An EfficientNetV2-Based for Alzheimer’s Disease Classification

In Indonesia, Alzheimer’s disease has emerged as a critical public health priority. This neurodegenerative disorder is characterized by the gradual erosion of memory, linguistic capabilities, and problem-solving skills resulting from irreversible neuronal damage. Magnetic Resonance Imaging (MRI) is commonly used for early diagnosis; however, manual interpretation of MRI scans is time-consuming and subject to inter-observer variability among medical professionals. Recent advances in artificial intelligence have enabled automated analysis of MRI images for Alzheimer’s disease detection, yet many existing approaches rely on deep learning architectures with high computational complexity. To address this limitation, this study proposes a lightweight deep convolutional network based on EfficientNetV2 for Alzheimer’s disease classification using brain MRI images. Data augmentation techniques, including random rotation, affine transformation, horizontal and vertical flipping and normalization are applied to enhance model generalization. Two EfficientNetV2 variants, EfficientNetV2_s and EfficientNetV2_m, are evaluated and compared using accuracy, precision, recall, and F1-score metrics. Experimental results demonstrate that EfficientNetV2_s achieves superior performance, attaining an accuracy, precision, recall, and F1-score of approximately 0.90, while EfficientNetV2_m achieves corresponding values of approximately 0.81, indicating lower generalization capability. These results confirm that the smaller EfficientNetV2_s model provides more accurate and reliable classification performance despite its reduced computational complexity. Keywords - Alzheimer’s Disease, Classification, Convolutional Neural Networks, Deep Learning.