Optimization of activity-driven event detection for long-term ambulatory urodynamics.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Signal Processing

Background:

• Lower urinary tract dysfunction (LUTD) affects millions globally, impacting quality of life.
• Catheter-free implantable devices offer long-term ambulatory bladder monitoring for LUTD diagnosis and treatment.
• Motion artifacts in bladder pressure signals can lead to misclassification and inaccurate diagnosis.

Purpose of the Study:

• To investigate the utility of integrating inertial measurement unit (IMU) data into bladder event classification.
• To assess the effectiveness of digital signal processing (DSP) and machine learning (ML) for activity recognition (AR) and artifact detection.
• To improve the accuracy of bladder event classification in ambulatory monitoring systems.

Main Methods:

• Collected simultaneous bladder pressure and IMU data from an ambulating Yucatan minipig.
• Identified 10 computationally inexpensive signal features for activity classification.
• Employed DSP and ML techniques for signal processing and activity classification.

Main Results:

• Achieved an average 91.5% accuracy in classifying patient activities using IMU data.
• Improved bladder event classification accuracy from 81% to 89.0% when classified activities were integrated.
• Demonstrated that specific IMU features enhance bladder event classification with low computational cost.

Conclusions:

• Activity recognition using IMU data effectively distinguishes bladder contractions from motion artifacts.
• This approach reduces misclassification errors in single-channel bladder event detection systems.
• Findings are relevant for emerging intravesical pressure sensors and ambulatory bladder pressure analysis.