B.M. Ellingson 1,2, M.G. Malkin 1,3,4, S.D. Rand 1,2, J.M. Connelly 3, C. Quinsey 4 P.S. LaViolette 1,5, D.P. Bedekar 1,2, and K.M. Schmainda 1,2,5 B.M. Ellingson 1,2, M.G. Malkin 1,3,4, S.D. Rand 1,2, J.M. Connelly 3, C. Quinsey 4 P.S. LaViolette 1,5, D.P. Bedekar 1,2, and K.M. Schmainda 1,2,5 Validation of functional diffusion maps (fDMs) as a biomarker for human glioma cellularity Validation of functional diffusion maps (fDMs) as a biomarker for human glioma cellularity 1 Translational Brain Tumor Program, 2 Dept. of Radiology, 3 Dept. of Neurology, 4 Dept. of Neurosurgery, 5 Dept. of Biophysics, Medical College of Wisconsin, Milwaukee, WI 1 Translational Brain Tumor Program, 2 Dept. of Radiology, 3 Dept. of Neurology, 4 Dept. of Neurosurgery, 5 Dept. of Biophysics, Medical College of Wisconsin, Milwaukee, WI INTRODUCTION METHODS  Patient Population: Sixty-nine (n = 69) patients with gliomas were enrolled in the current retrospective study. All patients gave informed written consent according to guidelines approved by the institutional review board at our institution.  Clinical MRI scans included a 3D-SPGR, FLAIR, diffusion, and pre-/post-contrast T1-weighted images on a 1.5T MR scanner (Excite, Cvi, or LX; GE Medical Systems, Waukesha, WI). ADC was calculated from b = 0 and 1,000 s/mm 2 images.  Functional Diffusion Maps (fDMs): All images for each patient were registered to their own baseline, postsurgical, pretreatment, SPGR anatomical images using a mutual information algorithm and a 12-degree of freedom transformation using FSL (FMRIB, Oxford, UK). o Voxel-wise subtraction was then performed between ADC maps acquired at subsequent time points and the baseline, postsurgical, pretreatment, ADC maps to create  ADC images. o Individual voxels were stratified into 3 categories: 1. Hypocellular (Red)  ADC >  ADC thresh 2. Hypercellular (Blue)  ADC < -  ADC thresh 3. No Change (Green) To comprehensively validate fDMs as a biomarker for glioma cellularity and clarify the biological and clinical sensitivity to tumor progression PURPOSE RESULTS REFERENCES 1 Sugahara, JMRI, 1999; 2 Lyng, MRM, 2000; 3 Chenevert, JNCI, 2000; 4 Hayashida, AJNR, 2006; 5 Manenti, Radiol Med, 2008; 6 Gauvain, AJR, 2001; 7 Chenevert, Clin Cancer Res, 1997; 8 Moffat, Proc Nat Acad Sci, 2005; 9 Moffat, Neoplasia, 2006; 10 Hamstra, JCO, 2008; 11 Hamstra, Proc Nat Acad Sci, Figure 1: A) ADC map from a glioblastoma patient showing regions of edema, necrosis, and viabile tumor. B) Cartoon demonstrating how ADC changes for each tissue type within the tumor region. C) Calculation of functional diffusion maps from sequential apparent diffusion coefficient (ADC) maps. Voxels are classified as hypercellular (blue) if ADC decreases, hypocellular (red) if ADC increases, or there is no change (green) beyond a particular  ADC threshold. Acknowledgements: SUMMARY Funding support was provided by NIH/NCI R01CA082500, NIH/NCI R21 CA109280, MCW Translational Brain Tumor Program, Advancing a Healthier Wisconsin and a Fellowship from the MCW Cancer Center. Hypothesis 1: Glioma cell density is inversely proportional to ADC measurement  17 patients (WHO grades II-IV) who underwent closed diagnostic stereotactic biopsy by means of StealthSystem™ surgical navigation were retrospectively examined.  Pre-operative ADC measurements spatially matched to biopsy samples were extracted.  Cell density measurements were taken from H&E-stained slides from biopsy specimens using MetaMorph™ image analysis software.  Linear regression was performed between the mean ADC measurement spatially matched to the biopsy site and the mean cell density obtained from the biopsy for each patient.  Results support the hypothesis that tumor cell density is inversely correlated with ADC measurements in human gliomas (Fig.2A; r 2 = , P<0.0001).  Biological Sensitivity = 2.14x10 -5 [mm 2 /s]/[nuclei/HPF] A B C Hypothesis 2: ADC variability across scan days must be determined to properly set  ADC thresholds  The 95% confidence intervals (C.I.s) for different mixtures of normal-appearing white matter (NAWM), gray matter (NAGM), and cerebrospinal fluid (CSF) were calculated for 69 patients with post-baseline times ranging from 1 wk to 1 yr.  Results suggest no statistically significant differences in  ADC variability over time (Bartlett’s test, P>0.05); however,  ADC variability was dependent on tissue type. Figure 2: Correlation between spatially matched ADC measurements and cell density from stereotactic biopsy samples. A) Postoperative SPGR images showing biopsy location. B) Representative histological images (H&E, x20 mag) showing increasing cellularity with increasing tumor grade (scale bar = 50 um). C) Correlation between mean ADC and mean cell density in nuclei per HPF (Pearson’s correlation coefficient, r 2 = , P<0.0001). Hypothesis 3: fDMs created with different  ADC thresholds reflect different sensitivity and specificity to brain tumor progression  Recurrent glioblastoma patients (n = 5) having similar abnormal FLAIR volumes were used to test whether the 95%C.I.s for  ADC (Hypothesis 2) produced different fDMs than the current  ADC threshold of 0.55 um 2 /ms.  Examination of 50 patients, including 19 patients who eventually showed disease progression, were performed to characterize the clinical sensitivity of fDMs.  ROC analysis was used to determine the sensitivity and specificity of the rate of change in hypercellular volume (in uL/day) for different  ADC thresholds to progressive disease (PD).  Results: The 95% C.I. for NAWM and NAGM were statistically different than the current  ADC threshold. Other  ADC thresholds produced statistically similar volumes of hyper- and hypocellularity.  ROC curves classifying PD from stable disease (SD) using the rate of change in hypercellular volume suggested all thresholds were better than chance (AUC, P<0.001).   ADC = 0.4 um 2 /ms is recommended for best balance of sensitivity and specificity.  The median hypercellular volume correlates with histological grade (Spearman, P=0.0016), while the rate of change in hypercellular volume correlates with Karnofsky Performance Status (Pearson, P<0.0001).  fDMs can be used as a predictive tool for progression free survival (PFS) following chemotherapy and radiotherapy (Log-rank, P=0.0032). High Sensitivity High Specificity NAWM NAGM NAGM+ NAWM NAGM+ NAWM+ CSF Figure 3: Distribution of  ADC for different tissue types, polled across different post-baseline time points (n = 69 patients; post-baseline time points from 1 wk to 1 yr. A) Regions of interest (ROIs) containing mistues of different tissue types used in estimation of  ADC variability.  In neoplasms, a decrease in apparent diffusion coefficient (ADC is believed to reflect an increase in tumor cellularity 1-6, and an increase in ADC is believed to reflect necrosis or decreases in cellularity as a result of successful cytotoxic treatment 2,7.  Functional diffusion maps (fDMs) were developed to exploit these principles on a voxel-wise basis and may be useful for predicting response to chemotherapy and radiotherapy  Despite positive initial results and clinical enthusiasm, a comprehensive validation of the assumptions made when performing fDM analysis has not been performed. Figure 4: Comparison of fDMs for different  ADC thresholds and clinical sensitivity to tumor progression. A) Volume of hypercellularity showing differences in NAWM/NAGM  ADC thresholds compared to standard threshold (n=5). B) Volume of hypocellularity shows differences between NAWM  ADC threshold and standard threshold (n=5). C) ROC curves for different  ADC thresholds (n=33). D) fDM volume characteristics for primary gliomas (n=50). E) Rate of change in hypercellular volume vs. Karnofsky performance status (KPS). F) Kaplan-Meier survival curves for progression-free survival (PFS) in 19 of 50 patients. * P < 0.05; ** P< 0.01; *** P<0.001 ADC is correlated with glioma cell density ADC is correlated with glioma cell density ADC varies with tissue type, but not over time  ADC varies with tissue type, but not over time fDMs are a valid biomarker for human glioma cellularity, fDMs are a valid biomarker for human glioma cellularity, fDM metrics correlate with histological grade, neurological status, and can predict progression-free survival in patients with gliomas fDM metrics correlate with histological grade, neurological status, and can predict progression-free survival in patients with gliomas