Active tumor boundary in preclinical FDG-PET imaging of TNBC PDX: Optimal definition and discrepancy between multi-parametric MR and caliper measurements

1389 Introduction: Delineation of tumor boundary in preclinical positron emission tomography (PET) is critical for quantitative imaging. Numerous methodologies have been proposed, including threholding methods based on maximum voxel value and automatic segmentation of tumors. The purpose of this study was to optimize the intensity threshold to match the active tumor boundary in simultaneous PET/MR imaging of triple negative breast cancer (TNBC) patient derived tumor xenografts (PDX). A secondary objective was to characterize variability in multi-parametric MR measures of tumor and standard caliper measures. Methods: TNBC PDX were generated through the Washington University in St. Louis co-clinical imaging research resource program. A total of 19 mice at two time points (total of 38 datasets) were imaged. PDX mice were injected with ~200µCi/100µL of FDG via tail vein for a 30min static multi-frame PET image acquisition 30min post injection. In parallel with PET acquisition, a multi-parametric (MP) MR acquisition was performed including T1w, T2w, and apparent diffusion coefficient (ADC) maps. In plane, resolution of T2w and T1w MR Images were 0.25×0.25×1 mm3 and PET images were 0.42×0.42×0.42 mm3. Tumor volumes were segmented manually from T1w, T2w, ADC, PET images individually and on co-registered (T2w, ADC and PET) images. T2w images were used as “true” tumor volume to optimize PET intensity threshold to define active tumor boundaries. PET tumor volume intensity were normalized (Max-Min) to remove bias from outlier intensity. Thresholding technique was applied to discard tumors boundary voxels to match the T2w tumor volume. A flood-fill operation was applied on the necrotic PET tumors to keep low intensity near the tumor center as part of the tumor volume. The PET tumor volume and intensity was also calculated after using a fixed threshold of 25% based on previous reports. Results: An average PET tumor intensity threshold value of 26.9% was optimized to match T2w MR tumor volumes, albeit with high variability (coefficient of variance 37.9%) (Figure 1 (A)). Based on previous reports, the volume determined using the 25% threshold correlates well with the T2w volumes (Figure1 (B)).The correlation between the mean tumor FDG uptake defined by optimal threshold correlates with mean FDG uptake defined by 25% threshold; however, there is 30% bias compared to optimally defined tumor volume (Figure 1 (C)). The linear regression plots of T2w and caliper volume shows that caliper volume over estimates (slope 1.5 and R² = 0.6) the tumor volume compares to T2w volume (Figure 1 (E)). The correlation of T1w, ADC, PET, registered (T2w, ADC and PET), and caliper volume with T2w (Figure (D) displays high variability in multi-parametric MR and caliper measurements. Over estimation becomes much prominent towards large tumor volume. Conclusions: While 25% FDG threshold provides high correlation to T2w-defined tumor volume, our data suggests using single threshold value in determining total tumor mean uptake value. There is high variability in multi-parametric measures of tumor volume suggesting a need for an integrated metric. Caliper volumes generally overestimate MR-defined tumor volume in TNBC PDX, but simple regression model can be used to derive estimated volumes.