Automatic lung cancer detection from CT image using improved deep neural network and ensemble classifier

The development of the computer-aided detection system placed an important role in the clinical analysis for making the decision about the human disease. Among the various disease examination processes, lung cancer needs more attention because it affects both men and women, which leads to increase the mortality rate. Traditional lung cancer prediction techniques failed to manage the accuracy because of low-quality image that affects the segmentation process. So, in this paper new optimized image processing and machine learning technique is introduced to predict the lung cancer. For recognizing lung cancer, non-small cell lung cancer CT scan dataset images are collected. The gathered images are examined by applying the multilevel brightness-preserving approach which effectively examines each pixel, eliminates the noise and also increase the quality of the lung image. From the noise-removed lung CT image, affected region is segmented by using improved deep neural network that segments region in terms of using layers of network and various features are extracted. Then the effective features are selected with the help of hybrid spiral optimization intelligent-generalized rough set approach, and those features are classified using ensemble classifier. The discussed method increases the lung cancer prediction rate which is examined using MATLAB-based results such as logarithmic loss, mean absolute error, precision, recall and F-score.