Deep learning-based classification of multi-categorical Alzheimer's disease data

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) poses significant social challenges due to unknown etiopathologies.
• Early detection of mild cognitive impairment (MCI) is crucial for potentially delaying or preventing AD progression.

Purpose of the Study:

• To apply deep learning (DL) techniques for multiclass classification of normal controls, MCI, and AD subjects.
• To evaluate the efficacy of DL models in analyzing comprehensive Alzheimer's Disease Neuroimaging Initiative (ADNI) data.

Main Methods:

• Utilized a multi-categorical dataset from ADNI, including brain imaging, cognitive tests, cerebrospinal fluid biomarkers, ApoE4 status, and age.
• Developed and implemented artificial neural network (ANN) and 1D convolutional neural network (1D CNN) classifiers.
• Performed multiclass classification to differentiate between normal, MCI, and AD patient groups.

Main Results:

• Achieved an overall accuracy of 87.197% with the ANN classifier.
• Obtained a similar overall accuracy of 88.275% with the 1D CNN classifier.
• Demonstrated the capability of DL models to effectively analyze complex ADNI datasets.

Conclusions:

• Deep learning techniques are powerful tools for analyzing ADNI data for early AD and MCI detection.
• Further refinements in DL methodologies are necessary to enhance classification accuracy and clinical utility.
• The study highlights the potential of AI in advancing Alzheimer's disease diagnostics.