Enhancing Early Detection of Brain Aneurysms: A CNN-Driven, Real-Time Approach with Angiography Imaging

Brain aneurysm is a swell or balloon in a blood vessel in the brain that can cause severe health problems to the patients. Due to its potentiality, if it ruptures it leads to serious neurological disorder such as: coma, hemorrhagic stroke, and even death. Early identification of brain aneurysm plays a crucial role in treatment. However, the traditional methods such as Computerized Tomographic Angiography (CTA), Magnetic Resonance Imaging (MRI) takes more time and complex procedures. The relentless advancement of medical imaging technologies continues to transform diagnostic paradigms in neurology. In this context, this study integrates the convolutional neural networks (CNNs) into real-time brain aneurysm detection from angiography image. This paper presents a CNN-based model designed to rapidly and accurately identify cerebral aneurysms from angiography images. The proposed method, initially prepares the dataset by collecting diverse angiography images from medical archives. Later, the images undergoes with various preprocessing methods such as: denoising, contrast enhancement, normalization, resizing, and augmentation to enhance the quality of data, and model generalization. Then, the method automatically extracts the blood vessel structure and aneurysm characteristics such as: irregular shapes, unusual bulges, and densities within the vessels as features. Finally, the extracted features passed through the connected layers and the CNN classifies the whether an image contains aneurysm or not. The CNN model is trained with stochastic gradient descent (SGD) algorithm, and the trained model is deployed on a standard Graphics Processing Unit (GPU) for real time inference. The results showed that, the performance of the method was superior compared to the traditional methods.