A fully-automated computer-aided breast lesion detection and classification system. (September 2020)

• Record Type: Journal Article
• Authors: Mutlu, Fuldem; Çetinel, Gökçen; Gül, Sevda
• Is Part Of: Biomedical signal processing and control. Volume 62(2020)
• Publication Date: 2020-09
• Subjects: ACC Accuracy -- AUC Area under care -- BPE Background parenchymal enhancement -- CAD Computer-aided detection and diagnosis -- CNN Convolutional neural network -- DCE-MRI Dynamic contrast enhanced-magnetic resonance imaging -- DN Dependence non-uniformity -- DV Dependence variance -- FCM Fuzzy c-means -- FN False negative -- FP False positive -- FS Fisher score -- GLCM Gray level co-occurrence matrix -- GLDM Gray level dependence matrix -- GLN Gray level non-uniformity -- GLRLM Gray level run length matrix -- GLV Gray level variance -- ICA Independent component analysis -- KNN K-nearest neighbor -- LDE Large dependence emphasis -- LOO Leave-one-out -- MRF Markov random field -- MRI Magnetic resonance imaging -- NB Naïve Bayes -- NGTDM Neighboring gray tone difference matrix -- PRE Precision -- QDA Quadratic discriminant analysis -- RF Random forest -- RG Region growing -- ROI Region of interest -- SDE Small dependence emphasis -- SEN Sensitivity -- SPE Specificity -- TN True negative -- TP True positive -- WHO World health organization; k-means clustering -- Texture analysis -- Shape analysis -- Fisher score -- Cross-Validation -- Classification; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.102157
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A comprehensive and fully automatic decision support system is designed for breast lesion detection and classification. A two-stage segmentation procedure is proposed. Histogram, shape, GLCM, GLRLM, NGTDM, and GLDM based 88 features are derived to characterize the breast lesions. SVM, KNN, RF and NB techniques are performed to classify breast lesions as benign or malignant. Abstract: This study presents an automatic computer-aided detection and diagnosis system which consists of two parts. The first part is for breast lesion characterization developed in pattern recognition framework (K-means clustering method) which is important to provide useful information for breast lesion characterization. Characterization of the detected lesion areas is done based on 6 parameters that are: (1) histogram, (2) shape, (3) gray level co-occurrence matrix, (4) gray level run length matrix, (5) neighboring gray tone difference matrix, and (6) gray level dependence matrix features. The second part of the system is developed based on machine learning algorithms and serves for the classification of localized breast lesions as benign and malignant. For classification, 4 different machine learning algorithms were investigated: (1) support vector, (2) k-nearest neighbors, (3) random forest, and (4) naïve Bayes classifiers. 84 histopathologically proven breast lesions were analyzed in the study. The proposed system compensates the motion artifacts, segments breast lesions, and classifies the lesions as benign and malignant. The results prove that the developed comprehensive system can detect and classifies breast lesions without any intervention. The best accuracy, sensitivity, specificity, and precision values to decide the tumor aggressiveness are 90.36%, 96.25%, 83.33%, and 92%, respectively.