Segmentations of MRI images of the female pelvic floor: A study of inter- and intra-reader reliability

Lennox Hoyte, Wen Ye, Linda Brubaker, Julia R. Fielding, Mark E. Lockhart, Marta E. Heilbrun, Morton B. Brown, Simon K. Warfield

Research output: Contribution to journalArticlepeer-review

40 Scopus citations


Purpose: To describe the inter- and intra-operator reliability of segmentations of female pelvic floor structures. Materials and Methods: Three segmentation specialists were asked to segment out the female pelvic structures in 20 MR datasets on three separate occasions. The STAPLE algorithm was used to compute inter- and intra-segmenter agreement of each organ in each dataset. STAPLE computed the sensitivity, specificity, and positive predictive values (PPV) for inter- and intra-segmenter repeatability. These parameters were analyzed using intra-class correlation analysis. Correlation of organ volume to PPV and sensitivity was also computed. Results: Mean PPV of the segmented organs ranged from 0.82 to 0.99, and sensitivity ranged from 33 to 96%. Intra-class correlation ranged from 0.07 to 0.98 across segmenters. Pearson correlation of volume to sensitivity were significant across organs, ranging from 0.54 to 0.91. Organs with significant correlation of PPV to volume were bladder (-0.69), levator ani (-0.68), and coccyx (-0.63). Conclusion: Undirected manual segmentation of the pelvic floor organs are adequate for locating the organs, but poor at defining structural boundaries.

Original languageEnglish (US)
Pages (from-to)684-691
Number of pages8
JournalJournal of Magnetic Resonance Imaging
Issue number3
StatePublished - Mar 2011


  • Intraclass correlation
  • MRI
  • Pelvic floor muscles
  • Positive predictive value
  • Repeatability
  • Segmentation

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging


Dive into the research topics of 'Segmentations of MRI images of the female pelvic floor: A study of inter- and intra-reader reliability'. Together they form a unique fingerprint.

Cite this