Design: This is an inter-observer study using a curated sample of 31 anonymized magnetic resonance imaging (MRI) studies comprising 43 leiomyomas independently reviewed by a panel of three fellowship trained radiologists. The preoperative MRIs had been obtained for leiomyoma treatment planning at a single tertiary medical center. The investigators selected studies of small volume uteri containing one to three leiomyomas and ensured an even distribution amongst the various FIGO types; 0 through 8 and hybrids 2-5 and 3-5 lesions. A referent FIGO classification was established by two non-reader experts as determined by consensus. The reader panel reported their impression of the FIGO category of each leiomyoma using an electronic data collection system (Qualtrics) as well as other leiomyoma and uterine metrics. Variations in calculations and inter-reader reliability were assessed using kappa statistics.

Setting: Single tertiary academic institution

Patients or Participants: Three fellowship-trained abdominal/women’s imaging radiologists.

Interventions: N/A

Measurements and Main Results: Among the 43 leiomyomas reviewed, all three readers were in full agreement of FIGO classification in 16 cases (37%). Two reader agreement was noted in 18 cases (42%). The mean agreement was 54% for each individual reader exactly matching with the referent classification [63%, 58%, 42%]. Many of the classifications not in agreement were only one category off, but the clinical significance was not determined. The reasons for misclassification are likely related to reader perception, even among experts. Revised inter-reader reliability data will be presented.

Conclusion: There was a modest level of accuracy among radiologists and the FIGO Leiomyoma subclassification system, with uncertain clinical significance at this time. A revision of the FIGO system, improved instructions for use of the clinical system and continued training of imaging readers may be opportunities for ongoing improvement in future iterations.