Richard J Chappell

University of Wisconsin–Madison, Madison, Wisconsin, United States

Are you Richard J Chappell?

Claim your profile

Publications (78)380.53 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: The mild cognitive impairment (MCI) stage of Alzheimer's disease (AD) may be optimal for clinical trials to test potential treatments for preventing or delaying decline to dementia. However, MCI is heterogeneous in that not all cases progress to dementia within the time frame of a trial and some may not have underlying AD pathology. Identifying those MCIs who are most likely to decline during a trial and thus most likely to benefit from treatment will improve trial efficiency and power to detect treatment effects. To this end, using multimodal, imaging-derived, inclusion criteria may be especially beneficial. Here, we present a novel multimodal imaging marker that predicts future cognitive and neural decline from [F-18]fluorodeoxyglucose positron emission tomography (PET), amyloid florbetapir PET, and structural magnetic resonance imaging, based on a new deep-learning algorithm (randomized denoising autoencoder marker, rDAm). Using ADNI2 MCI data, we show that using rDAm as a trial enrichment criterion reduces the required sample estimates by at least five times compared with the no-enrichment regime and leads to smaller trials with high statistical power, compared with existing methods. Copyright © 2015 The Alzheimer's Association. Published by Elsevier Inc. All rights reserved.
    No preview · Article · Jun 2015 · Alzheimer's & dementia: the journal of the Alzheimer's Association
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The role of postoperative radiation (RT) in atypical meningioma remains controversial. We report a retrospective review of outcomes and prognostic factor analysis in 158 patients treated between 2000 and 2010, and extensively review the literature. Following resection, 23 patients received immediate RT, whereas 135 did not. Median progression-free survival (PFS) with and without RT was 59 (range 43-86) and 88 (range 64-123) months. For Simpson grade (G) 1-3 resection, with and without RT, median PFS was 48 (2-80) versus 96 (88-123) months and for Simpson G4, it was 59 (6-86) versus 47 (15-104) months (P = 0.4). The rate of 5-year overall survival (OS) with and without RT was 89% and 83%, respectively. On univariate analysis, Simpson G4 (HR 3.2, P = 0.0006) and brain invasion (HR 2.2, P = 0.03) were significantly associated with progression, whereas age >60 years (HR 9.7, P = 0.002), mitoses >5 per 10 high-power field (0.2, P = 0.0056), and Simpson G4 (HR 2.4, P = 0.07) were associated with higher risk of death. We summarized 22 additional reports, which provide very divergent results regarding the benefit of RT. In our series, adjuvant RT is surprisingly associated with worse PFS and OS, and this is more likely to be due to selection bias of referring tumors with more aggressive characteristics such as elevated Ki-67 and brain invasion for adjuvant RT, rather than a direct causal effect of adjuvant RT. Although there is a trend toward improved PFS with adjuvant RT after subtotal resection, no improvement was noted in OS. Multivariate analysis did not yield statistical significance for any of the factors including Simpson grades of resection, adjuvant RT, or six pathological defining features. The relatively divergent results in the literature are most likely explained by patient selection variability; therefore, randomized trials to adequately address this question are clearly necessary.
    Full-text · Article · Apr 2015 · Journal of cancer research and therapeutics
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Imaging biomarkers of resistance to radiation therapy can inform and guide treatment management. Most studies have so far focused on assessing a single imaging biomarker. The goal of this study was to explore a number of different molecular imaging biomarkers as surrogates of resistance to radiation therapy. Twenty-two canine patients with spontaneous sinonasal tumors were treated with accelerated hypofractionated radiation therapy, receiving either 10 fractions of 4.2 Gy each or 10 fractions of 5.0 Gy each to the gross tumor volume. Patients underwent fluorodeoxyglucose (FDG)-, fluorothymidine (FLT)-, and Cu(II)-diacetyl-bis(N4-methylthiosemicarbazone) (Cu-ATSM)-labeled positron emission tomography/computed tomography (PET/CT) imaging before therapy and FLT and Cu-ATSM PET/CT imaging during therapy. In addition to conventional maximum and mean standardized uptake values (SUVmax; SUVmean) measurements, imaging metrics providing response and spatiotemporal information were extracted for each patient. Progression-free survival was assessed according to response evaluation criteria in solid tumor. The prognostic value of each imaging biomarker was evaluated using univariable Cox proportional hazards regression. Multivariable analysis was also performed but was restricted to 2 predictor variables due to the limited number of patients. The best bivariable model was selected according to pseudo-R(2). The following variables were significantly associated with poor clinical outcome following radiation therapy according to univariable analysis: tumor volume (P=.011), midtreatment FLT SUVmean (P=.018), and midtreatment FLT SUVmax (P=.006). Large decreases in FLT SUVmean from pretreatment to midtreatment were associated with worse clinical outcome (P=.013). In the bivariable model, the best 2-variable combination for predicting poor outcome was high midtreatment FLT SUVmax (P=.022) in combination with large FLT response from pretreatment to midtreatment (P=.041). In addition to tumor volume, pronounced tumor proliferative response quantified using FLT PET, especially when associated with high residual FLT PET at midtreatment, is a negative prognostic biomarker of outcome in canine tumors following radiation therapy. Neither FDG PET nor Cu-ATSM PET were predictive of outcome. Copyright © 2015 Elsevier Inc. All rights reserved.
    Full-text · Article · Mar 2015 · International journal of radiation oncology, biology, physics
  • [Show abstract] [Hide abstract]
    ABSTRACT: Kinetic parameter variability may be sensitive to kinetic model choice, kinetic model implementation or patient-specific effects. The purpose of this study was to assess their impact on the variability of dynamic contrast-enhanced computed tomography (DCE-CT) kinetic parameters. A total of 11 canine patients with sinonasal tumours received high signal-to-noise ratio, test-double retest DCE-CT scans. The variability for three distributed parameter (DP)-based models was assessed by analysis of variance. Mixed-effects modelling evaluated patient-specific effects. Inter-model variability (CVinter ) was comparable to or lower than intra-model variability (CVintra ) for blood flow (CVinter :[4-28%], CVintra :[28-31%]), fractional vascular volume (CVinter :[3-17%], CVintra :[16-19%]) and permeability-surface area product (CVinter :[5-12%], CVintra :[14-15%]). The kinetic models were significantly (P<0.05) impacted by patient characteristics for patient size, area underneath the curve of the artery and of the tumour. In conclusion, DP-based models demonstrated good agreement with similar differences between models and scans. However, high variability in the kinetic parameters and their sensitivity to patient size may limit certain quantitative applications. © 2015 John Wiley & Sons Ltd.
    No preview · Article · Feb 2015 · Veterinary and Comparative Oncology
  • R. Jeraj · T. Bradshaw · L. Forrest · R. Chappell

    No preview · Article · Sep 2014 · International journal of radiation oncology, biology, physics

  • No preview · Article · Sep 2014 · International journal of radiation oncology, biology, physics
  • [Show abstract] [Hide abstract]
    ABSTRACT: Background: Prolonged radiation treatment time (RTT) in head and neck squamous cell carcinoma (HNSCC) is associated with inferior tumor control in patients treated with radiation therapy (RT) alone. However, the significance of prolonged RTT with concurrent chemotherapy is less clear. Methods: We reviewed outcomes for 171 patients with primary HNSCC treated with curative intent RT and concurrent drug therapy from 2001 to 2009. The effects of RTT and other variables on local control and survival were analyzed. Results: Patients with RTT >7 weeks had a significantly increased risk of local failure (hazard ratio [HR], 2.6; p = .018) and death (HR, 1.9 p = .035). These results retained significance even after adjustment for tumor stage (age was not significant). Conclusion: For patients treated with concurrent chemoradiotherapy (chemoRT), prolonged RTT may compromise tumor control as has been established in the setting of RT alone. Symptoms of patients with HNSCC undergoing definitive chemoRT should be managed aggressively to limit treatment interruptions.
    No preview · Article · Aug 2014 · Head & Neck
  • Source

    Preview · Article · Jul 2014 · Alzheimer's and Dementia
  • Source

    Full-text · Article · Jul 2014 · Alzheimer's and Dementia
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Local failure rates after radiation therapy (RT) for locally advanced non-small-cell lung cancer (NSCLC) remain high. Consequently, RT dose intensification strategies continue to be explored, including hypofractionation, which allows for RT acceleration that could potentially improve outcomes. The maximum-tolerated dose (MTD) with dose-escalated hypofractionation has not been adequately defined. Seventy-nine patients with NSCLC were enrolled on a prospective single-institution phase I trial of dose-escalated hypofractionated RT without concurrent chemotherapy. Escalation of dose per fraction was performed according to patients' stratified risk for radiation pneumonitis with total RT doses ranging from 57 to 85.5 Gy in 25 daily fractions over 5 weeks using intensity-modulated radiotherapy. The MTD was defined as the maximum dose with ≤ 20% risk of severe toxicity. No grade 3 pneumonitis was observed and an MTD for acute toxicity was not identified during patient accrual. However, with a longer follow-up period, grade 4 to 5 toxicity occurred in six patients and was correlated with total dose (P = .004). An MTD was identified at 63.25 Gy in 25 fractions. Late grade 4 to 5 toxicities were attributable to damage to central and perihilar structures and correlated with dose to the proximal bronchial tree. Although this dose-escalation model limited the rates of clinically significant pneumonitis, dose-limiting toxicity occurred and was dominated by late radiation toxicity involving central and perihilar structures. The identified dose-response for damage to the proximal bronchial tree warrants caution in future dose-intensification protocols, especially when using hypofractionation.
    Full-text · Article · Oct 2013 · Journal of Clinical Oncology
  • Yunzhi Lin · Menggang Yu · Sijian Wang · Richard Chappell · Thomas F Imperiale
    [Show abstract] [Hide abstract]
    ABSTRACT: Colorectal cancer is the second leading cause of death from cancer in the United States. To facilitate the efficiency of colorectal cancer screening, there is a need to stratify risk for colorectal cancer among the 90% of US residents who are considered "average risk." In this article, we investigate such risk stratification rules for advanced colorectal neoplasia (colorectal cancer and advanced, precancerous polyps). We use a recently completed large cohort study of subjects who underwent a first screening colonoscopy. Logistic regression models have been used in the literature to estimate the risk of advanced colorectal neoplasia based on quantifiable risk factors. However, logistic regression may be prone to overfitting and instability in variable selection. Since most of the risk factors in our study have several categories, it was tempting to collapse these categories into fewer risk groups. We propose a penalized logistic regression method that automatically and simultaneously selects variables, groups categories, and estimates their coefficients by penalizing the $$L1$$-norm of both the coefficients and their differences. Hence, it encourages sparsity in the categories, i.e. grouping of the categories, and sparsity in the variables, i.e. variable selection. We apply the penalized logistic regression method to our data. The important variables are selected, with close categories simultaneously grouped, by penalized regression models with and without the interactions terms. The models are validated with 10-fold cross-validation. The receiver operating characteristic curves of the penalized regression models dominate the receiver operating characteristic curve of naive logistic regressions, indicating a superior discriminative performance.
    No preview · Article · Jul 2013 · Statistical Methods in Medical Research
  • M La Fontaine · L McDaniel · L Kubicek · R Chappell · L Forrest · R Jeraj
    [Show abstract] [Hide abstract]
    ABSTRACT: Purpose: A mixed‐effects model can separate and quantify effects according to randomness and measurable patient parameters such as weight or blood pressure. Using data acquired with a high signal‐to‐noise ratio (SNR) and limited patient motion in test‐retest‐retest scanning, we investigated the impact of various patient parameters on the repeatability of dynamic contrast enhanced CT (DCE‐CT) kinetic analysis. Methods: Eleven canine patients with sinonasal tumors received three pre‐treatment DCE‐CT scans on consecutive days. To ensure sub‐millimeter reproducibility, patients were anesthetized and immobilized with a patient‐specific bite‐block, dental mold, and vacuum mattress. Kinetic analysis was performed using the distributed parameter model to compute values for blood flow (F), blood volume ratio (BVR), transit time, and permeability (PER). The repeatability of each parameter was evaluated by the Bland‐Altman repeatability coefficient. The impact of patient parameters was determined by a hierarchical mixed‐effects model. Results: The repeatability coefficients were calculated for F (45%), BVR (39%), transit time (29%), and PER (37%). Between patients, the patient parameters with significant influence on the kinetic parameters were patient weight, area underneath the curve (AUC) of the artery, and AUC of the vein (p<0.05). However, within a patient, patient weight did not have a prominent bearing on the repeatability as it was found stable with a standard error of measurement (SEM) of 2%. As the AUC of the artery increased by one SEM, F decreased by 24% and BVR by 40%. In addition, PS decreased by 14% when the AUC of the vein increased by one SEM. Conclusion: Despite high SNR and limited patient motion, the AUCs of the artery and vein had a notable impact on the repeatability within patients for F, BVR, and PER. This indicates the need for reliable vessel selection in order to minimize kinetic analysis uncertainty.
    No preview · Article · Jun 2013 · Medical Physics
  • Yunzhi Lin · Richard Chappell · Mithat Gönen
    [Show abstract] [Hide abstract]
    ABSTRACT: The tumor-node-metastasis (TNM) staging system has been the anchor of cancer diagnosis, treatment, and prognosis for many years. For meaningful clinical use, an orderly, progressive condensation of the T and N categories into an overall staging system needs to be defined, usually with respect to a time-to-event outcome. This can be considered as a cutpoint selection problem for a censored response partitioned with respect to two ordered categorical covariates and their interaction. The aim is to select the best grouping of the TN categories. A novel bootstrap cutpoint/model selection method is proposed for this task by maximizing bootstrap estimates of the chosen statistical criteria. The criteria are based on prognostic ability including a landmark measure of the explained variation, the area under the receiver operating characteristic (ROC) curve, and a concordance probability generalized from Harrell's c-index. We illustrate the utility of our method by applying it to the staging of colorectal cancer.
    No preview · Article · May 2013 · Statistical Methods in Medical Research
  • [Show abstract] [Hide abstract]
    ABSTRACT: Arterial stiffness may be associated with cognitive function. In this study, pulse wave velocity (PWV) was measured from the carotid to femoral (CF-PWV) and from the carotid to radial (CR-PWV) with the Complior SP System. Cognitive function was measured by 6 tests of executive function, psychomotor speed, memory, and language fluency. A total of 1433 participants were included (mean age 75 y, 43% men). Adjusting for age, sex, education, pulse rate, hemoglobin A1C, high-density lipoprotein cholesterol, hypertension, cardiovascular disease history, smoking, drinking, and depression symptoms, a CF-PWV>12 m/s was associated with a lower Mini-Mental State Examination score (coefficient: -0.31, SE: 0.11, P=0.005), fewer words recalled on Auditory Verbal Learning Test (coefficient: -1.10, SE: 0.43, P=0.01), and lower score on the composite cognition score (coefficient: -0.10, SE: 0.05, P=0.04) and marginally significantly associated with longer time to complete Trail Making Test-part B (coefficient: 6.30, SE: 3.41, P=0.06), CF-PWV was not associated with Trail Making Test-part A, Digit Symbol Substation Test, or Verbal Fluency Test. No associations were found between CR-PWV and cognitive performance measures. Higher large artery stiffness was associated with worse cognitive function, and longitudinal studies are needed to confirm these associations.
    No preview · Article · Apr 2013 · Alzheimer disease and associated disorders
  • [Show abstract] [Hide abstract]
    ABSTRACT: Environmental tobacco smoke (ETS) exposure has been associated with adverse health outcomes. Our goal was to determine if ETS exposure changed between 1998-2000 and 2003-2005 among participants in the population-based Epidemiology of Hearing Loss Study. ETS exposure was ascertained using a cotinine-validated questionnaire at the 5-year (1998-2000) and 10-year follow-up examinations (2003-2005). Non-smoking participants with data from both visits were included (n=1898; ages 53-96 years at 5-yr follow-up). McNemar's test was used to test differences in ETS exposure overall and in 3 settings: home, work, and social settings. Generalized estimating equations (GEE) were used for multivariate logistic regression models of exposure. The proportion of nonsmokers with no or little ETS exposure increased from 80% to 88% (P< 0.0001). The percent living in a home with no indoor smokers increased from 94% to 97% (P<0.0001). The percent reporting no exposure at work increased from 91% to 95% (P<0.0001). The percent reporting the lowest frequency of social exposure increased from 65% to 77% (P<0.0001). In the GEE model, age was inversely associated with overall ETS exposure (Odds Ratio [OR] per 5 yr= 0.87, 95% CI= 0.81, 0.94), as was education (OR for college vs < high school=0.25, 95% CI=0.16, 0.37), female gender (ORI= 0.41, 95% CI= 0.33, 0.51), and later examination period (OR =0.62, 95% CI= 0.53, 0.73); current employment was positively associated with exposure (OR=1.44, 95% CI=1.14, 1.83). Between the late 1990s and the mid-2000s, ETS exposure in older adults decreased. Decreasing exposures suggest there may be future declines in ETS-related adverse health outcomes.
    No preview · Article · Apr 2013 · WMJ: official publication of the State Medical Society of Wisconsin
  • Yunzhi Lin · Sijian Wang · Richard J Chappell
    [Show abstract] [Hide abstract]
    ABSTRACT: The tumor-node-metastasis staging system has been the lynchpin of cancer diagnosis, treatment, and prognosis for many years. For meaningful clinical use, an orderly grouping of the T and N categories into a staging system needs to be defined, usually with respect to a time-to-event outcome. This can be reframed as a model selection problem with respect to features arranged on a partially ordered two-way grid, and a penalized regression method is proposed for selecting the optimal grouping. Instead of penalizing the L1-norm of the coefficients like lasso, in order to enforce the stage grouping, we place L1 constraints on the differences between neighboring coefficients. The underlying mechanism is the sparsity-enforcing property of the L1 penalty, which forces some estimated coefficients to be the same and hence leads to stage grouping. Partial ordering constraints is also required as both the T and N categories are ordinal. A series of optimal groupings with different numbers of stages can be obtained by varying the tuning parameter, which gives a tree-like structure offering a visual aid on how the groupings are progressively made. We hence call the proposed method the lasso tree. We illustrate the utility of our method by applying it to the staging of colorectal cancer using survival outcomes. Simulation studies are carried out to examine the finite sample performance of the selection procedure. We demonstrate that the lasso tree is able to give the right grouping with moderate sample size, is stable with regard to changes in the data, and is not affected by random censoring.
    No preview · Article · Dec 2012 · Biostatistics
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Carotid atherosclerosis has been suggested to be involved in cognitive decline. The Epidemiology of Hearing Loss Study is a longitudinal study of aging among Beaver Dam residents, WI. In 1998-2000, carotid intima-media thickness (IMT) and plaque were measured by ultrasound; cognitive function was measured by the Mini-Mental State Examination (MMSE). Follow-up examinations were conducted in 2003-2005 and 2009-2010. Incidence of cognitive impairment was defined as an MMSE score <24 or reported physician-diagnosed dementia during the follow-up. In the last examination, five additional cognitive tests were added. The associations of carotid atherosclerosis with incident cognitive impairment and cognitive test performance ten years later were evaluated. A total of 1651 participants (mean age 66.8 years, 41% men) without cognitive impairment at baseline were included in the incidence analysis. IMT was associated with incidence of cognitive impairment after multiple adjustments (hazard ratio: 1.09, p = 0.02 for each 0.1 mm increase in IMT). A total of 1311 participants with atherosclerosis data at baseline had the additional cognitive tests 10 years later. Larger IMT was associated with longer time to complete the Trail-Making Test-part B after multiple adjustments (0.1 mm IMT: 2.3 s longer, p = 0.02). Plaque was not associated with incident cognitive impairment or cognitive test performance 10 years later. In this population-based longitudinal study, carotid IMT was associated with a higher risk of developing cognitive impairment during the 10-year follow-up, and was associated with poorer performance in a test of executive function 10 years later.
    Full-text · Article · Jul 2012 · Atherosclerosis
  • Lyndsay N Kubicek · Songwon Seo · Richard J Chappell · Robert Jeraj · Lisa J Forrest
    [Show abstract] [Hide abstract]
    ABSTRACT: The purpose of our study was to compare setup variation in four degrees of freedom (vertical, longitudinal, lateral, and roll) between canine nasal tumor patients immobilized with a mattress and bite block, versus a mattress alone. Our secondary aim was to define a clinical target volume (CTV) to planning target volume (PTV) expansion margin based on our mean systematic error values associated with nasal tumor patients immobilized by a mattress and bite block. We evaluated six parameters for setup corrections: systematic error, random error, patient-patient variation in systematic errors, the magnitude of patient-specific random errors (root mean square [RMS]), distance error, and the variation of setup corrections from zero shift. The variations in all parameters were statistically smaller in the group immobilized by a mattress and bite block. The mean setup corrections in the mattress and bite block group ranged from 0.91 mm to 1.59 mm for the translational errors and 0.5°. Although most veterinary radiation facilities do not have access to Image-guided radiotherapy (IGRT), we identified a need for more rigid fixation, established the value of adding IGRT to veterinary radiation therapy, and define the CTV-PTV setup error margin for canine nasal tumor patients immobilized in a mattress and bite block.
    No preview · Article · Jun 2012 · Veterinary Radiology & Ultrasound
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Purpose: To quantify associations between pre-radiotherapy and post-radiotherapy PET parameters via spatially resolved regression. Materials and methods: Ten canine sinonasal cancer patients underwent PET/CT scans of [(18)F]FDG (FDG(pre)), [(18)F]FLT (FLT(pre)), and [(61)Cu]Cu-ATSM (Cu-ATSM(pre)). Following radiotherapy regimens of 50 Gy in 10 fractions, veterinary patients underwent FDG PET/CT scans at 3 months (FDG(post)). Regression of standardized uptake values in baseline FDG(pre), FLT(pre) and Cu-ATSM(pre) tumour voxels to those in FDG(post) images was performed for linear, log-linear, generalized-linear and mixed-fit linear models. Goodness-of-fit in regression coefficients was assessed by R(2). Hypothesis testing of coefficients over the patient population was performed. Results: Multivariate linear model fits of FDG(pre) to FDG(post) were significantly positive over the population (FDG(post) ~ 0.17 · FDG(pre), p = 0.03), and classified slopes of RECIST non-responders and responders to be different (0.37 vs. 0.07, p = 0.01). Generalized-linear model fits related FDG(pre) to FDG(post) by a linear power law (FDG(post) ~ FDG(pre)(0.93),p<0.001). Univariate mixture model fits of FDG(pre) improved R(2) from 0.17 to 0.52. Neither baseline FLT PET nor Cu-ATSM PET uptake contributed statistically significant multivariate regression coefficients. Conclusions: Spatially resolved regression analysis indicates that pre-treatment FDG PET uptake is most strongly associated with three-month post-treatment FDG PET uptake in this patient population, though associations are histopathology-dependent.
    Full-text · Article · Jun 2012 · Radiotherapy and Oncology
  • [Show abstract] [Hide abstract]
    ABSTRACT: Background/Aims: Hypercholesterolemia in midlife increases risk for Alzheimer's disease (AD) and contributes to cerebrovascular dysregulation - an early finding in preclinical AD pathology. Statins improve vascular reactivity, but it is unknown if they increase regional cerebral blood flow (CBF) in individuals at risk for AD. Methods: In a randomized, controlled, double-blind pilot study, 16 asymptomatic middle-aged adults with parental history of AD were randomized to atorvastatin or placebo daily for 4 months. At baseline and month 4, regional CBF was measured using arterial spin-labeling magnetic resonance imaging and endothelial function was measured using brachial artery ultrasound. Results: At baseline, participants with low HDL-cholesterol, higher global vascular risk, and greater endothelial dysfunction had reduced regional CBF in areas of the brain related to memory and learning (all p < 0.03). Using voxel-based analysis, 4 months of atorvastatin increased CBF in bilateral hippocampi, fusiform gyrus, putamen and insular cortices compared to placebo. Conclusion: In this pilot study, atorvastatin increased regional CBF in persons at risk for AD. Further research is warranted to confirm whether statins increase CBF in areas of the brain related to memory and learning and whether such perfusion changes are associated with a delay in the onset of AD. Clinical Trial Registration: Identifier: NCT00751907.
    No preview · Article · Nov 2011 · Current Alzheimer research

Publication Stats

3k Citations
380.53 Total Impact Points


  • 1993-2015
    • University of Wisconsin–Madison
      • • Department of Biostatistics and Medical Informatics
      • • Department of Human Oncology
      • • Department of Ophthalmology and Visual Sciences
      Madison, Wisconsin, United States
  • 2006
    • Oregon Health and Science University
      Portland, Oregon, United States
  • 1996
    • Memorial Sloan-Kettering Cancer Center
      • Department of Pediatrics
      New York, New York, United States