Retrieval boosted computer-aided diagnosis of clustered microcalcifications for breast cancer

Instance-Based Cost-Sensitive Boosting

Many classification algorithms aim to minimize just their training error count; however, it is often desirable to minimize a more general cost metric, where distinct instances have different costs. In this paper, an instance-based cost-sensitive Bayesian consistent version of exponential loss function is proposed. Using the modified loss function, the derivation of instance-based cost-sensitive extensions of AdaBoost, RealBoost and GentleBoost are developed which are termed as ICSAdaBoost, ICSRealBoost and ICSGentleBoost, respectively. In this research, a new instance-based cost generation method is proposed instead of doing this expensive process by experts. Thus, each sample takes two cost values; a class cost and a sample cost. The first cost is equally assigned to all samples of each class while the second cost is generated according to the probability of each sample within its class probability density function. Experimental results of the proposed schemes imply 12% enhancement in terms of [Formula: see text]-measure and 13% on cost-per-sample over a variety of UCI datasets, compared to the state-of-the-art methods. The significant priority of the proposed method is supported by applying the pair of [Formula: see text]-tests to the results.

Machine learning techniques for breast cancer computer aided diagnosis using different image modalities: A systematic review

BACKGROUND AND OBJECTIVE

The high incidence of breast cancer in women has increased significantly in the recent years. Physician experience of diagnosing and detecting breast cancer can be assisted by using some computerized features extraction and classification algorithms. This paper presents the conduction and results of a systematic review (SR) that aims to investigate the state of the art regarding the computer aided diagnosis/detection (CAD) systems for breast cancer.

METHODS

The SR was conducted using a comprehensive selection of scientific databases as reference sources, allowing access to diverse publications in the field. The scientific databases used are Springer Link (SL), Science Direct (SD), IEEE Xplore Digital Library, and PubMed. Inclusion and exclusion criteria were defined and applied to each retrieved work to select those of interest. From 320 studies retrieved, 154 studies were included. However, the scope of this research is limited to scientific and academic works and excludes commercial interests.

RESULTS

This survey provides a general analysis of the current status of CAD systems according to the used image modalities and the machine learning based classifiers. Potential research studies have been discussed to create a more objective and efficient CAD systems.

Computer-Aided Breast Cancer Diagnosis with Optimal Feature Sets: Reduction Rules and Optimization Techniques

This chapter introduces a new method for knowledge extraction from databases for the purpose of finding a discriminative set of features that is also a robust set for within-class classification. Our method is generic and we introduce it here in the field of breast cancer diagnosis from digital mammography data. The mathematical formalism is based on a generalization of the k-Feature Set problem called (α, β)-k-Feature Set problem, introduced by Cotta and Moscato (J Comput Syst Sci 67(4):686-690, 2003). This method proceeds in two steps: first, an optimal (α, β)-k-feature set of minimum cardinality is identified and then, a set of classification rules using these features is obtained. We obtain the (α, β)-k-feature set in two phases; first a series of extremely powerful reduction techniques, which do not lose the optimal solution, are employed; and second, a metaheuristic search to identify the remaining features to be considered or disregarded. Two algorithms were tested with a public domain digital mammography dataset composed of 71 malignant and 75 benign cases. Based on the results provided by the algorithms, we obtain classification rules that employ only a subset of these features.

An Improved CAD System for Breast Cancer Diagnosis Based on Generalized Pseudo-Zernike Moment and Ada-DEWNN Classifier

In this paper, a novel framework of computer-aided diagnosis (CAD) system has been presented for the classification of benign/malignant breast tissues. The properties of the generalized pseudo-Zernike moments (GPZM) and pseudo-Zernike moments (PZM) are utilized as suitable texture descriptors of the suspicious region in the mammogram. An improved classifier- adaptive differential evolution wavelet neural network (Ada-DEWNN) is proposed to improve the classification accuracy of the CAD system. The efficiency of the proposed system is tested on mammograms from the Mammographic Image Analysis Society (mini-MIAS) database using the leave-one-out cross validation as well as on mammograms from the Digital Database for Screening Mammography (DDSM) database using 10-fold cross validation. The proposed method on MIAS-database attains a fair accuracy of 0.8938 and AUC of 0.935 (95 % CI = 0.8213-0.9831). The proposed method is also tested for in-plane rotation and found to be highly rotation invariant. In addition, the proposed classifier is tested and compared with some well-known existing methods using receiver operating characteristic (ROC) analysis using DDSM- database. It is concluded the proposed classifier has better area under the curve (AUC) (0.9289) and highly précised with 95 % CI, 0.8216 to 0.9834 and 0.0384 standard error.

Analysis of perceived similarity between pairs of microcalcification clusters in mammograms

PURPOSE

Content-based image retrieval aims to assist radiologists by presenting example images with known pathology that are visually similar to the case being evaluated. In this work, the authors investigate several fundamental issues underlying the similarity ratings between pairs of microcalcification (MC) lesions on mammograms as judged by radiologists: the degree of variability in the similarity ratings, the impact of this variability on agreement between readers in retrieval of similar lesions, and the factors contributing to the readers' similarity ratings.

METHODS

The authors conduct a reader study on a set of 1000 image pairs of MC lesions, in which a group of experienced breast radiologists rated the degree of similarity between each image pair. The image pairs are selected, from among possible pairings of 222 cases (110 malignant, 112 benign), based on quantitative image attributes (features) and the results of a preliminary reader study. Next, the authors apply analysis of variance (ANOVA) to quantify the level of variability in the readers' similarity ratings, and study how the variability in individual reader ratings affects consistency between readers. The authors also measure the extent to which readers agree on images which are most similar to a given query, for which the Dice coefficient is used. To investigate how the similarity ratings potentially relate to the attributes underlying the cases, the authors study the fraction of perceptually similar images that also share the same benign or malignant pathology as the query image; moreover, the authors apply multidimensional scaling (MDS) to embed the cases according to their mutual perceptual similarity in a two-dimensional plot, which allows the authors to examine the manner in which similar lesions relate to one another in terms of benign or malignant pathology and clustered MCs.

RESULTS

The ANOVA results show that the coefficient of determination in the reader similarity ratings is 0.59. The variability level in the similarity ratings is proved to be a limiting factor, leading to only moderate correlation between the readers in their readings. The Dice coefficient, measuring agreement between readers in retrieval of similar images, can vary from 0.45 to 0.64 with different levels of similarity for individual readers, but is higher for average ratings from a group of readers (from 0.59 to 0.78). More importantly, the fraction of retrieved cases that match the benign or malignant pathology of the query image was found to increase with the degree of similarity among the retrieved images, reaching average value as high as 0.69 for the radiologists (p-value <10(-4) compared to random guessing). Moreover, MDS embedding of all the cases shows that cases having the same pathology tend to cluster together, and that neighboring cases in the plot tend to be similar in their clustered MCs.

CONCLUSIONS

While individual readers exhibit substantial variability in their similarity ratings, similarity ratings averaged from a group of readers can achieve a high level of intergroup consistency and agreement in retrieval of similar images. More importantly, perceptually similar cases are also likely to be similar in their underlying benign or malignant pathology and image features of clustered MCs, which could be of diagnostic value in computer-aided diagnosis for lesions with clustered MCs.

Regularization in retrieval-driven classification of clustered microcalcifications for breast cancer

We propose a regularization based approach for case-adaptive classification in computer-aided diagnosis (CAD) of breast cancer. The goal is to improve the classification accuracy on a query case by making use of a set of similar cases retrieved from an existing library of known cases. In the proposed approach, a prior is first derived from a traditional CAD classifier (which is typically pre-trained offline on a set of training cases). It is then used together with the retrieved similar cases to obtain an adaptive classifier on the query case. We consider two different forms for the regularization prior: one is fixed for all query cases and the other is allowed to vary with different query cases. In the experiments the proposed approach is demonstrated on a dataset of 1,006 clinical cases. The results show that it could achieve significant improvement in numerical efficiency compared with a previously proposed case adaptive approach (by about an order of magnitude) while maintaining similar (or better) improvement in classification accuracy; it could also adapt faster in performance with a small number of retrieved cases. Measured by the area of under the ROC curve (AUC), the regularization based approach achieved AUC = 0.8215, compared with AUC = 0.7329 for the baseline classifier (P-value = 0.001).