Three-dimensional pelvimetry by computed tomography

Abstract

This study was undertaken to assess the agreement of computed tomography (CT) pelvimetry with different postprocessing techniques. CT data sets of 25 patients were retrospectively analysed. There were no CT examinations performed solely for pelvimetry, and there was no radiation exposure for study purposes. Six pelvimetric measurements were obtained by two independent observers in four data sets of each patient, i.e. on biplanar topograms, multiplanar reconstructions of 1-mm slices, volume-rendered images of the same data and volume-rendered images based on 5-mm-thick images. Interobserver agreement and variability were determined by Bland-Altman analysis. A human skeleton was also scanned and measured with the same techniques and by ruler as reference. With a correlation coefficient of 0.98, interobserver agreement was best for assessing 3D volume-rendered images reconstructed from 1-mm-thick slices. Interobserver variability was very good for sagittal outlet and midpelvic diameter, transverse inlet diameter and obstetric conjugate (correlation coefficients 0.96-0.99) but limited for intertuberous and interspinous distance. CT and ruler measurements of the skeleton showed excellent agreement. Pelvimetry can be obtained with low interobserver variability on 3D volume-rendered CT reconstructions. Thus, CT pelvimetry is suitable to gain exact knowledge of pelvic anatomy to identify relevant parameters for dystocia in retrospective studies.