ZFDMN: a hybrid deep learning approach for lung cancer detection using MapReduce framework

This research presents a hybrid deep learning approach, named Zeiler and Fergus deep maxout network (ZFDMN) for lung cancer detection using MapReduce framework with computed tomography (CT) images. Initially, the input big data is partitioned using Bayesian fuzzy clustering (BFC) and is fed into the MapReduce framework. In the mapper phase, pre-processing of images is performed using the Kalman filter and from the pre-processed images, the lung lobes are segmented using Psi-Net. Then, the features are extracted from the segmented lung lobe using various feature extractors. In the reducer phase, lung cancer detection is performed using hybrid ZFDMN, which is the integration of Zeiler and Fergus network (ZFNet) and deep maxout network (DMN). The outcomes of the experiment show that ZFDMN attained better performance with maximum sensitivity, specificity, and accuracy of 91.98%, 91.43%, and 90.88%.