Pulmonary Nodule Classification Using Feature and Ensemble Learning-Based Fusion Techniques

The Pulmonary nodule indicates the presence of lung cancer. deep convolutional neural networks (DCNNs) have been widely used to classify pulmonary as benign or malignant. However, an individual learner usually performs unsatisfactorily due limited response space, incorrect selection hypothesis falling into local minimums. To investigate these issues, we propose ensemble learners fusion techniques based on averaging prediction score and maximum vote (MAX-VOTE). First, support vector machine (SVM) AdaBoostM2 learning algorithms are trained features from DCNNs. results both classifiers fused separately score. Secondly, feature technique is developed by fusing three DCNNs (AlexNet, VGG-16 VGG-19) through predefined rules. After that, SVM independently build multiple DCNN learners. predictions all MAX-VOTE. show that MAX-VOTE yields better performance out twelve single for binary class classification nodules. proposed also tested multi-class problem. compared implemented state-of-the-art techniques. achieved accuracy, AUC specificity scores 96.89%±0.25, 99.21%±0.10 97.70%±0.21, respectively.