The power laws: Zipf and inverse Zipf for automated segmentation and classification of masses within mammograms

Abstract

Breast cancer is becoming the leading form of cancer among women worldwide, indeed, there are no effective ways to prevent this disease at present, therefore, it's early screening and detection is the key to rise the success of treatment, hence, the reduce of the associated mortality rates. This work aims to improve the performance of the current computer-aided detection and diagnosis approaches (CADe/CADx) of breast cancer which involve the application of the computer technology in mammograms analysis and understanding; for this purpose, we deal with the power laws: Zipf and inverse Zipf. The originality of this research lays in the contribution of the power laws for mammograms analysis; it is the first attempt to use them in the field of mammograms masses segmentation and classification, indeed, these laws characterize the structural complexity of texture within mammograms and provide the curves of Zipf and inverse Zipf which carry significant information that could be used to mammograms masses detection and classification along a new set of textural features extracted from the curves of Zipf and inverse Zipf. According to our experiments conducted on a mammogram database used in the framework of a bilateral project between our university and the hospital CHU at Algeria, we can assert that our approach based Zipf's and inverse Zipf's laws is a powerful and efficient approach for automated mammograms masses detection and classification.

A Review of Machine Learning Techniques for the Classification and Detection of Breast Cancer from Medical Images

“…To classify tumors into benign or malignant tissue, authors [42] built Fuzzy C-Means (FCM) on DDSM and MIAS datasets, and they verified that accuracy is 87%, sensitivity value (90.47%), and specificity value (84.84%).…”

A Comparative Study of Machine Learning and Deep Learning Algorithms for the Detection and Classification of Breast Cancer From Mammography Images and Ultra Sound Images

“…They are classified according to the CNN architectures of GoogleNet and VGGNet. The GoogLeNet, VGGNet, and ResNet architectures all achieve an average classification of 93.5%.To identify tumors as benign or malignant tissue, the author [19] developed fuzzy C-means (FCM) on DDSM and MIAS datasets and found that the accuracy was 87%, the cutoff value (90.47%) and the specific value (84.84%). Accuracy is 87% Sn value is 90 to 47% Sp value is 84 to 84% [20] DDSM, MIAS LS SVM, KNN, Random Forest, and Naive Bayes Accuracy 92% [21] Private-1896 cases GLCM SFFS (sequential floating forward selection) the bilateral CC and MLO view images Sn-value is 68.8% Sp value is 95.0% The AUC value is 0.85 ± 0.046 [22] The Datasets were collected at two medical institutions CNN 87.68%…”

A Brief Survey on Breast Cancer Diagnostic With Deep Learning Schemes Using Multi-Image Modalities

“…Different similarity criteria are based on the type of objects which include Euclidean distance and cosine similarity. Fuzzy-C Mean is one of the leading clustering methods which widely used for breast cancer classification [60], [61], [69], [104]. In this study, 12 publications used fuzzy (fuzzy c-mean) algorithms that are 5% of the total reviewed studies.…”

“…It widely used some visual information in the image, such as color, shape, texture, and other information. M. Hamoud et al [104] and M. Kallenberg et al [130] extracted some textural features which include contrast, correlation, the sum of (average entropy, variance), entropy, homogeneity, maximum correlation coefficient, correlation, variance, inertia, inverse difference, entropy difference, variance difference. W. Sun et al [131] used DLL techniques to extract geometric features such as size, circularity, sphericity, irregularity.…”

“…17 studies calculate both SN and SP [32], [63]. 57 studies calculates the Acc, SN, and SP [43], [60], [61], [104], [123]. 13 studies calculate the SN, SP, and AUC [51], [147], [148].…”