Program Information
BEST IN PHYSICS (IMAGING): Correlation Between FDG-PET SUVmax and DCE-MRI Microvascular Parameters in Non-Small Cell Lung Cancer (NSCLC)
S Lee*, A Rimner , E Gelb , S Hayes , J Deasy , N Tyagi , Memorial Sloan Kettering Cancer Center, New York, NY
Presentations
WE-F-201-1 (Wednesday, August 2, 2017) 1:45 PM - 3:45 PM Room: 201
Purpose: To assess the relationship between tumor vascularity and metabolic activity measured by DCE-MRI and FDG-PET in NSCLC.
Methods: Pretreatment FDG-PET and DCE-MRI examinations were performed in 18 patients with NSCLC. The maximum standardized uptake value (SUVmax) was measured by FDG-PET. For DCE-MRI, a four-dimensional T1-weighted high resolution imaging with volume excitation (4D THRIVE) sequence with 2-second temporal resolution was performed on a 3T MRI scanner. DCE-MRI data were analyzed using dual-input Tofts-Kety (TK), extended TK, two compartment exchange, adiabatic approximation to the tissue homogeneity (AATH), distributed parameter (DP) models, and their best and weighted model approaches based on Akaike information criterion. Mean, median, standard deviation (SD) and median absolute deviation of total blood flow (BF), pulmonary arterial flow fraction, pulmonary BF, systemic BF, blood volume, mean transit time, permeability-surface area product, fractional interstitial volume (v_I), extraction fraction, volume transfer constant and efflux rate constant (k_ep) were calculated for each model. Univariate analysis (UVA) was performed using Spearman rank-order correlation. Variable selection in regression of SUVmax with DCE-MRI parameters was performed for each model by multivariate analysis (MVA) using least absolute shrinkage and selection operator (LASSO) in combination with leave-one-out cross-validation (LOOCV). The best-fitting model for predicting SUVmax was chosen by LOOCV-based adjusted R².
Results: In UVA, the TK-model-derived mean(k_ep) (ρ=-0.653, p=0.004), the AATH-model-derived median(k_ep) (ρ=0.554, p=0.019) and SD(v_I) (ρ=-0.585, p=0.012), and the DP-model-derived SD(v_I) (ρ=-0.589, p=0.012) were significantly correlated with SUVmax. In MVA, SUVmax was best described by a linear regression model with the TK-model-derived mean(k_ep) (LOOCV-based adjusted R²=0.384, p=0.004); the decrease in k_ep was significantly associated with increase in SUVmax.
Conclusion: Negative correlation was found between the TK-model-derived mean(k_ep) and SUVmax, implying that slower washout time course in the impulse tissue residue function of the dual-input TK model is associated with higher metabolic activity in NSCLC.
Funding Support, Disclosures, and Conflict of Interest: This project was supported from Cycle for Survival and MSK Imaging and radiation science (IMRAS) grants.
Contact Email: