Semi-Automated Detection of Bladder Neck Funneling and Measurement of Posterior Urethrovesical Angle in Females

Abstract

The pathophysiology of stress urinary incontinence is poorly understood but bladder neck funneling (BNF) and posterior urethrovesical angle (PUVA) enlargement have been implicated. Methods to measure these phenomena are poorly established. The aim of this thesis was to develop and evaluate a semi-automated method to analyze BNF and PUVA from ultrasound images acquired transperineally and test its repeatability and concurrent validity compared to manual segmentation. Agreement between the semi-automated and manual methods was assessed by kappa statistics and intraclass correlation coefficients (ICCs). The repeatability of detection of BNF using the semi-automated approach was almost perfect (ĸC = 1.00 (p<0.001)), while the reliability of semi-automated detection of PUVA was good (ICC(3,1) = 0.860 (0.784 – 0.910)). Concurrent validity of BNF classification was almost perfect (ĸL = 1.00 (p<0.001)), while PUVA estimation was moderate (ICC(2,1) = 0.610 (0.514 – 0.705)). The method presented here is an acceptable proof of concept; further development is recommended.