Research On Artificial Intelligence Algorithm For MRI Assisted Diagnosis Of Alzheimer’s Disease

Structural magnetic resonance imaging（MRI）provides useful information for biomarker exploration and intelligent classification of Alzheimer’s disease（AD）.In recent years,machine learning and deep learning methods have been widely used in feature extraction and computer-aided diagnosis or prediction of the transition from mild cognitive impairment（MCI）to AD.In this research,we aim to develop new artificial intelligence algorithms to detect or predict AD in an effective way.First,we proposed a modified 3D Efficient Net with sequentially connected 3D mobile inverted bottleneck convolution（MBConv）block to explore multi-scale characteristics of brain MR images for AD classification.The 3D MBConv block consists of depthwise convolution and squeeze-and-excitation module to transform the input features into more compact features with fewer parameters than standard convolution.We collected MR images of 218 normal controls（NC）,99 stable MCI subjects（s MCI）,80 progressive MCI subjects（p MCI）,and 182 mild AD patients from Alzheimer’s Disease Neuroimaging Initiative（ADNI）database.Our method was evaluated on three classification tasks:（1）NC versus AD,（2）NC versus p MCI,and（3）s MCI versus p MCI.The proposed 3D Efficient Net achieved 95.00% accuracy for classification of NC versus AD,86.67% for NC versus p MCI,and 83.33% for s MCI versus p MCI.Compared with classic convolutional neural network（CNN）and residual neural network（Res Net）as well as some published methods in recent studies,the classification performance of our proposed algorithm was among the top ranks,especially showed improvement in discriminating MCI subjects who were in high risks of conversion to AD.Second,we explored Vision Transformer and MLP-Mixer,which are artificial intelligence algorithms that do not use convolution,to detect or predict AD,and explored the effect of their hyperparameters on their performance。Our proposed 3D Efficient Net,3D Vision Transformer and 3D MLP-Mixer have demonstrated competitive results in multiple experiments,demonstrating their potential to provide a powerful tool for early diagnosis and prediction of AD.