Ct Image Lung Cancer Segmentation and Classification Using Canny-Based Expectation Maximization (CEM) and Machine Learning Algorithm

• Published in: NeuroQuantology Vol. 20; no. 10; p. 175
• Main Authors: Maheswari, S, Sundar, C, SThanabal, M
• Format: Journal Article
• Language: English
• Published: Bornova Izmir NeuroQuantology 01.01.2022
• Subjects: Algorithms; Classification; Computed tomography; Feature extraction; Image contrast; Image segmentation; Lung cancer; Lung diseases; Machine learning; Maximization; Medical imaging; Methodology; Optimization; Parameter sensitivity; Principal components analysis; Statistical analysis; Wavelet transforms
• ISSN: 1303-5150
• DOI: 10.14704/nq.2022.20.10.NQ55016

Recently, lung cancer has become very popular due to the change in lifestyle of the people. Early prediction of human lung cancer couldassistance in improving the casual for the existence of the patients. Medical image analysisperformances as a treasured tool for thecredentials of numerouskinds of human organ cancer using Computer Aided Diagnosis (CAD). In the proposed research, aninnovativestructurecould be proposed for the analysis of human unisex lung disease such as cancer using computed tomography (CT) medical images. Primarily, the CT images were pre-processed using 2D Improved AnisotropicBilateral (2D-IAB) filter and a new Edge Sharpening based Contrast and Brightness Improved Histogram Equalization (ES-CBIHE) technique. The de-noised images were then subject to segmentation using a novel Canny-based Expectation Maximization (CEM) algorithm. The segmented regions were utilized for feature extraction. Two types of statistical features were extracted from the segmented data, namely, the texture structure based feature called Gray Level Co-Occurrence Matrix (GLCM) and wavelet features. The Daubechies wavelet transform was used for the extraction of wavelet features. The dimension of the extracted features areminimized using principal component analysis (PCA) scheme. The reduced features were then classified using logistic regression machine learning classification algorithm. The segmented images are further classified into normal and abnormal. The simulation results are generated using publically available dataset. The output of the proposed methodology shows that it is an efficient classification methodology of tumors. The performance of proposed methodology classification is validated using various parameters like accuracy, specificity, sensitivity, precision, recall and F-score. The experimental results show that the proposed methodology performance is better than other existing methodologies and outperforms are state-of-the-art work of the research.