Optimizer analysis on efficient-net architecture for Alzheimer’s classification based on magnetic resonance imaging (MRI)

Alzheimer's can permanently destroy brain cells that are involved in memory and thinking ability. Alzheimer's is a neurological disease that has a major influence on sufferers’ life. Early identification and treatment of Alzheimer's patients is critical since it can help postpone the disease's progression and symptoms. This study designed a classification system for Alzheimer's disease into four classes; Very Mild Demented, Mild Demented, Moderate Demented and Non Demented. The system designed using CNN (Convolutional Neural Network) method with Efficient-Net architecture based on MRI (Magnetic Resonance Imaging). The input is taken from the Kaggle Alzheimer's Dataset with a total of 1600 images. The optimization is performed using Adam, Adamax, Nadam, RMSprop, and SGD. The best system performance has been carried out with Nadam optimizer with highest accuracy value of 0.97, precision value of 0.97, recall value of 0.97, f1-score of 0.97 and loss of 0.1104. Based on the performance results, the system shows that the CNN model with Efficient-Net architecture can classify the conditions of Alzheimer's disease.