A Dual Stream Deep Learning Framework for Alzheimer's Disease Detection Using MRI Sonification

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's Disease (AD) is a progressive brain disorder impacting millions globally, characterized by memory loss and cognitive decline.
• Magnetic Resonance Imaging (MRI) is a key diagnostic tool, but current methods primarily use visual data, overlooking other potential features.
• Exploring novel methods to enhance AD diagnosis is crucial for early intervention and patient management.

Purpose of the Study:

• To investigate the diagnostic potential of MRI sonification as a complementary approach to conventional image-based methods for Alzheimer's Disease.
• To develop and evaluate a novel dual-stream multimodal framework integrating 2D MRI slices and their audio representations.

Main Methods:

• A novel dual-stream multimodal framework was developed, transforming 2D MRI slices into audio signals using Gabor filtering and a Hilbert space-filling curve.
• The framework processed image and audio modalities using a Convolutional Neural Network (CNN) and YAMNet, respectively.
• Data fusion was achieved through logistic regression to combine features from both modalities.

Main Results:

• The multimodal framework achieved high accuracy in distinguishing between Alzheimer's Disease (AD) and Cognitively Normal (CN) subjects (98.2%).
• The system demonstrated strong performance in differentiating AD from Mild Cognitive Impairment (MCI) (94%) and MCI from CN subjects (93.2%).

Conclusions:

• MRI sonification offers a promising new avenue for extracting complementary diagnostic information from imaging data.
• This approach highlights the potential of audio transformation of imaging data for enhanced feature extraction and classification in neurological disorders.
• The developed multimodal framework shows significant potential for improving the accuracy of Alzheimer's Disease diagnosis.