Decision support detection system for lung nodule abnormalities based on machine learning algorithms
AbstractObjective Investigates the possibility of the early detection in the case of lung infection. Most cases of lung cancer are detected in advance stages as this type is hard to be detected in premature phases. The Zero-Change dataset was chosen to measure the systems' performance on nodule growth. The chosen dataset is assumed as a proven clinical dataset and was used by several researchers in their proposed systems. The designed detection technique has been considered to be used as a decision support tool. This technique is based on using two machine learning algorithms for classification purposes.
Methods Machine learning techniques was applied to detect interesting patterns and manipulate the dataset images in order to enhance the classification task. Preprocessing procedures also have been applied using different MATLAB functions. In addition two well-known techniques that related to the SVMs; the RBF kernel based support vector machines and the polynomial kernel based support vector machines have also applied using MATLAB© package named PRTools.
Results The performance of this paper proposed technique was evaluated based on several values of both chosen techniques. The procedure was implemented on the basis of leave-one-out-cross-validation procedure in order to generate unbiased outcomes. The results of cross validation procedure is averaged and presented as a classifier outcome. The misclassification error, sensitivity, specificity and accuracy are calculated to show a clear image about the two classifiers.
Conclusion The experimental results have shown that the proposed system has scored high accuracy by Polynomial kernel SVM. A set of distinguishable representative features which are correlated together by a statistics association. Also this designed system can be considered as a benchmark for developing of other tissues abnormalities signs detection systems.