| Failure of the storage and voiding functions of the urinary bladder due to spinal cord injury (SCI), neural diseases, health conditions, or aging, causes serious complications in a patient's health. Currently, it is possible to partially restore bladder functions in drug-refractory patients using implantable neurostimulators. Improving the efficacy and safety of these neuroprostheses used for bladder functions restoration requires a bladder sensor (BS) capable of detecting urine volume in real-time to implement a closed-loop system that applies electrical stimulation only when required. The BS can also trigger an early warning to advise patients with impaired sensations when the bladder should be voided or when an abnormally high post-voiding residual volume remains after an incomplete voiding. In this thesis, we present new measurement methods and a dedicated digital signal processor for real-time decoding of the bladder volume through afferent neural signals arising from natural receptors present in the bladder wall. The main contributions of this thesis have been reported in three peer-reviewed journal papers.;We first present a comprehensive literature review, including papers, patents and mainstay books of bladder anatomy, physiology, and pathophysiology. This review allowed us to identify the requirements (user needs) that a BS must meet for chronic applications, such as those proposed in this thesis. An exhaustive analysis of the particular anatomical and physiological characteristics of the bladder, which we realized had influenced or prevented the achievement of a BS for monitoring the bladder volume or pressure in past studies, are also presented. Based on this study and on a systematic assessment of the measurement methods published in past years, we determined the best measurement principle for chronic bladder volume monitoring: the detection, discrimination and decoding of the afferent neural activity stemming from specialized volume receptors (mechanoreceptors), on which some authors had hypothesized about its existence in the bladder inner mucosa.;Next, we present methods that allows for a real-time estimation of bladder volume through the afferent activity of the bladder mechanoreceptors. Our method was validated with data acquired from anesthetized rats in acute experiments. It was possible to qualitatively estimate three states of bladder fullness in 100% of trials when the recorded afferent activity exhibited a Spearman's correlation coefficient of 0.6 or better. Furthermore, we could quantitatively estimate the bladder volume, and also its pressure, using time-windows of properly chosen duration. The mean volume estimation error was 5.8 +/- 3.1%. Our results also allowed us to shed light on the controversial subject of the type of responses that are detectable from bladder afferent recordings. We demonstrated that it is possible to quantify not only phasic but also tonic bladder responses during slow filling and isovolumetric measurements, respectively.;Finally, we present a dedicated digital signal processor (DSP) capable of monitoring the bladder volume running the proposed qualitative and quantitative measurement methods. The DSP performs real-time detection and discrimination of extracellular action potentials (on-the-fly spike sorting) followed by neural decoding to estimate either three qualitative levels of fullness or the bladder volume value, depending on the selected output mode. The proposed DSP was tested using both realistic synthetic signals with a known ground-truth and real signals from bladder afferent nerves recorded during acute experiments with animal models. The spike-sorting processing circuit yielded an average accuracy of 92% using signals with highly correlated spike waveforms and low signal-to-noise ratios. The volume estimation circuits, which were tested with real signals, reproduced the accuracies achieved by offline simulations in Matlab, i.e., 94% and 97% for quantitative and qualitative estimations, respectively. To assess feasibility, the DSP was deployed in the Actel FPGA Igloo AGL1000V2, which showed a power consumption of 0.5 mW and a latency of 2.1 ms at a 333 kHz core frequency. These performance results demonstrate that an implantable bladder sensor that detects, discriminates and decodes afferent neural activity is feasible. |
