Structural MRI-based detection of Alzheimer's disease using feature ranking and classification error

Area of Science:

• Neuroimaging
• Medical Informatics
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) diagnosis relies on identifying structural brain changes.
• Structural magnetic resonance imaging (sMRI) provides detailed brain morphology data.
• Automated systems can aid in the early and accurate detection of AD.

Purpose of the Study:

• To develop an automatic computer-aided diagnosis (CAD) system for Alzheimer's disease detection.
• To utilize feature ranking methods for identifying discriminative features from sMRI data.
• To enhance classification performance through data fusion techniques.

Main Methods:

• Voxel-based morphometry (VBM) was used to analyze gray matter differences between AD patients and healthy controls.
• Seven feature ranking methods (SD, MI, IG, PCC, TS, FC, GI) were employed to extract significant features.
• A support vector machine (SVM) classifier was used, with optimal feature subset determined by classification error estimation.

Main Results:

• The proposed CAD system achieved a classification accuracy of up to 92.48% for AD detection.
• Evaluation was performed on the ADNI dataset, comprising 130 AD patients and 130 healthy controls.
• 10-fold cross-validation demonstrated the robustness of the system's performance.

Conclusions:

• An effective automatic CAD system for AD classification using sMRI and feature ranking was developed.
• The system's performance is comparable to existing state-of-the-art classification models.
• Feature ranking and classification error estimation are crucial for optimal AD detection.