OB-TSASA Guided Soft Voting Ensemble Classifier for Breast Cancer Classification Using Thermal Infrared Images

Abstract

Nowadays, the use of thermal infrared images has become more popular in breast cancer detection. Thermal image analysis with machine learning algorithms allows physicians to make accurate decisions on cancer diagnostics. Hence, this paper introduces a novel automated breast cancer classifier using thermal infrared images. Initially, the features from the DMR – IR dataset are extracted by an integrated approach of Kernel Principle Fast Independent-Component Analysis (KPFICA) and deep object-centric pooling convolution neural network (DOCP-CNN). Additionally, the DOCP-CNN hyperparameters are tuned by the Harris Hawks Optimization (HHO) algorithm. Next, the proposed oppositional based tunicate swarm algorithm with simulated annealing (OB-TSASA) optimization is used to select the relevant features from the extracted features. Finally, the OB-TSASA guided soft voting ensemble classifier is proposed to detect and classify cancerous patients. The experiment is conducted on the DMR – IR dataset to estimate the effectiveness of the proposed method. Furthermore, the proposed method produces outstanding classification results in terms of AUC, accuracy, F1-score, sensitivity, MCC, and kappa statistics. It is observed that the classification accuracy is 97.73% which is superior to other classifiers and the error rate is 2.27% which is very low compared to other methods.