Two low-pressure variable-reluctance pressure sensors (DP103, diaphragm dash number 10, Validyne Engineering, Los Angeles, CA 91324, USA) were used to acquire TMp vibration. A double-tubing stethoscope (Sprague rappaport stethoscope, ESR-112, Elite Medical Instrument Inc., Fullerton, CA 92831, USA) was connected to a hub of three 4-way stop valves [Fig. 2]. The middle valves were used to test the air tightness of the system by connecting it to a manometer and also to vent the pressure out of the system between data acquisition segments (as described below). By controlling the top 3 stop valves, TMp from both ears can be combined and acquired as one signal or separated and acquired independently by a different pressure transducer. The pressure transducers were mounted on an IV pole similar to clinical settings. Flexible PVC tubes connected pressure transducers and stop valves and were 6 feet long (diameter = 5 mm).
Subject Selection:
15 subjects (age 23-34 years) were recruited for the study after the experimental protocol was approved by University of Central Florida Institutional Review Board. The relatively young population may be preferred as studies have suggested that cochlear aqueduct patency decreases with age [28], which could potentially reduce signal quality. Future studies would consider a larger population with a wider age range.
Data Acquisition:
After obtaining informed consents from the subjects, stethoscope earpieces were securely placed in their ears to acquire the TMp signals. A lubricant (Oto-ease Earmold Lubricant, Westone Laboratories, Colorado Springs, CO 80906, USA) was rubbed on the earpieces before the experiment to ensure a proper seal between the earpieces and the external ear canal. An optical ear lobe pulse sensor (Sparkfun Electronics, Niwot, CO, USA) was clipped to the subject’s earlobe to monitor the ear lobe pulse. A nasal cannula was worn by the subject. The nasal cannula was connected to an end-Tidal CO2 (ETCO2) sensor (MicroCap9, MDPro, San Diego, CA 92117). The output from pressure sensors and earlobe pulse sensor was simultaneously acquired by an iWorx data acquisition system (IX-TA-220, iWorx Systems Inc., Dover, NH 03820, USA) and software (Labscribe version 3.611700) that allowed real-time monitoring of signals. To preserve time- and frequency-domain information, a sampling frequency of 2000 Hz was used. The experiment setup is shown in figure 3.
The air tightness of the whole system was checked at the beginning of each data acquisition session. Here, the tubing from the earpieces, the transducer, and the manometer were all connected together at the hub by properly adjusting all the valves. The venting to atmosphere port was closed and a small pressure (2-3 cm of water) was applied to the system by the syringe. Air tightness was confirmed when this pressure was maintained for one minute. If an air leak were detected, the manometer could be connected to each air space (left or right) separately to isolate the leak source. No leak was detected in any of the study subjects, suggesting the robustness of this system. Once the leak checking was done, the system was vented to the atmosphere by opening the venting valve then closing it once the manometer indicated zero pressure.
At the beginning of each trial, valves were set where each transducer was connected to one ear only to record TMp signals from the two ears separately (Fig 4.a). Subjects were asked to refrain from any movement to help minimize noise in the measured TMp. Baseline TMp and earlobe pulse were recorded for five minutes in the sitting position. Then, valves were set to acquire TMp signals from both ears collectively by a single transducer (Fig 4.b) and the system was vented again to the atmosphere then sealed before data was collected for another 5 minutes. Baseline data was used to select the optimal (i.e., highest signal quality, see equation 1) configuration to be used during testing.
To investigate the effect of ICP changes on the TMp waveform, data was recorded at different tilt angles and with and without hyperventilation (Fig 5). Here, subjects were asked to move to the tilt table and rest at 45° head-up-tilt (HUT) position. Subjects were secured to the table using shoulder straps after adjusting the footrest of the table to match subject height. Feet were kept secured by sliding them between the table’s foam leg roller and the feet holder.
After two minutes of rest, the system was vented to the atmosphere and then sealed again. The signal output from pressure transducer(s) and earlobe pulse sensor and partial pressure reading (in mmHg) from the ETCO2 sensor were recorded at 45° tilt table position for 1 minute. Then the table was tilted to the head-down-tilt (HDT) position (at -45° to -30° depending on the subject’s comfort) to acquire data at the elevated ICP state. Then, data was acquired for 30 seconds after 5 seconds of stabilization time. This was followed by subject hyperventilation while continuously monitoring the ETCO2 sensor reading. When the partial pressure reading dropped by 15-20 mmHg from the initial value (typically after 10-30 sec), subjects were asked to stop hyperventilating and data was recorded for another 30 seconds following 5 seconds of stabilization time at the same position. Then subjects were tilted back to 45° head-up position and, after stabilization for 10-15 seconds, 1 minute of data was acquired from the pressure, earlobe pulse sensor and ETCO2 sensor. All the procedures were performed following relevant regulations and guidelines.
Signal Processing and Segmentation:
The TMp signal preprocessing code was written using a commercial software package (Matlab, Mathworks, Natick, MA) [29]. The raw TMp signal was filtered to remove environmental, electronic and respiratory noises. ICP waveforms are known to have most of their energy below the hearing threshold of 20 Hz [30] and hence, higher frequencies would mostly contain noise. The cochlear aqueduct, which links between ICP and TM pulsations, also acts as a low-pass filter to filter out cardiac and respiration induced pulses and frequencies above 20 Hz [31]. Therefore, a 4th order band pass filter with a passband of 1 to 20 Hz was applied.
To segment TMp cycles, the earlobe pulse was used as a reference signal since it typically shows relatively repeatable waveforms with a clear positive peak. Matlab function ‘findpeaks’ was used to locate the peaks of the earlobe pulse signal as proposed in a previous study [32]. TMp events were identified by choosing a search interval of 0.7s before and after the earlobe pulse peaks. Minima in the search intervals were taken as the beginning and end of each TMp event, with each event corresponding to an earlobe pulse. Individual TMp events were aligned in time and were averaged to reduce noise. The mean TMp waveform was used to represent the TMp signal and extract signal features that may correlate with ICP changes.