Advanced Implementation of Convolutional Neural Networks for Alzheimer's Diseases Diagnosis

Advanced Implementation of Convolutional Neural Networks for Alzheimer's Diseases Diagnosis

연구 분야: Artificial Intelligence

학회: SN Computer Science

Alzheimer's Disease (AD) is a serious medical issue because of its progressive nature and the problem in diagnosis. One of the latest developments in artificial intelligence, the Convolutional Neural Networks(CNN), opens up great new possibilities in the development of methods that are being widely used for obtaining high performance in the diagnosis of AD. The new implementation of CNN with enhanced performance for the diagnosis of AD is proposed on brain Magnetic Resonance Imaging (MRI) copies. Pooling layers rectified linear unit (ReLU) activation functions and convolutional layers(Conv2D) comprise its fundamental architecture. The model here proposed is focused on the role of the activation function and convolution processes in feature extraction and nonlinear transforms but emphasizes the simplicity and efficiency of the architecture.In this paper, the model was processed and evaluated with a dataset of MRI pictures classified into various stages of AD to assess the effectiveness of the technique. The model's ability to categorize was assessed using performance metrics like recall, accuracy, precision, and F1-score. The results are promising, showing the efficient design of CNN for the analysis of medical images, considering that although simple, the performance of the architecture is competitive for distinguishing the stages of AD. This work presents the basic CNN components for medical diagnosis, thus opening a venue toward more complex and hybrid neural network architectures. This study evaluates that the advanced implementation of CNN, with 99.06% validation accuracy and 93.82% balanced accuracy, is less complex and is good for real-world uses in early diagnosis.