Efficient Hyperparameter Optimization of Convolutional Neural Networks on Classification of Early Pulmonary Nodules

Abstract

Lung cancer is the leading cause of cancer mortality, accounting for approximately 20% of all cancer-related deaths. Patients diagnosed in the early stages have a 1-year survival rate of 81-85% while in an advanced stage have 15-19% chances of survival. Therefore, it is very necessary to diagnose lung cancer in early stages in malignant or benign, when the nodules are still very small, but it is a complex task even for experienced specialists and presents some challenges. To assist specialists, computer-aided diagnosis systems have been used to improve the accuracy in the diagnosis. In this paper, we exploit the use of a technique of hyperparameter tuning to find the best architecture of a Convolutional Neural Network to classify small pulmonary nodules balanced with diameter 5-10mm. The best results achieved were an error rate of 12%, sensitivity of 94%, specificity of 83%, accuracy of 88% and F-measure of 89%