Alzheimer's Disease MRI Image Segmentation Based on the Enhanced U-Net

The most widespread form of dementia for which there is still no cure is Alzheimer's disease (AD). Choosing the best treatment options for AD can be aided by an accurate classification of the disease. Several studies from the past ten years have demonstrated the value of deep learning algorithms for diagnosing AD. In this study, we suggest a U-net style model for the diagnosis of AD utilising 3D T1-weighted magnetic resonance imaging (MRI). In this study, the Alzheimer is segmented from 2D brain scans using a U-Net Convolutional Network-based method. The first U-Net design executes a series of convolutional operations with a filter size of 3*3. In this paper we have used techiques of hybrid pooling. So instead of using max pooling we have used mixing of minimum pooling and average pooling. The updated architecture, which outperforms the original U-Net model clearly, was seen to obtain an admirable performance rate of 91.23%.

I. Introduction

Globally the prominent cause of dementia is Alzheimer's disease (AD), a progressive degenerative neurological ailment that causes the brain cells to gradually deteriorate. Early signs of AD include forgetfulness and inability to recall recent incidents. Additionally, as the illness worsens, it affects every aspect of cognition, ultimately rendering the patient wholly dependent on others for even the most fundamental activities of daily living [1]. There may be up to 55 million Alzheimer's patients globally, and it's predicted that figure will rise by 10 million per year. AD is commonly observed in old age. If left unchecked, it cause serious impact on future as world age increasing,