresholding approach is used for the detection of the focal region of a mass. After each step of thresholding, a region growing technique is applied on each focal region to detect potential sites of masses using gradient and intensity information. The performance of the proposed method is tested on 107 scanned-film images consisting of 52 images with masses and 55 normal images from the mini-MIAS database, and 158 digital radiography (DR) images containing 78 images with masses and 80 normal images from a local database. A sensitivity of 95% with 5.2 false positives (FPs)/image is obtained with the mini-MIAS images and a sensitivity of 98.8% with 2.1 FPs/image is obtained with the DR images.

Benign-malignant mass classification in mammogram using edge weighted local texture features

This paper introduces novel Discriminative Robust Local Binary Pattern (DRLBP) and Discriminative Robust Local Ternary Pattern (DRLTP) for the classification of mammographic masses as benign or malignant. Mass is one of the common, however, challenging evidence of breast cancer in mammography and diagnosis of masses is a difficult task. Since DRLBP and DRLTP overcome the drawbacks of Local Binary Pattern (LBP) and Local Ternary Pattern (LTP) by discriminating a brighter object against the dark background and vice-versa, in addition to the preservation of the edge information along with the texture information, several edge-preserving texture features are extracted, in this study, from DRLBP and DRLTP. Finally, a Fisher Linear Discriminant Analysis method is incorporated with discriminating features, selected by stepwise logistic regression method, for the classification of benign and malignant masses. The performance characteristics of DRLBP and DRLTP features are evaluated using a ten-fold cross-validation technique with 58 masses from the mini-MIAS database, and the best result is observed with DRLBP having an area under the receiver operating characteristic curve of 0.982.

Face detection using skin color modeling and geometric feature

The paper proposes a robust face detection technique based on skin color clustering and geometric feature. Skin regions are extracted using Gaussian skin color model in Cb-Cr space and likelihood ratio method is used to create a binary mask. Further, morphological operations and adaptive thresholding is applied to group the potential face regions and remove noisy pixels. The region thus extracted is operated by an edge filter to find the edges inside it. Finally, best ellipse searching is used to detect the frontal face outline. The skin color model was designed using a combination of two different databases, to encompass larger skin hues. Later, a total of 165 facial images from Caltech database were randomly selected to evaluate the performance of the proposed method and an accuracy of 95% was obtained.

Computer-aided detection and diagnosis of mammographic masses using multi-resolution analysis of oriented tissue patterns

Abstract In this article, a novel approach is proposed for automatic detection and diagnosis of mammographic masses, one of the common signs of non-palpable breast cancer. However, detection and diagnosis of mass are difficult due to its irregular shape, variability in size, and occlusion within breast tissue. The main aim of this study is to classify masses into benign and malignant after detecting them automatically. We propose an iterative method of high-to-low intensity thresholding controlled by radial region growing for the detection of masses. Based on the observation that in presence of mass orientation of tissue patterns changes, which may differ from benign to malignant, a multi resolution analysis of orientation of tissue patterns is then performed to categorize them. The performance of the proposed algorithm is evaluated with images from the digital database for screening mammography (DDSM), containing 450 benign masses, 440 malignant masses, and 410 normal images. A sensitivity of 85.0% is achieved at 1.4 false positives per image in mass detection, whereas an area under the receiver operating characteristic curve of 0.92 with an accuracy of 83.30% is achieved for the diagnosis of malignant masses.