Developing and validating a multivariable prediction model to improve the diagnostic accuracy in determination of cervical versus endometrial origin of uterine adenocarcinomas: A prospective MR study combining diffusion-weighted imaging and spectroscopy

Abstract

Background: A triage test to assist clinical decision-making on choosing primary chemoradiation for cervical carcinomas or primary surgery for endometrial carcinomas is important.

Purpose or hypothesis: To develop and validate a multiparametric prediction model based on MR imaging and spectroscopy in distinguishing adenocarcinomas of uterine cervical or endometrial origin.

Study type: Prospective diagnostic accuracy study.

Population: Eighty-seven women: 25 cervical and 62 endometrial adenocarcinomas divided into training (n = 43; cervical/endometrial adenocarcinomas = 11/32) and validation (n = 44; 14/30) datasets.

Field strength/sequence: The 3T diffusion-weighted (DW) MR imaging and MR spectroscopy.

Assessment: Morphology, volumetric DW MR imaging and spectroscopy (MDS) scoring system with total points 0-5, based on presence of the following MR features assessed independently by two radiologists: (a) epicenter at the cervix, (b) rim enhancement, (c) disrupted cervical stromal integrity, (d) mean volumetric apparent diffusion coefficient values (ADCmean) higher than 0.98 × 10-3 mm² /s, (e) fatty acyl δ 1.3 ppm more than 161.92 mM. Histopathology as gold standard.

Statistical tests: Logistic regression and receiver operator characteristic (ROC) curves analysis.

Results: For both the training and validation datasets, the MDS score achieved an accuracy of 93.0% and 84.1%, significantly higher than that of morphology (88.4% and 79.5%), ADC value (74.4% and 68.2%), and spectroscopy (81.4% and 68.2%; P < 0.05 for all). The performances of the scoring were superior to the morphology in the training dataset (areas under the receiver operating characteristics curve [AUC] = 0.95 vs. 0.89; P = 0.046), but not in the validation dataset (AUC = 0.90 vs. 0.85; P = 0.289).

Data conclusion: MDS score has potentials to improve distinguishing adenocarcinomas of cervical or endometrial origin, and warrants large-scale studies for further validation.

Level of evidence: 1 Technical Efficacy: Stage 3 J. Magn. Reson. Imaging 2018;47:1654-1666.

Keywords: cervical carcinoma; diagnostic accuracy; diffusion-weighted imaging; endometrial carcinoma; magnetic resonance imaging; magnetic resonance spectroscopy.