Os06.7.a somatostatin receptor subtype expression and radiomics from dwi-mri represent standardized uptake values (suv) of68ga-dotatoc pet in patients with meningioma, a translational study

BACKGROUND In this translational study, we first aimed to analyse the expression of somatostatin receptor subtype (SSTR) 2A, 2B and 5 and their correlation with the maximum standardized uptake value (SUVmax) of 68Gallium DOTA-D-Phe1-Tyr3-octreotide Positron Emission Tomography (68Ga-DOTATOC PET) in patients with meningioma. Secondly, we aimed to develop a radiomic model based on apparent diffusion coefficient (ADC) maps derived from diffusion weighted magnetic resonance images (DWI MRI) to reproduce SUVmax. MATERIAL AND METHODS A total of 51 patients who underwent MRI and 68Ga-DOTATOC PET prior to surgical resection of intracranial meningioma were included in the imaging part of the study. SUVmax were quantified from the PET images. Segmentation of the post-contrast T1-weighted (cT1) MR images served as regions of interest and were mapped to the co-registered ADC maps. A total of 1940 radiomic features were quantified from the tumor area for each ADC map. A random forest regression model was trained to reproduce SUVmax values. The model was evaluated within a repeated nested cross-validation design. The training included a feature selection was done in two-steps using a minimum redundancy maximum relevance feature selection and a recursive feature elimination based. The expression of SSTR subtypes 2A, 2B and 5 was quantified in 18 surgical specimen of meningioma and compared to SUVmax values. RESULTS The SUVmax values predicted by the random forest models correlated significantly with the observed values (p < 0.05) for all 100 repeats with a mean Pearson’s r = 0.42 ± 0.07 SD and a RMSE = 28.46 ± 0.16. The expression of SSTR subtypes 2A, 2B and 5 correlated significantly with the SUVmax values (p<0.001, R2=0.669; p=0.001, R2=0.393; and p=0.012, R2=0.235, respectively). CONCLUSION In this translational study, SSTR subtypes 2A, 2B and 5 correlate significantly with SUVmax. We developed a random forest regression model using radiomic features to reproduce SUVmax from 68Ga-DOTATOC PET in meningiomas.