Fuzzy-VGG: A fast deep learning method for predicting the staging of Alzheimer's disease based on brain MRI

•Fuzzy theory used to highlight key MRI info, improving training results while reducing time and effort.•Two-stage cutout method enhances medical images with diverse distribution, preserving original characteristics.•2D and 3D MRI images used to classify Alzheimer's with good accuracy; fusion of models improves results.•Method combining with blockchain tech allows easy updates; suitable for quasi-medical care in hospitals. Alzheimer's disease, a primary degenerative encephalopathy, predominantly affects elderly and pre-elderly individuals and is characterized by persistent anxiety-like brain activity. Despite magnetic resonance imaging emerging as a crucial tool for examining brain tissues and diagnosing Alzheimer's disease, most existing studies solely focus on detecting the disease's presence, neglecting its staging. Furthermore, these methods generally focus on the improvement of an algorithm architecture rather than on improving the quality of medical images. This paper aims to address this limitation by presenting a novel method, Fuzzy-VGG. With this approach, the image pixels are reordered based on their gray levels using fuzzy theory, placing a high priority on the most significant local information in each pixel. Following this, a VGG-based primary framework is employed to train the datasets, and a two-stage cutout strategy is used to augment the dataset and enhance training results. The results demonstrate substantial improvements in classification performance and accelerated model convergence. The primary novelties and contributions of this work encompass: 1) a focus on local key areas to expedite training and enhance results, 2) the utilization of a two-stage cutout data enhancement strategy tailored to medical images' characteristics, and 3) seamless integration with blockchain technology to facilitate scalability and continuous model improvement.