Screening glioma and glioblastoma brain tumors using dual deep learning algorithm incorporated correlative GAN and BrainNet through the probability segmentation.
The earlier identification of the tumors in human brain can improve the life time of the affected patients. Mainly, Glioma and Glioblastoma are the primary type of brain tumors where the survival rate of the patient is low and hence it's earlier screening is important. This research work proposes Dual Deep Learning (DDL) based Glioma and Glioblastoma brain tumor detection methodology. The main objective of this research work is for performing multi class brain image classification process. The proposed tumor detection system contains preprocessing, data augmentation and the proposed DDL algorithm module in training of the system for generating the training values. The testing system of the proposed work contains preprocessing, the proposed DDL algorithm module along with the probability segmentation algorithm to perform both classification and segmentation process. The preprocessing is used here to enhance the brain imaging quality to improve the tumor detection performance and the data augmentation increases the brain images count for neglecting the issues of the overfitting during the training stage of the classifier only. The proposed DDL algorithm module is designed with Correlative Generative Adversarial Networks (CGAN) and BrainNet classification algorithms, where as CGAN is proposed for computing the discriminative features which are mainly used for differentiating the Glioma and Glioblastoma. The computed discriminative features are classified by the proposed BrainNet classification algorithm which produces the classification results. The Empirical-Axiomatic Probability Segmentation Algorithm (EAPSA) have been constructed for segmenting the region of tumor pixels in both Glioma and Glioblastoma images. The ablation parameter study of the proposed DDL classification algorithm is performed and its experimental results are achieved by testing the different brain MRI images which are available on standard benchmarked brain MRI imaging datasets.