Microcalcification (MC) detection is an important component of breast cancer diagnosis. However, visual analysis of mammograms is a difficult task for radiologists. Computer Aided Diagnosis (CAD) technology helps in identifying lesions and assists the radiologist in making his final decision. This work is a part of a CAD project carried out at the Imaging Science Research Division (ISRD), Digital Medical Imaging Program, Moffitt Cancer Research Center, Tampa, FL. A CAD system had been previously developed to perform the following tasks: (a) pre-processing, (b) segmentation and (c) feature extraction of mammogram images. Ten features covering spatial, and morphological domains were extracted from the mammograms and the samples were classified as Microcalcification (MC) or False alarm (False Positive microcalcification/ FP) based on a binary truth file obtained from a radiologist's initial investigation. The main focus of this work was two-fold: (a) to analyze these features, select the most significant features among them and study their impact on classification accuracy and (b) to implement and compare two machine-learning algorithms, Neural Networks (NNs) and Support Vector Machines (SVMs) and evaluate their performances with these features. The NN was based on the Standard Back Propagation (SBP) algorithm. The SVM was implemented using polynomial, linear and Radial Basis Function (RBF) kernels. A detailed statistical analysis of the input features was performed. Feature selection was done using Stepwise Forward Selection (SFS) method. Training and testing of the classifiers was carried out using various training methods. Classifier evaluation was first performed with all the ten features in the model. Subsequently, only the features from SFS were used in the model to study their effect on classifier performance. Accuracy assessment was done to evaluate classifier performance. Detailed statistical analysis showed that the given dataset showed poor discrimination between classes and proved a very difficult pattern recognition problem. The SVM performed better than the NN in most cases, especially on unseen data. No significant improvement in classifier performance was noted with feature selection. However, with SFS, the NN showed improved performance on unseen data. The training time taken by the SVM was several magnitudes less than the NN. Classifiers were compared on the basis of their accuracy and parameters like sensitivity and specificity. Free Receiver Operating Curves (FROCs) were used for evaluation of classifier performance. The highest accuracy observed was about 93% on training data and 76% for testing data with the SVM using Leave One Out (LOO) Cross Validation (CV) training. Sensitivity was 81% and 46% on training and testing data respectively for a threshold of 0.7. The NN trained using the 'single test' method showed the highest accuracy of 86% on training data and 70% on testing data with respective sensitivity of 84% and 50%. Threshold in this case was -0.2. However, FROC analyses showed overall superiority of SVM especially on unseen data. Both spatial and morphological domain features were significant in our model. Features were selected based on their significance in the model. However, when tested with the NN and SVM, this feature selection procedure did not show significant improvement in classifier performance. It was interesting to note that the model with interactions between these selected variables showed excellent testing sensitivity with the NN classifier (about 81%). Recent research has shown SVMs outperform NNs in classification tasks. SVMs show distinct advantages such as better generalization, increased speed of learning, ability to find a global optimum and ability to deal with linearly non-separable data. Thus, though NNs are more widely known and used, SVMs are expected to gain popularity in practical applications. Our findings show that the SVM outperforms the NN. However, its performance depends largely on the nature of data used.

This thesis presents techniques constructed to aid the radiologists in detecting breast cancer, the second largest cause of cancer deaths for western women. In the first part of the thesis, a computer-aided detection (CAD) system constructed for the detection of stellate lesions is presented. Different segmentation methods and an attempt to incorporate contra-lateral information are evaluated. In the second part, a new method for evaluating such CAD systems is presented based on constructing credible regions for the number of false positive marks per image at a certain desired target sensitivity. This method shows that the resulting regions are rather wide and this explains some of the difficulties encountered by other researchers when trying to compare CAD algorithms on different data sets. In this part an attempt to model the clinical use of CAD as a second look is also made and it shows that applying CAD in sequence to the radiologist in a routine manner, without duly altering the decision criterion of the radiologist, might very well result in suboptimal operating points. Finally, in the third part two dual-energy imaging methods optimized for contrast-enhanced imaging of breast tumors are presented. The first is based on applying an electronic threshold to a photon-counting digital detector to discriminate between high- and low-energy photons. This allows simultaneous acquisition of the high- and low-energy images. The second method is based on the geometry of a scanned multi-slit system and also allows single-shot contrast-enhanced dual-energy mammography by filtering the x-ray beam that reaches different detector lines differently. / QC 20100819

<p>The first part of this thesis is concerned with trend modelling of breast cancer mortality rates. By using an age-period-cohort model, the relative contributions of period and cohort effects are evaluated once the unquestionable existence of the age effect is controlled for. The result of such a modelling gives indications in the search for explanatory factors. While this type of modelling is usually performed with 5-year period intervals, the use of 1-year period data, as in Paper I, may be more appropriate.</p><p>The main theme of the thesis is the evaluation of the ability to detect signals in x-ray images of breasts. Early detection is the most important tool to achieve a reduction in breast cancer mortality rates, and computer-aided detection systems can be an aid for the radiologist in the diagnosing process.</p><p>The evaluation of computer-aided detection systems includes the estimation of distributions. One way of obtaining estimates of distributions when no assumptions are at hand is kernel density estimation, or the adaptive version thereof that smoothes to a greater extent in the tails of the distribution, thereby reducing spurious effects caused by outliers. The technique is described in the context of econometrics in Paper II and then applied together with the bootstrap in the breast cancer research area in Papers III-V.</p><p>Here, estimates of the sampling distributions of different parameters are used in a new model for free-response receiver operating characteristic (FROC) curve analysis. Compared to earlier work in the field, this model benefits from the advantage of not assuming independence of detections in the images, and in particular, from the incorporation of the sampling distribution of the system's operating point.</p><p>Confidence intervals obtained from the proposed model with different approaches with respect to the estimation of the distributions and the confidence interval extraction methods are compared in terms of coverage and length of the intervals by simulations of lifelike data.</p>

The first part of this thesis is concerned with trend modelling of breast cancer mortality rates. By using an age-period-cohort model, the relative contributions of period and cohort effects are evaluated once the unquestionable existence of the age effect is controlled for. The result of such a modelling gives indications in the search for explanatory factors. While this type of modelling is usually performed with 5-year period intervals, the use of 1-year period data, as in Paper I, may be more appropriate. The main theme of the thesis is the evaluation of the ability to detect signals in x-ray images of breasts. Early detection is the most important tool to achieve a reduction in breast cancer mortality rates, and computer-aided detection systems can be an aid for the radiologist in the diagnosing process. The evaluation of computer-aided detection systems includes the estimation of distributions. One way of obtaining estimates of distributions when no assumptions are at hand is kernel density estimation, or the adaptive version thereof that smoothes to a greater extent in the tails of the distribution, thereby reducing spurious effects caused by outliers. The technique is described in the context of econometrics in Paper II and then applied together with the bootstrap in the breast cancer research area in Papers III-V. Here, estimates of the sampling distributions of different parameters are used in a new model for free-response receiver operating characteristic (FROC) curve analysis. Compared to earlier work in the field, this model benefits from the advantage of not assuming independence of detections in the images, and in particular, from the incorporation of the sampling distribution of the system's operating point. Confidence intervals obtained from the proposed model with different approaches with respect to the estimation of the distributions and the confidence interval extraction methods are compared in terms of coverage and length of the intervals by simulations of lifelike data.