Automated systems based on machine learning present more accurate diagnosis and easy analysis of medical images. In real world applications especially in medical applications we are usually faced with the problem of imbalanced data sets. In lung CT nodule detection systems, nodules are a part of minority class and other non-nodule organs are a member of the majority class. So to detect nodules without ignoringthem we should improve the predictive accuracies over the minority class. Conventional machine learning classifiers have high predictive accuracy over the majority class and less predictive accuracy over the class with minor members. Ensemble of classifiers is one of the most important solutions.In this work researchers used the LogitBoost ensemble classifier to diagnosis nodules from non-nodule objects. LogitBoost optimized the AdaBoostM1 by minimizing the binomial deviance which causes incorrect classification over the minority observations. LogitBoost can give better average accuracy than AdaBoostM1 for data with poorly separable classes. One of the problems in ensemble classifiers is estimating the cost parameters. In this paper Particle Swarm Optimization (PSO) algorithm was used to optimize these parameters. The extracted features from lung nodules were applied to the proposed optimized LogitBoost algorithm as its predictor variables. Then the data set was randomly divided into two sets: 80% as the training set and the remaining 20% as testing set. Simulation results revealedthat the proposed algorithm could detect nodules with high performance of AUC=86%.

Computer Aided Diagnosis (CAD), Lung Nodule, Ensemble Classification, Learning Classifiers, LogitBoost, EstimateParameter, PSO