A Population-Based Gaussian Mixture Model Incorporating 18F-FDG PET and Diffusion-Weighted MRI Quantifies Tumor Tissue Classes

• DOI: 10.2967/jnumed.115.163972
• Publication Date: 2015-12-11T03:47:35Z
• Authors: Mathew R. Divine; Prateek Katiyar; Ursula Kohlhofer; Leticia Quintanil...; Bernd J. Pichler; Jonathan A. Disse...
• Keywords: Histology

The aim of our study was to create a novel Gaussian mixture modeling (GMM) pipeline model the complementary information derived from¹⁸F-FDG PET and diffusion-weighted MRI (DW-MRI) separate tumor microenvironment into relevant tissue compartments follow development these longitudinally. <b>Methods:</b> Serial ¹⁸F-FDG apparent diffusion coefficient (ADC) maps from DW-MR images NCI-H460 xenograft tumors were coregistered, population-based GMM implemented on imaging data. segmented 3 distinct regions correlated with histology. ANCOVA applied gauge how well total volume predictor for ADC ¹⁸F-FDG, or if good average values in whole necrotic viable tissues. <b>Results:</b> coregistered PET/MR excellent agreement histology, both visually quantitatively, allowed validation last-time-point measurements. Strong correlations found (<i>r</i> = 0.88) fractions 0.87) between histology clustering. provided probabilities each compartment uncertainties expressed as tissues which resolution scans inadequate accurately suggested that (<i>P</i> 0.0009, <i>P</i> 0.02) 0.008, 0.003, 0.01) positive, linear function volume. proved be positive 0.001) 0.0001) <b>Conclusion:</b> longitudinal measurements allows segmentation when using pipeline. Leveraging power multiparametric PET/MRI this manner has potential take assessment disease outcome beyond RECIST could provide an important impact field precision medicine.