Evaluation of Decoding Algorithms for Estimating Bladder Pressure from Dorsal Root Ganglia Neural Recordings

Area of Science:

• Biomedical Engineering
• Neuroscience
• Urology

Background:

• Closed-loop bladder control devices offer potential clinical advantages over open-loop systems.
• Previous research confirmed the feasibility of using single-unit recordings from dorsal root ganglia (DRG) to decode bladder pressure.
• Online sorting of single units is computationally intensive and often unreliable, prompting investigation into multi-unit activity.

Purpose of the Study:

• To investigate the feasibility of utilizing DRG multi-unit recordings for decoding bladder pressure.
• To compare the performance of different algorithms in estimating bladder pressure from DRG multi-unit activity.

Main Methods:

• Seven anesthetized feline experiments were conducted to collect DRG multi-unit and bladder pressure data.
• Various feature selection methods were employed alongside three algorithms: multivariate linear regression, Kalman filter, and nonlinear autoregressive moving average (NARMA).
• Models were trained and validated against bladder pressure data.

Main Results:

• The NARMA model with regularization achieved the most accurate bladder pressure estimation, with a normalized root-mean-squared error (NRMSE) of 17 ± 7%.
• A basic Kalman filter demonstrated the highest correlation with bladder pressure (CC = 0.81 ± 0.13).
• In a chronic feline experiment, a model trained two weeks prior achieved an NRMSE of 10.7% and CC of 0.61.

Conclusions:

• DRG multi-unit recordings can be effectively used to decode bladder pressure, offering a viable alternative to single-unit analysis.
• The NARMA model shows significant potential for robust bladder contraction detection in closed-loop systems.
• Further research is needed to assess the long-term stability of parameters and signals for clinical integration into bladder neuroprostheses.