Parkinson’s Disease Monitoring System

Objective: Our system is designed to ensure accurate wrist placement for wearable devices—an essential factor for effective Parkinson’s disease monitoring. Correct device positioning enables reliable capture of movement data, which is critical for assessing motor symptoms and optimizing treatment plans.

Methodology: We developed a deep learning solution using real patient accelerometer data. By training our models on diverse patient datasets, we focused on recognizing proper device positioning. Our approach combines:

• Convolutional Neural Networks (CNNs): For extracting key features from raw accelerometer signals.
• Recurrent Neural Networks (RNNs): Specifically LSTM and GRU architectures to capture temporal movement patterns over time.

This dual approach ensures that our system can differentiate between correct and improper wrist placement in real time.

Solution: The solution implements a real-time verification system that continuously checks wearable positioning during use. When the device deviates from the optimal placement, the system immediately triggers an alert, enabling both patients and caregivers to take corrective action. This proactive monitoring enhances the overall accuracy of the data used for disease management.

Impact:

• 45% Improvement in Data Accuracy: Reliable sensor data leads to better monitoring of motor fluctuations.
• Enhanced Patient Adherence: Real-time feedback encourages proper device usage, ensuring consistent monitoring.
• Proactive Treatment Adjustments: Accurate data enables clinicians to make timely modifications to treatment plans, improving patient outcomes.

Technologies: TensorFlow, PyTorch, Python, LSTM, GRU, AWS