Automated Brain Aneurysm Classification Using Deep Convolutional Neural Networks

The prevention of brain aneurysm rupture morbidity and mortality requires the early and proper identification of brain aneurysms. The present piece of work postulates an automated brain aneurysm segmentation and classification system founded on the deep convolutional neural networks, created in MATLAB 2021. The suggested system has a structured pipeline that integrates image preprocessing, handcrafted feature extraction, and deep learning-based classification to enhance the diagnostic performance. Firstly, the RGB images of the brain are turned into grayscale images and denoised with a hybrid median filter with wiener estuarization. Contrasts in images are then increased to emphasize on anatomical features. The local binary patterns (LBP) features are computed on the basis of circular neighborhoods and a Fuzzy-LBP-based segmentation scheme is implemented to precisely locate aneurysm location. The refinement of post-segmentation is performed with the help of morphological operations and thresholding, and the features of Gray-Level Co-occurrence Matrix (GLCM) are applied to describe the information on texture. An aneurysm classification into the various categories is then done automatically using a deep CNN model. The accuracy, specificity and execution time metrics are used to measure the suggested approach. The experimental results show better performance than the other existing methods and achieve an accuracy of 94.49, specificity of 91.42 with an execution time of 30.94 s. These results prove the usefulness of the suggested automated system in assisting the early and valid diagnosis of brain aneurysms.