Deep Multiple Instance Learning for Automatic Breast Cancer Assessment Using Digital Mammography

In this paper we will present a computer-aided detection (CAD) system for mass detection and classification. This CAD system performs mass detection on regions of interest (ROIs) and performs the normal-abnormal classification on detected masses. To enhance the ROIs, histogram equalization and average filtering were employed. ROIs or suspicious lesions were detected by an edge-base segmentation technique. Each ROI was represented by 12 textural features calculated from spatial gray-level co-occurrence matrix (GLCM). Finally, a back propagation neural network was used to classify a ROI to normal or abnormal one. Performance of the proposed system was analyzed in mini-MIAS data set by means of receiver operating characteristics curve (ROC curve) and free-response ROC curve (FROC curve). We archived area under the ROC curve AZ = 0,815. This value lied in the range from 0.7 to 0.9 so that our CAD system could be considered accurate enough.

Computer-Aided Mass Detection Based on Ipsilateral Multiview Mammograms

Digital CT Scan images are exchanged among the medical practitioners for consultative purposes or discussing diagnostic and therapeutic measures. This exchange of medical images imposes two main constraints for medical information: (i) the Content Authentication and (ii) Copyright protection. To address these two issues, we have proposed a technique of watermarking for medical images which embeds two watermarks in the CT scan image. The binary pattern is embedded in region of interest (ROI) for content authentication and the composite watermark is embedded in the region of non interest (RONI) for copyright protection. Initially CT Scan image is divided in ROI and RONI by segmentation process. The segmented ROI contains some holes which are first filled. The image morphology is then used and closing operation is applied on the resultant image which eliminates small holes and fills gaps in the contour. Thus binary mask is obtained which is used to separate ROI from RONI. Before embedding the binary pattern in ROI for data authentication purpose, the Least Significant Bit (LSB) information of ROI is extracted first and is saved in separate store. Also, before embedding the composite watermark, pixels in RONI are scrambled for providing the further security. The payload in both ROI and RONI is embedded with simple LSB substitution technique. The extraction procedure is the reverse of embedding process. However few steps of extraction procedure are same as embedding procedure. We have used CT scan image database obtained from Radiology Department, Ackron University Ohio USA. The experimental results show that the proposed algorithm provides better security of medical information during transmission by addressing both issues of data authentication and copyright protection.KeywordsMedical image watermarkingLeast significant bit techniqueImage morphologyImage segmentationData authenticationCopyright protectionRegion of interest (ROI)Region of non interest (RONI)

Mass localization is important in computer-aided detection (CAD) system for the diagnosis of suspicious regions in mammograms. In this paper, we present an automated classification system for the detection of masses in mammographic images. Suspicious regions are located with an adaptive region growing firstly. Then, the initial regions are further refined with narrow band based active contour, which can improve the segmentation accuracy of masses. CLBP (Complete Local Binary Pattern) texture features are extracted from the ROIs (regions of interest) containing the segmented suspicious regions. Finally, the ROIs are classified by means of support vector machine (SVM), with supervision provided by the radiologist's diagnosis. The method was evaluated on a dataset with 231 images, containing 245 masses. Among them, 125 images containing 133 masses are used to optimize the parameters and are used to train SVM. The remaining 106 images are used to test the performance. It obtained 1.36 FPsI at the sensitivity 76.8%. It shows that the proposed method is a promising approach to achieve low FPsI while maintain a high sensitivity.

To compare the influence of two different regions of interest (ROIs) on diffusion tensor metrics in dystrophic thigh muscles using acustom-made (whole muscle) ROI including and aselective ROI excluding areas of fatty replacement. Diffusion tensor imaging (DTI) and chemical-shift-encoded water-fat magnetic resonance imaging (MRI) of the thigh was conducted on a3-Tesla system in 15cases with muscular dystrophy and controls. The ROIs were chosen according to patterns of fatty replacement on co-registered axial DTI and gradient echo sequence (GRE) images. Fractional anisotropy (FA), apparent diffusion coefficient (ADC), fiber track length (FTL), and muscle fat fractions (MFF) were compared between both ROI segmentations. These comparisons, muscle-specific correlation coefficients, and the influence of ROI localization on tensor metrics were derived based on linear mixed effects regression models. In the cases ahigh correlation was observed for ADC and FA with MFF using acustom ROI. The correlation was weaker but still significant with aselective ROI method. Using the custom ROI, FTL correlated significantly with MFF in 3 out of 4muscles (r ≤ -0.51). Acorrelation was not found for the selective ROI method. Interaction analysis revealed that the association of ADC and FA with MFF was not significantly influenced by the ROI localization. For FTL the ROI localization significantly reduced the negative association with MFF. The DTI metrics and FTL of custom ROI segmentation are significantly influenced by MFF. Contrary to ADC and FA, the effect of MFF on FTL is significantly reduced when applying selective ROI segmentation, which could therefore be abetter option for MR tractography.

In this paper, we propose a Computer-Aided Detection (CAD) system based on the novel Multi-Classifier Fusion-based Classification Model (MCFM) of the breast lesions using textural features in Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI). The proposed system consists of four principal phases: Region Of Interest (ROI) segmentation, feature extraction, fusion and selection and finally Classification. We have extracted a complementary set of features from the DCE-MRI database on two statistical descriptors such as the Gray Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP) methods. Then, we have concatenated the extracted features using Serial method to exploit the complementary between these sets. Next, we have applied a features selection phase through Principal Component Analysis (PCA) and Feature Subset Selection (FSS) techniques to remove redundant information and improve the overall performance of breast DCE-MRI in a computer-aided system. For the classification, we have used a set of classifiers (K-Nearest Neighbors (K-NN), Support Vector Machines (SVM) and Multilayer Perceptron-Artificial Neural Network (MLP-ANN)). Then, a fusion Model between them using several metrics is applied. Experimental validation is performed over an MRI dataset, which is comprised of 286 patients with 143 Masses and 143 No Masses (Normal) breast tissues. These experiments report encouraging performances for mass detection, achieved using ten-fold cross-validation model, in terms of the averaged measurements of different metrics such as Sensitivity, Specificity, Area Under Curve (AUC) and Accuracy, which are 0.8972, 0.9572, 0.9630 and 0.9501, respectively. The proposed DCE-MRI analysis model can offer an efficient system for radiologists to identify the breast tissues.

An automated computer-aided detection (CAD) system is being proposed for identification of lung nodules present in computed tomography (CT) images. This system is capable of identifying the region of interest (ROI) and extracting the features from the ROI. Feature vectors are generated from the gray-level covariance matrix using the statistical properties of the matrix. The relevant features are identified by adopting principle component analysis algorithm on the feature space (the space formed from the feature vectors). Support vector machine and fuzzy C-means algorithms are used for classifying nodules. Annotated images are used to validate the results. Efficiency and reliability of the system are evaluated visually and numerically using relevant measures. Developed CAD system is found to identify nodules with high accuracy.

The objective of this study was to evaluate the performance of a computer-aided detection (CAD) system for the detection of breast cancer, based on mammographic appearance and histopathology. From 1000 consecutive screening mammograms from women with biopsy-proven breast carcinoma, 273 mammograms were selected randomly for retrospective evaluation by CAD. The sensitivity of the CAD system for breast cancer was assessed from the proportion of masses and microcalcifications detected. The corresponding tumor histopathologies also were evaluated. Normal mammograms (n = 155 patients) were used to determine the false-positive rate of the system. Of the 273 breast carcinomas, 149 appeared mammographically as masses, and 88 appeared as microcalcifications, including 36 carcinomas that presented as mixed lesions. The CAD system marked 125 of 149 masses correctly (84%), marked 86 of 88 microcalcifications correctly (98%), and marked 32 of 36 of mixed lesions correctly (89%.). The system showed a high sensitivity for the detection of ductal carcinoma in situ (95%; 73 of 77 lesions), invasive lobular carcinoma (95%; 18 of 19 lesions), invasive ductal carcinoma (85%; 125 of 147 lesions), and invasive mammary carcinoma (90%; 27 of 30 lesions). The highest CAD system sensitivity was for all invasive carcinomas that presented as microcalcifications (100%). On normal mammograms, there was an average of 1.3 false-positive CAD marks per image. The CAD system correctly marked a large majority of biopsy-proven breast cancers, with a greater sensitivity for lesions with microcalcifications and without significant impact of performance based on tumor histopathology. CAD was highly effective in detecting invasive lobular carcinoma (sensitivity, 95%) and ductal carcinoma in situ (sensitivity, 95%). CAD represents a useful tool for the detection of breast cancer.

Socioeconomic inequalities in breast and thyroid cancer screening in Korea: A nationwide cross-sectional study

To study the clinical value of computer aided detection (CAD) in direct full-field digital mammography. 480 digital mammograms of 120 patients with pathologically proved breast cancer were randomly selected and interpreted by an expert radiologist without knowing the pathologic diagnosis, then analyzed by CAD system, and finally read by the radiologist again to make the final diagnosis. The sensitivity for breast cancer detection of the radiologist in diagnosis of breast cancer was calculated before and after using CAD. Data of the breast density, diameters of mass and calcification, being spiculate or not, and pathology of each case were reviewed to determine how they impacted on the improvement of sensitivity with the assistance of CAD. The sensitivity for breast cancer detection by the radiologist before using CAD was 74.2%, significantly lower than that after using CAD (89.2%, chi(2) = 37.628, P = 0.000). Breast density (chi(2) = 7.352, P = 0.007), diameter of calcification (chi(2) = 10.549, P = 0.014) and pathologic pattern (chi(2) = 13.751, P = 0.003) were highly correlated with the improvement of sensitivity. Use of CAD might overcome type 1 missed diagnosis, and not type 2 missed diagnosis. Computer-aided detection helps radiologist improve their diagnostic accuracy of breast cancer, however, is not capable of differentiating benign from malignant lesions.

Effect of CAD on Radiologists' Detection of Lung Nodules on Thoracic CT Scans: Analysis of an Observer Performance Study by Nodule Size

Mammographic percent density (PD%) is known to be a strong risk factor for breast cancer. Recent studies also suggest that parenchymal texture features, which are more granular descriptors of the parenchymal pattern, can provide additional information about breast cancer risk. To date, most studies have measured mammographic texture within selected regions of interest (ROIs) in the breast, which cannot adequately capture the complexity of the parenchymal pattern throughout the whole breast. To better characterize patterns of the parenchymal tissue, the authors have developed a fully automated software pipeline based on a novel lattice-based strategy to extract a range of parenchymal texture features from the entire breast region. Digital mammograms from 106 cases with 318 age-matched controls were retrospectively analyzed. The lattice-based approach is based on a regular grid virtually overlaid on each mammographic image. Texture features are computed from the intersection (i.e., lattice) points of the grid lines within the breast, using a local window centered at each lattice point. Using this strategy, a range of statistical (gray-level histogram, co-occurrence, and run-length) and structural (edge-enhancing, local binary pattern, and fractal dimension) features are extracted. To cover the entire breast, the size of the local window for feature extraction is set equal to the lattice grid spacing and optimized experimentally by evaluating different windows sizes. The association between their lattice-based texture features and breast cancer was evaluated using logistic regression with leave-one-out cross validation and further compared to that of breast PD% and commonly used single-ROI texture features extracted from the retroareolar or the central breast region. Classification performance was evaluated using the area under the curve (AUC) of the receiver operating characteristic (ROC). DeLong's test was used to compare the different ROCs in terms of AUC performance. The average univariate performance of the lattice-based features is higher when extracted from smaller than larger window sizes. While not every individual texture feature is superior to breast PD% (AUC: 0.59, STD: 0.03), their combination in multivariate analysis has significantly better performance (AUC: 0.85, STD: 0.02, p < 0.001). The lattice-based texture features also outperform the single-ROI texture features when extracted from the retroareolar or the central breast region (AUC: 0.60-0.74, STD: 0.03). Adding breast PD% does not make a significant performance improvement to the lattice-based texture features or the single-ROI features (p > 0.05). The proposed lattice-based strategy for mammographic texture analysis enables to characterize the parenchymal pattern over the entire breast. As such, these features provide richer information compared to currently used descriptors and may ultimately improve breast cancer risk assessment. Larger studies are warranted to validate these findings and also compare to standard demographic and reproductive risk factors.

For the first time, full-field digital mammography (FFDM) allows computer-aided detection (CAD) analysis of directly acquired digital image data. The purpose of this study was to evaluate a CAD system in patients with histologically correlated breast cancer depicted with FFDM. Sixty-three cases of histologically proven breast cancer detected with FFDM (Senographe 2000D, GE Medical Systems, Buc, France) were analyzed using a CAD system (Image Checker V2.3, R2 Technology, Los Altos, Calif.). Fourteen of these malignancies were characterized as microcalcifications, 37 as masses, and 12 as both. The mammographic findings were categorized as BI-RADS 3 (n=5), BI-RADS 4 (n=17) and BI-RADS 5 (n=40). The sensitivity for malignant lesions and the rate of false-positive marks per image were calculated. The sensitivity and its 95% confidence interval (CI) were estimated. The sensitivity of the CAD R2 system in breast cancer seen on FFDM was 89% for microcalcifications [CI(95%)=(70%; 98%)] and 81% for masses [CI(95%)=(67%; 91%)]. As expected, the detection rate was higher in lesions categorized as BI-RADS 5 (37 of 40) compared with lesions categorized as BI-RADS 4 (11 of 17). In the group categorized as BI-RADS 3 the detection rate was 4 of 5 lesions; however, this group was very small. The rate of false-positive marks was 0.35 microcalcification marks/image and 0.26 mass marks/image. The overall rate of false-positive marks was 0.61 per image. CAD based on FFDM provides an optimized work flow. Results are equivalent to the results reported for CAD analysis of secondarily digitized image data. Sensitivity for microcalcifications is acceptable and for masses is low. The number of false-positive marks per image should be reduced.

For the radiologists, it is difficult to identify the mass on a mammogram since the masses are surrounded by the mammary gland and blood vessel. In current breast cancer screening, about 10 by radiologists owing to the ambiguous margins of lesions and the visual fatigue of radiologists resulting from the long-time diagnosis. For these reasons, many computer-aided detection (CADe) systems have been developed to aid radiologists in detecting mammographic lesions that may indicate the presence of breast cancer. The purpose of this study is to construct an automated CADe system using a new feature extraction method for mammographic mass detection. In this system, some adaptive square regions of interest (ROIs) are segmented according to the size of suspicious areas. Then a new feature extraction method adopting grey level co-occurrence matrix and optical density features called GLCM-OD features is applied to each ROI. The GLCM-OD features describe local grey level texture characteristics and the whole photometric distribution of the ROI. Finally, the stepwise linear discriminant analysis is applied to classify abnormal regions by selecting and rating individual performance of each feature. This system is trained and tested by 358 mammographic cases from the digital database of screening mammography (DDSM). The proposed system averagely provides sensitivity of 97.3% with 4.9 false positives per image and the Az is 0.981. The results prove that the proposed system achieves satisfactory detection performance.Keywordscomputer-aided detection systemgrey level co-occurrence matrixoptical densityGLCM-OD features

Artificial intelligence in breast imaging