Development of a computer-aided classification system for cancer detection from digital mammograms

Abstract

Mammogram-breast x-ray is considered the most effective, low cost, and reliable method in early detection of breast cancer. Although general rules for the differentiation between benign and malignant breast lesion exist, only 15 to 30% of masses referred for surgical biopsy are actually malignant. Computer-Aided Classification system was used to help in diagnosing abnormalities faster than traditional screening program without the drawback attribute to human factors. In this work, an approach is proposed to develop a computer-aided classification system for cancer detection from digital mammograms. The proposed system consists of three major steps. The first step is region of interest (ROI) extraction of 256×256 pixels size. The second step is the feature extraction; we used a set of 88 features and we found that 78 of these feature are capable of differentiating between normal and cancerous breast tissues in order to minimize the classification error. The third step is the classification process; we used the technique of the k-Nearest Neighbor (k-NN) to classify between normal and cancerous tissues. The proposed system was shown to have the large potential for cancer detection from digital mammograms.

A Hybrid Deep Transfer Learning of CNN-Based LR-PCA for Breast Lesion Diagnosis via Medical Breast Mammograms

“…These solutions can be classified into classical and deep learning models. Early CAD systems have been developed using the classical machine learning models such as the research conducted in [18][19][20][21][22][23][24][25][26][27][28][29][30]. The basic framework in these studies consists of three stages including Region of Interest (ROI) extraction, feature extraction and selection, and classification stage using well-known classifiers such as K Nearest Neighbor (KNN), Support Vector Machine (SVM), Neural Network (NN), and Ensemble classifiers such as the work conducted by Kadah et al [18,21,24,25,31].…”

A methodology for customizing clinical tests for esophageal cancer based on patient preferences

“…Data mining techniques have also revealed unknown causes and helped in detecting breast cancer. Alolfe et al [11] used k-NN algorithm from the extracted features to classify if a particular region of interest (ROI) of the digital image, an output of mammogram, is carrying benign or malign masses to determine breast cancer and reported that the algorithm at k = 1 gave the best result. Abreu et al [12] tried to find out overall survival rate for woman, suffering from breast cancer using 3 ensemble methods (TreeBagger, LPBoost and Subspace) considering 25% missing data and found Treebagger with 3 neighbor was the best among the above three methods.…”

“…First we study the designing of classification methodologies for electronic health records, a well studied problem for many diseases e.g. heart disease [6,7,8,9,10], breast cancer [11,12,13], etc. (ref section 1.1).…”

Investigation on ROI size and location to classify mammograms

“…Results showed that the largest ROI size provides the worst detection while the best results were obtained for middle size ROI of size 512×512 pixels and proposed multi size analyses for further improvement. Alolfe et al (2008) selected an ROI of size 256×256 pixels from the mammograms in order to extract features and to classify normal and cancerous breast tissue. The results were less satisfactory due to small number of cases used in learning phase and due to variation in mammogram database.…”