Alzheimer's disease diagnosis based on multiple cluster dense convolutional networks

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) diagnosis relies on identifying neuroanatomical changes via structural magnetic resonance images (MRI).
• Current machine learning methods often require extensive preprocessing like registration and segmentation.
• Deep learning offers potential for automated feature extraction and classification in AD diagnosis.

Purpose of the Study:

• To propose a novel deep learning classification method for Alzheimer's disease (AD) diagnosis using structural MRI.
• To develop a method that reduces reliance on traditional image preprocessing steps.
• To enhance the accuracy of computer-aided diagnosis for AD and mild cognitive impairment (MCI).

Main Methods:

• A classification method based on multiple cluster Dense convolutional neural networks (DenseNets) was developed.
• Brain MRI images were partitioned into local regions, and 3D patches were extracted and clustered using K-Means.
• DenseNets were employed to learn features from clusters, with regional and final classification results ensembled.

Main Results:

• The method achieved 89.5% accuracy and 92.4% AUC for Alzheimer's disease (AD) vs. normal control (NC) classification.
• For mild cognitive impairment (MCI) vs. NC classification, accuracy was 73.8% and AUC was 77.5%.
• The approach demonstrated promising classification performance without rigid registration and segmentation.

Conclusions:

• The proposed multiple cluster DenseNet method effectively classifies Alzheimer's disease (AD) and mild cognitive impairment (MCI) from MRI.
• This deep learning approach offers an efficient alternative to traditional preprocessing-intensive methods.
• The findings highlight the potential of advanced machine learning for neurodegenerative disease diagnosis.