1706P - Radiomics improves response evaluation for desmoid tumours treated with chemotherapy

Abstract

BackgroundResponse of desmoid tumors (DT) to chemotherapy (CT) is evaluated according to RECIST criteria in daily practice and clinical trials. However, MRI demonstrates early change in heterogeneity in responding tumor, due to decrease in cellular area and increase in fibro-necrotic content, before dimensional response. Heterogeneity can be quantified through radiomics approach. Aim was to develop radiomics-based response criteria and to compare their performances with usual criteria. Methods43 patients (28 women, median age: 38.2) were prospectively included in this ancillary multicentre study of a randomized phase II trial (NCT01876082) as they presented with progressive DT, contrast-enhanced MRI at baseline (MRI-0) and early evaluation (3 months later, MRI-1). After signal intensities normalization, voxel size standardization and segmentation of whole DT volume on fat-suppressed contrast-enhanced T1-weighted imaging, 90 baseline and delta 3D-radiomics features (RF) were extracted. Using least absolute shrinkage and selection operator with Cox regression model and cross validation, a radiomics score based on selected and weighted RF was generated. The performances of prognostic models based on radiomics score, RECIST, mRECIST, EASL, Cheson, Choi and modified-Choi criteria from MRI-0 to MRI-1 to predict progression-free survival (PFS) were assessed with concordance-index. All the results were adjusted for ECOG performance status, initial tumor volume, prior CTs, current CT and B-catenin mutation. Usual response criteria were evaluated in both disease control and objective response settings. ResultsThere were 11 progressions. The radiomics score included 4 variables (1 baseline RF and 3 delta-RF). A high radiomics score indicated a poorer prognosis. The radiomics score significantly correlated with PFS (adjusted hazard ratio = 3.39, 95%CI=[1.47-7.79], p=0.0043) while none of the usual response criteria was (range of p-values=[0.166-0.750]). The prognostic model based on radiomics score had the highest concordance-index (0.794, 95%CI=[0.678-0.910]). ConclusionsQuantifying heterogeneity with MRI at baseline and its early change through a dedicated radiomics score can improve the response evaluation for DT treated with CT. Legal entity responsible for the studyAntoine Italiano. FundingHas not received any funding. DisclosureAll authors have declared no conflicts of interest.