Analyse d'un nouveau modèle radiomic sur l'IRM 2D pour l'évaluation de l'envahissement myométrial profond, l'infiltration lymphovasculaire et l'histologie de haut grade chez les patientes atteintes du cancer de l'endomètre

 

Caroline Reinhold

Centre universitaire de santé McGill [CUSM]

 

Domaine :  cancer

Programme recherche en radiologie

Concours 2017-2018

PURPOSE : To evaluate a novel MRI radiomic model for the assessment of deep myometrial invasion, lymphovascular space invasion and histologic high tumor grade in patients with endometrial carcinoma.

BACKGROUND : Endometrial carcinoma is the most common gynecological cancer in developed countries. The standard treatment is the surgical removal of the uterus and both ovaries. Additional procedures can be benificial
in high-risk cases but increase operating time and morbidity. Patients must therefore undergo accurate
risk assessment to determine the best treatment plans. However, a number of prognostic factors can only be assessed from surgical specimens. In this context, a comprehensive noninvasive diagnostic tool for preoperative risk stratification could have important clinical implications.

MATERIALS AND METHODS : Between January 2011 and December 2015, 212 consecutive patients who had undergone MRI before treatment for endometrial carcinoma at the McGill University Health Center will be included in this retrospective study. Texture analysis of endometrial carcinoma will be assessed quantitatively using a commercially available software (TexRad®). Random forest models will be constructed using various texture features for outcome prediction. The predictive accuracy of the model will be compared with the current gold standard, which is the visual assessments by board-certified radiologists.

ADVANCEMENT OF KNOWLEDGE : We anticipate that the accuracy of predicting deep myometrial invasion applying mathematical modeling will be at least comparable, if not superior, to that achieved by a standard qualitative review by board certified subspecialty radiologists with expertise in gyne-oncological imaging. This model has the potential to change current practice: once validated, it will be able to provide better preoperative risk assessment and therefore better clinical treatment planning for patients with endometrial carcinoma.