An improved method for diagnosis of Parkinson's disease using deep learning models enhanced with metaheuristic...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• Early diagnosis of Parkinson's disease (PD) remains a clinical challenge.
• Medical imaging techniques like MRI and SPECT offer non-invasive quantitative brain health measures.
• Machine and deep learning models are crucial for accurate PD diagnosis using imaging data.

Purpose of the Study:

• To propose and evaluate deep learning models, including a hybrid approach, for the early detection of Parkinson's disease.
• To enhance model performance through hyperparameter optimization using Grey Wolf Optimization (GWO).

Main Methods:

• Four deep learning models and one hybrid model were developed for PD detection.
• Grey Wolf Optimization (GWO) was employed for automatic hyperparameter tuning.
• Models were applied to T1, T2-weighted MRI and SPECT DaTscan datasets.

Main Results:

• All proposed models demonstrated high performance, achieving accuracies near or above 99%.
• The hybrid GWO-VGG16+InceptionV3 model achieved 99.94% accuracy and 99.99% AUC on the T1,T2-weighted dataset.
• The GWO-VGG16+InceptionV3 model achieved 100% accuracy and 99.92% AUC on the SPECT DaTscan dataset.

Conclusions:

• Deep learning models, particularly the hybrid GWO-VGG16+InceptionV3, show exceptional efficacy in early Parkinson's disease detection.
• Optimized deep learning models using GWO significantly enhance diagnostic accuracy for PD using medical imaging.
• These findings suggest a promising avenue for improving early PD diagnosis and patient management.