TY - JOUR T1 - Reproducibility of FDG-PET radiomic parameters for triple negative breast cancer (TNBC) sub-types JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1757 LP - 1757 VL - 59 IS - supplement 1 AU - Timothy Whitehead AU - Madhusudan Savaikar AU - Lori Strong AU - Amber Salter AU - Jingqin Luo AU - Yu Tao AU - Nicole Fetting AU - Richard Laforest AU - Richard Wahl AU - Shunqiang Li AU - Kooresh Shoghi Y1 - 2018/05/01 UR - http://jnm.snmjournals.org/content/59/supplement_1/1757.abstract N2 - 1757Objectives: ntroduction Numerous recent works highlight the limited utility of established tumor cell lines in recapitulating the heterogeneity of tumors in patients. More realistic preclinical cancer models are thought to be provided by transplantable, patient-derived tumor xenografts (PDX). Importantly, the use of PDX ushers new paradigms involving co-clinical imaging where novel quantitative imaging (QI) methodologies developed and validated using PDX mice can be implemented in a clinical setting, and vice versa. The objective of this work was to assess the reproducibility of QI radiomic metrics in five subtypes of TNBC. Methods: Gene expression analysis of 93 TNBCs (29657 unique genes/probes) was used to determine the correlation to pubslihed expression pattern of 5 TNBC subtypes including 2 basal-like (BL1 and BL2), an immunomodulatory (IM), a mesenchymal (M), a mesenchymal stem-like (MSL), and a luminal androgen receptor (LAR) subtype (see heatmap in fig1). Tumors with the highest correlation to each subtype were used to generate PDX (N=4-12 per subtype, total N=53). Tumors in the range of 100 - 450 mm3 were chosen for imaging. Mice were fasted for four hours prior to imaging with 200μCi/100μL of 18F-FDG via tail vein catheter for a 60min dynamic PET-CT acquisition. Body temperature was monitored and recorded throughout the imaging procedure using a rectal thermometer. After completion of the imaging protocol mice were returned to normal husbandry and standard rodent diet. This protocol was repeated the following day using the same animals, at the same time when possible. A test-retest protocol was implemented for this Purpose: A radiomics approach was used to assess the reproducibility of PET image parameters at 45-60min post-injection including SUVavg, median SUV (SUVmed), SUVmax, SUVpeak, CV as well as muscle and liver normalized parameters. Bland-Altman plots, Lin’s concordance correlation coefficient (CCC), and the interclass correlation coefficient (ICC) were used to assess the repeatability of the various parameters. A CCC > 0.7 was considered acceptable. Stata 12 and MATLAB were used for statistical analysis. Results: SUVmed provided the highest repeatability (CCC=0.86) followed by SUVavg (CCC=0.84) and SUVmax (CCC=0.73). The Bland-Altman plot for the median SUV is shown in fig 2. Liver was shown to be more reliable than muscle for normalization, although the reproducibility (i.e. CCC) was reduced by normalization with both tissues. The data were stratified by PDX subtype, and the analysis repeated for subtypes with sufficient statistics. The analysis generally agreed with overall test-retest results. Muscle normalization resulted in CCC<0.7 for all parameters, and liver normalization resulted in CCC>0.7. Conclusion: We identified 5 TNBC subtypes following gene expression analysis of 93 TNBC tumors for which we generated PDX. Numerous quantitative parameters displayed CCC>0.7. Normalization to muscle or liver generally lowered CCC compared to unnormalized parameters, but liver normalization displayed a CCC>0.7. These results will be used to assess the suitability of the metrics to assess response to therapy using TNBC PDX. ER -