Shape factors for analysis of breast tumors in mammograms
Abstract
Distinguishing between benign and malignant breast tumors requires careful analysis of their shape complexity and radiographic definition. In this paper we examine the usefulness of shape factors such as compactness, moments, Fourier descriptors, and statistics of chord lengths in distinguishing between circumscribed/spiculated and benign/malignant masses. A database of 54 tumors was used in pattern classification experiments. Classification accuracies of 95% for circumscribed/spiculated, 76% for benign/malignant, and 77% for four-group classification were obtained, which indicate the usefulness of the proposed methods in breast cancer diagnosis.Department
Electrical and Computer EngineeringPublisher
IEEEScopus Count
Collections
Related items
Showing items related by title, author, creator and subject.
-
Breast cancer detection from mammograms using artificial intelligenceBreast cancer is one of the fastest-growing forms of cancer in the world today. Breast cancer is primarily found in women, and its frequency has been gaining significantly in the last few years. The key to tackle the rising cases of breast cancer is early detection. Many studies have shown that early detection significantly reduces the mortality rate of those affected. Machine learning and deep learning techniques have been adopted in the present scenario to help detect breast cancer in an early stage. Deep learning models such as the convolutional neural networks (CNNs) are suited explicitly to image data and overcome the drawbacks of machine learning models. To improve upon conventional approaches, we apply deep CNNs for automatic feature extraction and classifier building. In this chapter, we have demonstrated thoroughly the use of deep learning models through transfer learning, deep feature extraction, and machine learning models. Computer-aided detection or diagnosis systems have recently been developed to help health-care professionals increase diagnosis accuracy. This chapter presents early breast cancer detection from mammograms using artificial intelligence (AI). Various models have been presented along with an in-depth comparative analysis of the different state-of-the-art architectures, custom CNN networks, and classifiers trained on features extracted from pretrained networks. Our findings have indicated that deep learning models can achieve training accuracies of up to 99%, while both validation and test accuracies up to 96%. We conclude by suggesting various improvements that could be made to existing architectures and how AI techniques could help further improve and help in the early detection of breast cancer.
-
Measures of acutance and shape for classification of breast tumorsMost benign breast tumors possess well-defined, sharp boundaries that delineate them from surrounding tissues, as opposed to malignant tumors. Computer techniques proposed to date for tumor analysis have concentrated on shape factors of tumor regions and texture measures. While shape measures based on contours of tumor regions can indicate differences in shape complexities between circumscribed and spiculated tumors, they are not designed to characterize the density variations across the boundary of a tumor. Here, the authors propose a region-based measure of image edge profile acutance which characterizes the transition in density of a region of interest (ROI) along normals to the ROI at every boundary pixel. The authors investigate the potential of acutance in quantifying the sharpness of the boundaries of tumors, and propose its application to discriminate between benign and malignant mammographic tumors. In addition, they study the complementary use of various shape factors based upon the shape of the ROI, such as compactness. Fourier descriptors, moments, and chord-length statistics to distinguish between circumscribed and spiculated tumors. Thirty-nine images from the Mammographic Image Analysis Society (MIAS) database and an additional set of 15 local cases were selected for this study. The cases included 16 circumscribed benign, 7 circumscribed malignant, 12 spiculated benign, and 19 spiculated malignant lesions. All diagnoses were proven by pathologic examinations of resected tissue. The contours of the lesions were first marked by an expert radiologist using X-Paint and X-Windows on a SUN-SPARCstation 2 Workstation. For computation of acutance, the ROI boundaries were iteratively approximated using a split/merge and end-point adjustment technique to obtain the best-fitting polygonal approximation. The jackknife method using the Mahalanobis distance measure in the BMDP (Biomedical Programs) package was used for classification of the lesions using acutance and the shape factors as features in various combinations. Acutance alone resulted in a benign/malignant classification accuracy of 95% the MIAS cases. Compactness alone gave a circumscribed/spiculated classification rate of 92.3% with the MIAS cases. Acutance in combination with a moment-based shape measure and a Fourier descriptor-based measure gave four-group classification rate of 95% with the MIAS cases. The results indicate the importance of including lesion edge definition with shape information for classification of tumors, and that the proposed measure of acutance fills this need.
-
Diagnosis of breast cancer from histopathological images with deep learning architecturesBreast cancer is one of the most common cancer types among women worldwide. If not treated in earlier stages, it may be fatal. Therefore early diagnosis of breast cancer can minimize the human life risk. Mammograms and ultrasound imaging technologies play a crucial role to detect intraductal papillomas. However, the determination process of intraductal papillomas requires histopathological image analysis which may be mostly time-consuming, subjective, and tedious if carried out manually by the experts. To cover issue, computer-aided diagnosis (CAD) systems came into consideration. However, earlier CAD systems could not achieve significant improvement in the diagnosis process and their usage of them did not become widespread for more than a decade. Since deep learning has made so many significant advances in a wide variety of image applications, CAD systems that use its principles perform as well as the experts in stand-alone mode, and even perform better when used in support mode. In this chapter, we utilized various deep learning architectures for the detection process of breast cancer on the invasive ductal carcinoma (IDC) dataset which is one of the most popular and remarkable datasets in this field. According to the results, the pretrained VGG16 and MobileNet architectures obtain the best detection performance, reaching nearly 92% classification accuracy.
