br In the first scenario experiments are conducted on total
In the first scenario, experiments are conducted on total 20 features using three different classifiers, i.e., SVM, NB, and RF with k-fold cross validation (k = 10) without using CAGA. According to the bar chart in Fig. 7, the accuracy, sensitivity, specificity, and F-Measure for each of the given classifier, i.e., SVM, NB, and RF, are determined for the evaluation of their performances. It can be observed from Fig. 7, that the average performance results of ac-curacy, sensitivity, specificity, and F-Measure are 93.99%, 96.22%, and 94.10%, achieved for SVM, NB, and RF classifier, respectively. In the second scenario, CAGA based feature optimization is applied during experimentation, where 6 optimal features are selected out of a total 20 features for classification purpose. The optimal features such as Feret y , Feret angle, Min-Feret, Aspect Ratio, Roundness, and Solidity are extracted and used in three different classifiers, i.e., SVM, NB, and RF, for the classification with cross-validation of 10-fold. It can be observed from Fig. 8 that using CAGA based selected features, the average perfor-mance results of accuracy, sensitivity, specificity, and F-Measure are 95.99%, 98.80%, and 93.25%, achieved for SVM, NB, and RF classifier, respectively.
The classification results obtained in both scenarios, i.e., with or without CAGA, indicates the advantage of CAGA over non-CAGA in the classification of malignant cells. From Figs. 7 and 8, it can be observed that in both scenarios the NB classifier performs better than SVM and RF due to its high ratio in accuracy, sensitiv-ity, specificity and F-Measure. Therefore, by applying the majority voting algorithm, NB is the best classifier for the malignant Pyocyanin detection in breast cytology images.
In this paper, we have proposed a framework for an e-Health care service to efficiently and accurately detect and classify ma-lignant cells in breast cancer in breast cytology images. In the proposed framework, state-of-the-art CAGA technique is used for the optimal selection of features from different shape based and textured based features to increase the classification accuracy
and reduce the computational complexity. Similarly, three well-known classification techniques, i.e., SVM, NB, and RF, are applied to classify tumor cells into malignant and benign. The results obtained using real data sets during experimentation validate the classification accuracy up to 98% for NB as compared to SVM and RF. In future, the proposed framework will be improved by incor-porating deep learning techniques in cloud-based e-Health care servers supporting IoMT for the accurate classification of other medical diseases, which can assist medical physicians efficiently and accurately to make more intelligent decisions in diagnosing breast cancer.
 M.G. Gafar, M. Elhoseny, M. Gunasekaran, Modeling neutrosophic variables based on particle swarm optimization and information theory measures for forest fires, J. Supercomput. (2018) 1–18.
Mr. Sana Ullah Khan completed his M.S. in Computer Science from Islamia College University Peshawar in 2014. He is currently perusing his Ph.D. in the field Medical Image Processing from the same university. His research interest includes Computer vision, Medical imaging, Machine learning, Cloud computing and Deep Learning. He is the author of various research articles of journals and conferences.
Dr. Naveed ISLAM completed his Ph.D. in Computer Science from University of Montpellier II, France. His research interests include Data Protection and Secu-rity, Computer Vision and Artificial Intelligence. He is the author of numerous international journal and conference articles. He has worked as a computer vi-sion R&D engineer at different organizations in France. is currently an Assistant Professor at Islamia College University, Peshawar, Pakistan.
Dr. Zahoor Jan is currently holding the rank of an associate professor in computer science at Islamia College Peshawar, Pakistan. He received his M.S. and Ph.D. degree from FAST University Islamabad in 2007 and 2011, respectively. He is also chairman of Depart-ment of Computer Science at Islamia College Peshawar, Pakistan. His areas of interests include image pro-cessing, machine learning, computer vision, artificial intelligence and medical image processing, biologically inspired ideas like genetic algorithms and artificial neural networks, and their soft-computing applications,
biometrics, solving image/video restoration problems using combination of classifiers using genetic programming, optimization of shaping functions in digital watermarking and image fusion.
Dr. Ikram Ud Din received his Ph.D. degree in Com-puter Science from the School of Computing, Universiti Utara Malaysia (UUM). He was a member of the In-terNetWorks Research Laboratory, UUM from 2014 to 2016. He is currently a Lecturer at the Department of Information Technology, The University of Haripur. He received his M.Sc. in Computer Science and M.S. in Computer Networking from the Department of Com-puter Science, University of Peshawar, Pakistan in 2006 and 2011, respectively. He also served as IEEE UUM Student Branch Professional Chair. Ikram has 10 years