Architecture-Aware Augmentation: A Hybrid Deep Learning and Machine Learning Approach for Enhanced Parkinson's...

Area of Science:

• Neuroscience
• Medical Imaging Analysis
• Artificial Intelligence in Healthcare

Background:

• Parkinson's Disease (PD) is a progressive neurodegenerative disorder with millions affected globally.
• Early detection of PD is critical for enhancing patient outcomes and treatment efficacy.
• Spiral drawing analysis offers a non-invasive method for identifying early motor impairments in PD.

Purpose of the Study:

• To evaluate the performance of hybrid deep learning and machine learning models for early Parkinson's Disease detection using spiral drawings.
• To investigate the impact of various data augmentation techniques on model accuracy.
• To compare the effectiveness of different model combinations (ViT-KNN, CNN-SVM, ResNet-50-LR) with and without data augmentation.

Main Methods:

• Comparison of three hybrid models: Vision Transformer (ViT) with K-Nearest Neighbors (KNN), Convolutional Neural Networks (CNN) with Support Vector Machines (SVM), and Residual Neural Networks (ResNet-50) with Logistic Regression.
• Evaluation of model performance on both augmented and non-augmented spiral drawing datasets.
• Application of various data augmentation techniques, including rotation and flipping.

Main Results:

• The ViT-KNN model achieved high initial accuracy (96.77%) but showed decreased performance with data augmentation, indicating reliance on global patterns.
• The ResNet-50-LR model demonstrated consistent performance improvement with data augmentation, reaching 93.55% accuracy with rotation and flipping.
• Different hybrid models exhibited varied responses to data augmentation, emphasizing the importance of technique selection.

Conclusions:

• Hybrid models show potential for early Parkinson's Disease detection via spiral drawing analysis.
• Data augmentation strategies significantly influence the performance of these hybrid models.
• Careful selection and application of augmentation techniques are crucial for optimizing diagnostic tools in medical image analysis for PD.