FCN-PD: An Advanced Deep Learning Framework for Parkinson's Disease Diagnosis Using MRI Data.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder impacting motor function and quality of life.
• Current diagnostic methods for PD often involve subjective assessments, leading to delays and inaccuracies.
• Accurate and early diagnosis is crucial for effective Parkinson's disease management.

Purpose of the Study:

• To introduce FCN-PD, a novel deep learning framework for precise Parkinson's disease diagnosis.
• To leverage MRI data for enhanced accuracy in identifying PD.
• To overcome limitations of traditional diagnostic techniques through advanced computational methods.

Main Methods:

• Developed the FCN-PD framework utilizing a hybrid feature extraction approach.
• Integrated EfficientNet for local spatial detail capture and attention mechanisms for global context.
• Employed a Fully Connected Network (FCN) for final classification of MRI data.

Main Results:

• FCN-PD achieved high diagnostic accuracy across three public MRI datasets: 97.2% (PPMI), 95.6% (OASIS), and 96.8% (MIRIAD).
• The framework outperformed traditional Convolutional Neural Network (CNN)-based models by 5.3% on the PPMI dataset.
• Demonstrated superior performance in representing hierarchical features and handling high-dimensional MRI data.

Conclusions:

• FCN-PD offers significant advancements in diagnostic accuracy and efficiency for Parkinson's disease.
• The framework's ability to capture both local and global features makes it a promising tool for clinical application.
• FCN-PD has the potential to facilitate earlier PD detection, leading to improved patient outcomes.