Ingfrid S. Haldorsen MD, Presenter: Nothing to Disclose

Jenny Aase Husby MD, Abstract Co-Author: Nothing to Disclose

Henrica Werner MD, Abstract Co-Author: Nothing to Disclose

Inger Johanne Magnussen MD, Abstract Co-Author: Nothing to Disclose

Jarle Tor Rorvik MD, Abstract Co-Author: Nothing to Disclose

Harald Helland MD, Abstract Co-Author: Nothing to Disclose

Jone Trovik MD, Abstract Co-Author: Nothing to Disclose

Ansgar Espeland MD, PhD, Abstract Co-Author: Nothing to Disclose

Helga Salvesen MD, PhD, Abstract Co-Author: Nothing to Disclose

To evaluate agreement between radiologists on pelvic MR imaging findings of deep myometrial invasion, cervical stroma invasion, and lymph node metastases in endometrial carcinomas, and to compare these imaging findings with final histopathological results.

We included 57 histologically confirmed endometrial carcinoma patients in a study of preoperative pelvic 1.5T MR imaging after informed consent and IRB approval. The protocol included axial oblique T1-weighted sequences before and after i.v. contrast injection (2 min delay) and sagittal and axial oblique T2-weighted images. Four radiologists (three consultants with experience in pelvic imaging and one resident), blinded to clinical and histopathological data, independently reviewed the images for presence of deep myometrial invasion, cervical stroma invasion, and lymph node metastases. We calculated kappa-coefficients [with 95% confidence intervals] for interobserver agreement, and sensitivity and specificity values for each observer using histopathological data from surgical specimen as reference standard.

Overall agreement between all observers was fair (κ= 0.39 [0.25-0.53]) for deep myometrial invasion and moderate for cervical stroma invasion (κ=0.51 [0.25-0.73]) and lymph node metastases (κ=0.55 [0.15-0.83]). Pairwise agreement between resident and consultants was similar to pairwise agreement between consultants. Sensitivity (specificity) values for the four observers ranged from 72-92% (39-63%) for deep myometrial invasion, 38-63% (82-94%) for cervical stroma invasion, and 25-38% (89-100%) for lymph node metastases.

Pelvic MR imaging showed only modest interobserver agreement and diagnostic accuracy for deep myometrial invasion, cervical stroma invasion, and lymph node metastases in patients with endometrial carcinomas. Such patients need improved pre-operative imaging methods.

A considerable interobserver variation exists in treatment relevant preoperative staging evaluation of endometrial carcinomas based on MR imaging, even among experienced radiologists.

Haldorsen, I, Husby, J, Werner, H, Magnussen, I, Rorvik, J, Helland, H, Trovik, J, Espeland, A, Salvesen, H, Modest Agreement between Radiologists on Pelvic MR Imaging Findings in Endometrial Carcinomas.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11005107.html