Deep learning and spark architecture based intelligent brain tumor MRI image severity classification

The emerging technologies have faster growth in and have acquired a fundamental position in analyzing novel views in the anatomy of brains. The imaging modality has prevalent use in medical science for earlier detection and diagnosis. However, precise and timely diagnosis of brain tumors is a challenging task. This paper presents a novel method, namely Border Collie Firefly Algorithm-based Generative Adversarial network (BCFA-based GAN) using spark framework for effective severity level classification in brain tumor. Here, a set of slave nodes and master node is employed for performing severity classification. Here, the pre-processing is done using Laplacian filter to eradicate clatter present in image. The generated image as a result of pre-processing is segmented wherein Deep Joint model is adapted for generating segments. Thereafter, the feature extraction is performed wherein statistical features, Texton features and Karhunen-Loeve Transform-based features are extracted using slave nodes. Support vector machine (SVM) is fed with the obtained features, wherein tumor classification is done in the master node. Finally, the result is fed to BCFA-based GAN for severity level classification. The devised BCFA is used in tuning the GAN, the devised BCFA is obtained by integrating Border Collie Optimization (BCO) into Firefly Algorithm (FA). The proposed BCFA-based GAN offered the best performance and produced high values of accuracy at 97.515%, sensitivity at 97.515% as well as specificity at 97.515%.