Enhanced Deep Learning Methodology for Detection and Identification of Brain Tumor using CNN

Introduction: Brain tumours are a significant global public health issue that require precise and effective diagnosis techniques for well-thought-out treatment strategies. Three separate works for the automatic identification and categorization of brain tumours in magnetic resonance imaging (MR) datasets are presented in this dissertation. Objectives: The primary objective of this work was to explore uses of the version of the discrete orthonormal S-transform (DOST) to extract texture characteristics and fuzzy C-means (FCM) for image segmentation. By using Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA) we can minimize the dimensional of the recovered features. The proposed Adaboost method is run with the reduced features, and a Random Forest (ADBRF) classifier is then employed. The proposed Adaboost method is run with the reduced features, and a Random Forest (ADBRF) classifier is then employed. Methods: A range of machine learning algorithms Reduced Complexity Spatial Fusion CNN is applied for the brain to classify and identify the brain tumor. Results: The results revealed significant performance enhancements, particularly for boosting algorithms. Mean, Variance, standard deviation, entropy, energy has been calculated and by using proposed design it has high accuracy rate. Conclusions: This study highlights the potential of advanced machine learning techniques for classifying and identification of brain tumor. Accurate analysis is often hampered by the complexity and diversity of image data. To tackle this, we created the RCSF-CNN, a proposed method that combines the extraction of complexity features with CNN for brain cancer detection and classification. Notably, our methodology enhances the resilience and accuracy of detection and classification procedures by utilizing the DOST as an intermediary stage. 21% of the dataset was set aside for rigorous testing, while the remaining 79% was used for training