Abstract:
Pelvic floor muscle exercises, are an important component of women's health in general. The FemFit® a new intra-vaginal pressure sensor array has been developed by the Auckland Bioengineering Institute, to accurately assess pelvic floor muscle function. A pilot study was conducted to assess the reliability of the FemFit® pressure measurements generated by an pelvic floor exercise. There was a need to establish the necessary computational and statistical methods to enable analysis of the pressure data produced by the FemFit®, then to apply these methods to the data before analysing the repeatability of the FemFit®. The statistical methods were designed to segment the FemFit® data into baselines and events, transform the FemFit® data to interpret the pressure measurements as a change from baseline, and a proposal to quantify the repeatability of the pressure data from the FemFit® as a proportion. The segmentation utilised the data's first principal component to identify partitions within it, and naively labels these partitions as baselines or events based on characteristics of the data. The transformation methods produced the anticipated output, yet there was concern over the validity of the statistical framework used to create it. The proposal for the repeatability analysis utilised Hidden Markov Models to make inference about the FemFit's® reliability in capturing the same exercise pressure trace between sessions. The output was in the form of the proportion of similarity, which could be practically interpreted in a clinical setting. Therefore, the practical interpretation of the results supported that the FemFit® was in fact reliable. In conclusion, the research output generated has provided statistical inference to answer the questions posed by the pilot study.