OBJECTIVE: To determine the frequency and to assess the interobserver agreement of identification of muscular and ligamentous pelvic floor structures using magnetic resonance imaging. METHOD: Twenty asymptomatic female volunteers (aged 20--80 years old; mean: 50) were submitted to magnetic resonance imaging (1.5 T) examinations of the pelvis. Turbo spin-echo sequences were employed to obtain T1 and T2 weighted images on axial and sagittal planes. Two independent observers evaluated the scans in order to identify the levator ani (coccygeal, pubococcygeal, iliococcygeal and puborectalis muscles), obturatorius internus and urethral sphincter muscles, and the pubovesical and pubourethral ligaments. The frequency and the interobserver agreement of the identification of the anatomical structures were assessed (kappa statistic -- kappa). RESULTS: The frequency of identification of the structures ranged from 50 to 100%, and was slightly lower for identification of the ligaments. Interobserver agreement was as follows: levator ani and obturatorius internus muscle (kappa=1), pubococcygeal (kappa=0.62), iliococcygeal (kappa=0.86), puborectalis (kappa= 0.27), coccygeal (kappa=0) and urethral sphincter muscles (kappa=1), pubovesical (kappa=0.50) and pubourethral (kappa=0.58) ligaments. CONCLUSION: Magnetic resonance imaging of the pelvis allowed precise identification of the main muscular and ligamentous pelvic floor structures in most individuals, whereas interobserver agreement was considered good.

Keywords: Pelvic structures, Magnetic resonance imaging, Pelvic floor