Intelligent Automated System for Detecting Diagnostically Challenging Breast Cancers

Nonmass-enhancing (NME) lesions constitute a diagnostic challenge in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) of the breast. Computer-aided diagnosis (CAD) systems provide physicians with advanced tools for analysis, assessment, and evaluation that have a significant impact on the diagnostic performance. Here, we propose a new approach to address the challenge of NME lesion detection and segmentation, taking advantage of independent component analysis (ICA) to extract data-driven dynamic lesion characterizations. A set of independent sources was obtained from the DCE-MRI dataset of breast cancer patients, and the dynamic behavior of the different tissues was described by multiple dynamic curves, together with a set of eigenimages describing the scores for each voxel. A new test image is projected onto the independent source space using the unmixing matrix, and each voxel is classified by a support vector machine (SVM) that has already been trained with manually delineated data. A solution to the high false-positive rate problem is proposed by controlling the SVM hyperplane location, outperforming previously published approaches.

Accurate methods for computer aided diagnosis of breast cancer increase accuracy of detection and provide support to physicians in detecting challenging cases. In dynamic contrast enhancing magnetic resonance imaging (DCE-MRI), motion artifacts can appear as a result of patient displacements. Non-linear deformation algorithms for breast image registration provide with a solution to the correspondence problem in contrast with affine models. In this study we evaluate 3 popular non-linear registration algorithms: MIRTK, Demons, SyN Ants, and compare to the affine baseline. We propose automatic measures for reproducible evaluation on the DCE-MRI breast-diagnosis TCIA-database, based on edge detection and clustering algorithms, and provide a rank of the methods according to these measures.

Computer aided applications in Dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) are increasingly gaining attention as important tools to asses the risk of breast cancer. Chest wall detection and whole breast segmentation require effective solutions to increase the potential benefits of computer aided tools for tumor detection. Here we propose a 3D extension of Gabor filtering for detection of wall-like regions in medical imaging, and prove its effectiveness in chest-wall detection.

Imbalanced datasets constitute a challenge in medical-image processing and machine learning in general. When the available training data is highly imbalanced, the risk for a classifier to find the trivial solution increases dramatically. To control the risk, an estimate on the prior class probabilities is usually required. In some medical datasets, such as breast cancer imaging techniques, estimates on the priors are intractable. Here we propose a solution to the imbalanced support vector classification problem when prior estimations are absent based on a case-dependent transformation on the decision function.

Non-mass enhancing lesions (NME) constitute a diagnostic challenge in dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) of the breast. Computer Aided Diagnosis (CAD) systems provide physicians with advanced tools for analysis, assessment and evaluation that have a significant impact on the diagnostic performance. Here, we propose a new approach for the specific problem of NME detection and segmentation, by taking advantage of independent component analysis (ICA) to extract a data-driven dynamic characterization of tissue. A set of independent sources was obtained from a dataset of patients, and the dynamic behavior of the different tissues was described by multiple dynamic curves, together with a set of eigenimages describing the scores for each voxel. A new test image is projected onto the independent source space using the unmixing matrix, and each voxel is classified by a support vector machine (SVM) that has already been trained with manually delineated data. A solution to the high false positive rate problem is proposed by controlling the SVM hyperplane location. The CAD system is trained and validated, reaching a DSC coefficient of 0.7215 for NME segmentation.