Rodrigo Pereira Ramos

Segmentation and detection of breast cancer in mammograms combining wavelet analysis and genetic algorithm

In Brazil, the National Cancer Institute (INCA) reports more than 50,000 new cases of the disease, with risk of 51 cases per 100,000 women. Radiographic images obtained from mammography equipments are one of the most frequently used techniques for helping in early diagnosis. Due to factors related to cost and professional experience, in the last two decades computer systems to support detection (Computer-Aided Detection - CADe) and diagnosis (Computer-Aided Diagnosis - CADx) have been developed in order to assist experts in detection of abnormalities in their initial stages. Despite the large number of researches on CADe and CADx systems, there is still a need for improved computerized methods. Nowadays, there is a growing concern with the sensitivity and reliability of abnormalities diagnosis in both views of breast mammographic images, namely cranio-caudal (CC) and medio-lateral oblique (MLO). This paper presents a set of computational tools to aid segmentation and detection of mammograms that contained mass or masses in CC and MLO views. An artifact removal algorithm is first implemented followed by an image denoising and gray-level enhancement method based on wavelet transform and Wiener filter. Finally, a method for detection and segmentation of masses using multiple thresholding, wavelet transform and genetic algorithm is employed in mammograms which were randomly selected from the Digital Database for Screening Mammography (DDSM). The developed computer method was quantitatively evaluated using the area overlap metric (AOM). The mean ± standard deviation value of AOM for the proposed method was 79.2 ± 8%. The experiments demonstrate that the proposed method has a strong potential to be used as the basis for mammogram mass segmentation in CC and MLO views. Another important aspect is that the method overcomes the limitation of analyzing only CC and MLO views.

Texture extraction: An evaluation of ridgelet, wavelet and co-occurrence based methods applied to mammograms

Image processing algorithms can be used in computer-aided diagnosis systems to extract features directly from digitized mammograms. Typically, two classes of features are extracted from mammograms with these algorithms, namely morphological and non-morphological features. Image texture analysis is an important technique that represents gray level properties of images used to describe non-morphological features. This technique has shown to be a promising technique in analyzing mammographic lesions caused by masses. In this paper, we evaluate texture classification using features derived from co-occurrence matrices, wavelet and ridgelet transforms of mammographic images. In particular, we propose a false positive reduction in computer-aided detection of masses. The data set consisted of 120 cranio-caudal mammograms, half containing a mass, rated as abnormal images, and half with no lesions. The following texture descriptors were then calculated to analyze the regions of interest (ROIs) texture patterns: entropy, energy, sum average, sum variance, and cluster tendency. To select the best set of features for each method, we applied a genetic algorithm (GA). In the ROIs classification stage, we used the Random Forest algorithm, a data mining technique that separates the data into non-overlapping segments. Experimental results showed that the best classification rates were obtained with the wavelet-based feature extraction using GA for selection of the most relevant features, giving an AUC=0.90.

Classification of masses in mammographic image using wavelet domain features and polynomial classifier

Breast cancer is the most common cancer among women. In CAD systems, several studies have investigated the use of wavelet transform as a multiresolution analysis tool for texture analysis and could be interpreted as inputs to a classifier. In classification, polynomial classifier has been used due to the advantages of providing only one model for optimal separation of classes and to consider this as the solution of the problem. In this paper, a system is proposed for texture analysis and classification of lesions in mammographic images. Multiresolution analysis features were extracted from the region of interest of a given image. These features were computed based on three different wavelet functions, Daubechies 8, Symlet 8 and bi-orthogonal 3.7. For classification, we used the polynomial classification algorithm to define the mammogram images as normal or abnormal. We also made a comparison with other artificial intelligence algorithms (Decision Tree, SVM, K-NN). A Receiver Operating Characteristics (ROC) curve is used to evaluate the performance of the proposed system. Our system is evaluated using 360 digitized mammograms from DDSM database and the result shows that the algorithm has an area under the ROC curve Az of 0.98+/-0.03. The performance of the polynomial classifier has proved to be better in comparison to other classification algorithms.

LBP operators on curvelet coefficients as an algorithm to describe texture in breast cancer tissues

We present a method based on curvelet transform, LBP, ANOVA and PL classifier.We validate the proposed approach considering the metrics accuracy and AUC.The features was evaluated by applying the DT, RaF, SVM and PL classifiers.The proposed approach achieved AC values among 91% and 100%.The method was tested on the datasets: DDSM, BCDR-FMR, BCDR-DMR and UCSB-BB. In computer-aided diagnosis one of the crucial steps to classify suspicious lesions is the extraction of features. Texture analysis methods have been used in the analysis and interpretation of medical images. In this work we present a method based on the association among curvelet transform, local binary patterns, feature selection by statistical analysis and distinct classification methods, in order to support the development of computer aided diagnosis system. The similar features were removed by the statistical analysis of variance (ANOVA). The understanding of the features was evaluated by applying the decision tree, random forest, support vector machine and polynomial (PL) classifiers, considering the metrics accuracy (AC) and area under the ROC curve (AUC): the rates were calculated on images of breast tissues with different physical properties (commonly observed in clinical practice). The datasets were the Digital Database for Screening Mammography, Breast Cancer Digital Repository and UCSB biosegmentation benchmark. The investigated groups were normal-abnormal and benign-malignant. The association of curvelet transform, local binary pattern and ANOVA with the PL classifier achieved higher AUC and AC values for all cases: the obtained rates were among 91% and 100%. These results are relevant, specially when we consider the difficulties of clinical practice in distinguishing the studied groups. The proposed association is useful as an automated protocol for the diagnosis of breast tissues and may contribute to the diagnosis of breast tissues (mammographic and histopathological images).

Automatic Detection of Breast Masses Using Two-View Mammography

Computer-Aided Detection (CADe) and Computer- Aided Diagnosis (CADx) systems have been developed to increase the possible abnormalities diagnostic performance of radiologists in the initial stage. However, several computerized systems analyze the MedioLateral Oblique (MLO) view and CranioCaudal (CC) views independently or using only one of the views. A group of tools for computer-aided detection of lesions caused by breast masses in the two above projections are presented in this paper. A preprocessing stage was applied to enhance the breast information. Moreover, a method for automatic segmentation of lesions combining genetic algorithm and wavelet transform for multilevel threshold was proposed. Through preliminary tests, the method seems to meaningfully improve the diagnosis in the early breast cancer detection with multi-views. It was found a difference about 11% between the boundary regions identified by the proposed approach and that obtained by compared the chain-code available in the base DDSM (Digital Database for Screening Mammography).

Comparison of PCA and ANOVA for information selection of CC and MLO views in classification of mammograms

In this paper, we present a method for extraction and attribute selection for textural features classification using the fusion of information from the mediolateral oblique (MLO) view and craniocaudal (CC) views. In the extraction step, wavelet coefficients together with singular value decomposition technique were applied to reduce the number of textural attributes. For the selection stage and reduction of attributes, an evaluation of the Analysis of Variance (ANOVA) technique and Principal Component Analysis (PCA) is performed when used for textural information reduction. In the final step, it was used the Random Forest algorithm for classifying regions of interest (ROIs) of the set of images determined as normal, benign and malignant. The experiments showed that ANOVA reached the higher proportional attributes reduction and featured the best results for information fusion of CC and MLO views. The best classification rates were obtained with ANOVA for normal-benign images (area under the receiver operating characteristic curve - AUC = 0.78) and benign-malignant images (AUC = 0.83) and with the PCA method for normal-malignant images (AUC = 0.85).

Fusion of Two-View Information: SVD Based Modeling for Computerized Classification of Breast Lesions on Mammograms

Over the past few years, the cancer has been one of the most responsible for the high number of deaths, and could become one of the main responsible for most deaths in the next decades. According to the World Health Organization, the number of deaths due to cancer, which was just 13% in 2008, is currently having a significant increase and one estimates that this number could reach approximately 12 million until 2030 (Tang et al., 2009).

MÉTODO SVD: ANÁLISE WAVELETS-MÃE NA EXTRAÇÃO DE TEXTURA APLICADAS EM MAMOGRAMAS

Resumo: A análise de texturas é uma importante etapa empregada nos sistemas CAD de imagens mamográficas. Este trabalho apresenta um estudo comparativo entre diferentes tipos de wavelets-mãe para a extração de características de texturas por meio do método SVD. O método ANOVA foi aplicado para redução no número de atributos gerados pela técnica SVD. Em seguida, o classificador SVM foi aplicado para avaliar o desempenho das diferentes funções waveletsmãe. Os experimentos realizados mostraram que a waveletsmãe Daubechies 4 proporcionou os melhores resultados com uma área ROC de 0,83. O método de redução baseado em SVD e ANOVA proporcionou uma redução de 71,6% do número de atributos utilizados na etapa de classificação do sistema. A base de imagens DDSM foi utilizada para avaliar os métodos propostos.