A new intra-vaginal pressure sensing device (FemFit) has been developed to measure the vaginal pressure profile (1). The FemFit utilises a linear array of eight sensors to differentiate between intra-abdominal pressure and pelvic floor pressure within the pressure profile. The overarching aim of the development of the FemFit is to help women manage pelvic floor health and this work aims to investigate the utility of a vaginal pressure profile. The pressure profile has the potential to determine the integrity of the pelvic floor muscles (PFM), as well as providing real time feedback for the user via Bluetooth to a Smart device. The ability to distinguish between PFM activation and abdominal pressure development can guide women not only on the correct performance of a pelvic floor muscle contraction (PFMC), but help with decision making on other exercises to assess if they are pelvic floor safe. The data collected by the FemFit is complex because it involves eight sensors, each measuring pressure at 100Hz. 

The aim of this study was to devise an appropriate methodology to investigate and quantify the repeatability of the pressures measured by the FemFit.

This was a test re-test observational experimental study using the FemFit. A sample of 28 women were invited to participate. Inclusion criteria were: > 18 years; not pregnant; have not had any vaginal surgery; have no symptoms of pelvic organ prolapse or chronic back pain; and who are comfortable using an intra-vaginal device. All participants self-inserted the device with the instruction, ‘insert as you would a tampon.’ Once the device was comfortably in-situ they were asked to perform a series of exercises in the standing position. These included PFMCs and Valsalvas under the instruction of two female researchers. After the completion of the exercises, the participant removed the FemFit to conclude the protocol. After 15 minutes, the participant repeated the protocol. As a result, each participant had two sets of data which are paired for the purposes of investigating and quantifying the repeatability of the FemFit. 

Preprocessing procedures for the data involved identifying the time intervals which correspond to the performed exercises, and transforming the recorded absolute pressure measurements to change in pressure measurements by subtracting the baseline pressure (measured just before an exercise began). A typical example of the pressure trace from sensor four (near the PFM region) is shown in Figure 1A. The duration of an exercise varies, which is also influenced by the researcher’s instruction. This can be corrected by standardising the time component (see Figure 1B). Likewise, the pressure generated varies for the same exercise between and within sessions, which may be due to the placement of the FemFit, participant fatigue, or an improvement in the execution of the exercise by the participant. Standardisation of the pressure generated by the exercise removed the influence of these effects. A Hidden Markov Model (HMM) is a statistical methodology that is suitable for comparing the shape of the two pressure profiles (2). A HMM can facilitate the ability to distinguish the similarities in the shape of the two profiles and quantify this as a proportion. The analyses of the data were done with the software package R version 3.4.3.

Four participants were excluded because of missing data. The identification of the exercise pressure profiles was carried out with statistical learning techniques supported by clinical verification. The proportion of similarity for each pair of exercise pressure profiles was calculated by dividing the standardised pressure profile into sections, identifying the sections in which the shapes were similar and expressing this as a proportion of the total number of sections, given the fitted HMM (see Figure 1B). Hence the proportion of similarity in the shapes of compared profiles is derived from these assignments made by the fitted HMM. Figure 2 visualises the 24 participants' proportions for a PFMC captured by sensor four and a Valsalva captured by sensor seven (a distal sensor).

Figure 1: Representative data from one participant. The solid line represents the pressure trace from the first data set, the dash-line is from the second dataset. A (left) raw pressure (mmHg) traces of a PFMC. B (right). Standardised pressure profile for the same PFMC utilising a HMM to compare the shape. Red indicates similar, blue not similar. 

Figure 2: Box plots of 24 participants’ proportions of similarity for three PFMCs and three Valsalvas.

A proportion was calculated for each participant’s PFMCs and Valsalvas. This is summarised visually in Figure 2. The median of the PFMC‘s proportion of similarity, for the 24 participants, varied between 0.60 and 0.66. The median of the Valsalva‘s proportion of similarity, for the 24 participants, varied between 0.62 and 0.64. The spread for each exercise indicates that there is notable variability in the proportion of similarity for the 24 participants. Further study is required to identify what contributes to the observed variability.

The output produced by this novel method to investigate and quantify the FemFit’s repeatability facilitated the interpretation of the whole pressure trace, rather than one summary number such as a mean or median which may obscure the status of the pelvic floor muscles.

Schell, A, Budgett, D, Nielsen, P, Smalldridge, J, Hayward, L, Dumoulin, C, & Kruger, J. Design and development of a novel intra-vaginal pressure sensor array. In: Annual Meeting of the International-Continence-Society (ICS); 2016.
Shumway, RH, & Stoffer, DS. Time series analysis and its applications: With R examples (Fourth ed.). New York: Springer; 2016. p. 334–345.