Article Text

Download PDFPDF
376 MRI-based radiomic features for the prediction of mmr status in endometrial cancer: first analyses of a prospective trial
  1. Alberto Farolfi1,
  2. Danila Diano2,
  3. Giacomo Feliciani3,
  4. Giorgia Gurioli2,
  5. Daniela Montanari3,
  6. Marianna Molari2,
  7. Alessia Filograna4,
  8. Andrea Amadori5,
  9. Paolo Maniglio6,
  10. Isabella Strada5,
  11. Alice Rossi1,
  12. Luca Savelli5,
  13. Domenico Barone1 and
  14. Ugo De Giorgi2
  1. 1Oncology Department, Meldola, Italy
  2. 2IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST), Meldola, Italy
  3. 3Radiology Department, Meldola, Italy
  4. 4Biolab Department, Meldola, Italy
  5. 5IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST), Forlì, Italy
  6. 6Oncology Department, Forlì, Italy


Introduction/Background Molecular profiling of endometrial cancer has entered the staging classification and provides information on the treatment decision. In a prospective study we are validating magnetic resonance imaging (MRI)-based radiomics models for the prediction of histopathological and molecular characteristics in early-stage endometrial cancer (EC) patients.

Methodology Patients with newly diagnosed EC were prospectively enrolled and underwent to 3-T MR imaging before surgery. The same day of MRI, patients underwent plasma sample collection for the analysis of cell-free DNA. An experienced radiologist contoured the tumor and Radiomics features were extracted with an IBSI compliant software. Radiomic features were associated with histology, grade, deep myometrial invasion (DMI), lymphovascular space invasion (LVSI), and molecular profiling through Mann-Whitney U statistical test and removed redundancy with Pearson ρ.

Results Among 17 patients enrolled (mean age 74, range 25–91), 13 (76.5%) patients had an endometrioid hystology, 4 (23.5%) patients had a high grade tumor, 5 (29.4%) presented LVSI and 6 (35.3%) patients had a MMR deficient (MMRd) tumor. Cell-free DNA was detectable in all patients, with a median value of 0.7 ng/µl. No differences were observed between median cell-free DNA values in terms of grade, LVSI or MMR status. Among radiomics features, first-order features such ADC variance and maximum intensity showed a trend for the prediction of grade with p-value of 0.07 and 0.09 respectively at U test, but not for LVSI. ADC median value was instead associated with MMRd with a p-value of 0.05.

Conclusion MRI-based radiomics has great potential in developing advanced prognostication and molecular information. Our preliminary results demonstrate that MRI-based radiomics might predict MMR status in patients with early EC, and we plan to expand this work in larger cohorts.

Disclosures A.F. has received personal honoraria for lectures from Astrazeneca, GSK-Tesaro, Clovis, and advisory board from Jannsen, Astrazeneca, GSK-Tesaro.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.