Walter R. Witschey
Associate Professor
Research Interests
Make your next move.
Prepare your application
Review drafts before you send them.
Explore related searches
Articles (13)
Purpose Portal hypertension, a major complication of chronic liver disease, leads to significant morbidity and mortality. While portal vein diameter measured on imaging has long been proposed as a non-invasive marker of portal hypertension, normative CT-based reference values and population-level associations remain incompletely characterized. Here, we aim to define contemporary reference values for portal vein diameter on clinically obtained CT and evaluate its associations with demographic, clinical, and imaging factors, as well as its diagnostic performance for portal hypertension. Methods We conducted a retrospective analysis of 20,225 clinically obtained CT scans at a single academic medical center. The main portal vein was automatically segmented using Total Segmentator, and maximum diameter extracted using the Vascular Modeling Toolkit. Associations with demographic and imaging factors were evaluated using linear mixed-effects models; prevalent liver disease and portal hypertension using logistic regression; risk of incident ascites and esophageal varices among participants with liver disease using Cox regression; and invasive hepatic venous pressures using correlation analysis and linear regression. Results The mean portal vein diameter was 12.4 mm (95% CI, 12.37–12.45). Larger diameter was independently associated with male sex (+1.4 mm), higher BMI (+0.11 mm/kg/m 2 ), greater height (+0.04 mm/cm), and older age (+0.05 mm/10 years) (all p < 0.001), and was substantially larger on contrast-enhanced abdomen/pelvis CT (+2.4 mm, p < 0.001). Each 1-mm increase in portal vein diameter was associated with higher odds of prevalent liver disease (OR 1.06; 95% CI, 1.04–1.08) and portal hypertension (OR 1.18; 95% CI, 1.12–1.28). Among individuals with liver disease, greater diameter predicted higher risk of incident esophageal varices (baseline diameter HR 1.50; 95% CI, 1.14–2.08) and ascites (HR per mm increase in diameter 1.06; 95% CI, 1.003–1.12). However, portal vein diameter demonstrated weak to no association with invasively measured hepatic venous pressures. Conclusion In this large, EHR-linked imaging cohort, the mean portal vein diameter on CT was 12.4 mm and varied with demographic and imaging factors. Larger diameter was associated with liver disease, portal hypertension, and subsequent development of varices and ascites, supporting use of portal vein diameter as a pragmatic screening or enrichment tool within multimodal clinical frameworks. Key Results Mean portal vein diameter on routine clinical CT was 12.4 mm (95% CI, 12.37–12.45) and varied with sex, height, BMI, exam type, contrast use, and clinical setting. Each 1-mm increase in portal vein diameter was associated with higher odds of prevalent liver disease (OR 1.06) and portal hypertension (OR 1.18). Among individuals with liver disease, larger portal vein diameter predicted higher risk of incident esophageal varices and ascites, independent of demographic and imaging factors.
Year:
2026
Aims The distribution of abdominal adipose depots and their mechanistic links to type 2 diabetes remain incompletely understood. This study elucidated the relationship between type 2 diabetes presence and quantitative abdominal imaging traits, including hepatic steatosis, liver and spleen size, and adipose distribution, using unenhanced computed tomography (CT) scans from a large-scale, racially diverse, disease-focused medical biobank. Materials and Methods Deep learning algorithms were applied to abdominal CT scans to automatically quantify image-derived phenotypes, including spleen-hepatic attenuation difference (SHAD) for hepatic steatosis, liver and spleen volumes (LV and SV, respectively), visceral and subcutaneous adipose tissue (VAT and SAT, respectively), and visceral-to-subcutaneous fat ratio (VSR). Results Diabetic individuals demonstrated a greater degree of hepatic steatosis and central adiposity than those without diabetes. Liver attenuation was lower (47.6 vs. 52.4 Hounsfield units (HU); lower values indicate greater steatosis), SHAD was higher (−5.41 vs. −8.41 HU; more positive values indicate greater steatosis), and steatosis prevalence increased (38.4% vs. 21.4%) (all p<2.2×10⁻¹⁶). VSR was also elevated (0.64 vs. 0.54, p=5.86×10⁻¹³). These trends remained significant after stratification by sex. Multivariate analyses revealed independent associations of diabetes with SHAD (OR 1.04), LV (OR 1.59), SV (OR 3.95), VAT (OR 1.23), SAT (OR 1.05), and VSR (OR 2.27), after adjusting for age, sex, race, and BMI. Conclusions Hepatic steatosis, hepatomegaly, and visceral adiposity on CT imaging are predictive of type 2 diabetes presence. Notably, VSR showed a stronger association with diabetes than BMI, underscoring how body-fat distribution, rather than overall adiposity, more accurately reflects metabolic disease risk.
Year:
2025
Purpose The purpose of this study was to investigate the effect of magnetization exchange on the measurement of tryptophan and NAD + T 1 relaxation times and to determine the magnetization exchange rates with a two‐spin system model using downfield 1 H MRS spectroscopy at 7 T in human brain. Methods We collected downfield 1 H MRS spectra in the human brain of eight healthy volunteers using a spectrally selective single‐slice sequence (excitation window: 9.7 ± 2 ppm) at 7 T. Alternating selective and broadband saturation recovery experiments were performed to probe magnetization transfer‐dependent changes in T 1 recovery. The apparent T 1 was modeled independently for each experiment for the tryptophan resonance at 10.1 ppm and for the NAD + resonances at 9.3, 9.1, and 8.9 ppm. The magnetization exchange rate was modeled explicitly using a two‐spin model to fit both experiments simultaneously. Results The apparent T 1 relaxation times measured from broadband saturation recovery experiments were significantly longer for each resonance compared to those measured from selective saturation ( p < 0.001). The ratio of broadband to selective T 1 was significantly larger for tryptophan than NAD + (TRP: 19 ± 6 vs. NAD + : 9 ± 3; p < 0.01). Using the two‐spin model, we modeled the T 1 of each resonance (ms): T 1,TRP = 622.3 ± 405.5; T 1,NAD,H2 = 924.9 ± 233.7; T 1,NAD,H6 = 1800.0 ± 985.5; T 1,NAD,H4 = 2057.3 ± 573.5. The chemical exchange rate of tryptophan was 12.3 ± 3.1 Hz; the cross‐relaxation rates of NAD + were (Hz): σ NAD,H2 = 5.7 ± 1.6, σ NAD,H6 = 3.3 ± 0.8, and σ NAD,H4 = 3.0 ± 1.5. The tryptophan exchange rate was significantly faster than the rates for NAD + ( p < 0.001). Conclusion Tryptophan chemical exchange and NAD + cross‐relaxation with water can be quantified in vivo in human brain at 7 T using downfield spectroscopy.
Year:
2025
Structured Abstract Background Malnutrition significantly impacts surgical outcomes yet is difficult to identify preoperatively. Few studies have investigated the association between comprehensive body composition assessment and malnutrition in males and females separately. This study evaluates sex-specific associations between preoperative imaging-derived body composition features and malnutrition in abdominal surgery patients. Methods We retrospectively analyzed patients who underwent computed tomography (CT) scans and elective abdominal surgery at a single institution (2018-2021). Preoperative CT scans were assessed using deep learning to quantify five muscle groups and two fat depots. Malnutrition was diagnosed by registered dietitians using standardized criteria. Sex-specific associations with malnutrition were evaluated using logistic regression. Results Among 1,143 patients (52% female), clinical malnutrition was diagnosed in 20.2% of patients, with prevalence varying by procedure type (3.5-38.2%). Malnutrition was associated with reduced muscle volume for both sexes; in contrast, malnutrition was associated with myosteatosis in 3 of 5 muscle groups for females only. In males, malnutrition was associated with decreased psoas volume (OR 0.59 SD, p<0.01), decreased quadratus lumborum volume (OR 0.59 SD, p<0.01), and reduced erector spinae attenuation (OR 0.66 SD, p=0.048). In females, decreased psoas volume (OR 0.55 SD, p<0.001) and attenuation (OR 0.64 SD, p<0.01) were associated with malnutrition. Both sexes demonstrated increased subcutaneous fat attenuation associated with malnutrition (males: OR 1.51 SD, p<0.01; females: OR 1.73 SD, p<0.001), while increased visceral fat attenuation (OR 1.4 SD, p=0.027) was associated with malnutrition only in females. Conclusions Males and females differ in baseline body composition and features associated with clinical malnutrition. Comprehensive deep learning analysis of muscle and fat characteristics from cross-sectional imaging provides insight into the sex-specific relationships between body composition and malnutrition in the preoperative setting and provides an opportunity for early identification of patients with greater nutrition-related surgical risk.
Year:
2025
Recent advancements in generative artificial intelligence have shown promise in producing realistic images from complex data distributions. We developed a denoising diffusion probabilistic model trained on the CheXchoNet dataset, encoding the joint distribution of demographic data and echocardiogram measurements. We generated a synthetic dataset skewed towards younger patients with a higher prevalence of structural left ventricle disease. A diagnostic deep learning model trained on the synthetic dataset performed comparably to one trained on real data producing an AUROC=0.75(95%CI 0.72-0.77), with similar performance on an internal dataset. Combining real data with positive samples from the synthetic data improved diagnostic accuracy producing an AUROC=0.80(95%CI 0.78-0.82). Subgroup analysis showed the largest performance improvement across younger patients. These results suggest diffusion models can increase diagnostic accuracy and fine-tune models for specific populations.
Year:
2024
Purpose This goal of this study was to optimize spectrally selective 1 H‐MRS methods for large‐volume acquisition of low‐concentration metabolites with downfield resonances at 7 T and 3 T, with particular attention paid to detection of nicotinamide adenine dinucleotide (NAD + ) and tryptophan. Methods Spectrally selective excitation was used to avoid magnetization‐transfer effects with water, and various sinc pulses were compared with a band‐selective, uniform response, pure‐phase (E‐BURP) pulse. Localization using a single‐slice selective pulse was compared with voxel‐based localization that used three orthogonal refocusing pulses, and low bandwidth refocusing pulses were used to take advantage of the chemical shift displacement of water. A technique for water sideband removal was added, and a method of coil channel combination for large volumes was introduced. Results Proposed methods were compared qualitatively with previously reported techniques at 7 T. Sinc pulses resulted in reduced water signal excitation and improved spectral quality, with a symmetric, low bandwidth‐time product pulse performing best. Single‐slice localization allowed shorter TEs with large volumes, enhancing signal, whereas low‐bandwidth slice‐selective localization greatly reduced the observed water signal. Gradient cycling helped remove water sidebands, and frequency aligning and pruning individual channels narrowed spectral linewidths. High‐quality brain spectra of NAD + and tryptophan are shown in 4 subjects at 3 T. Conclusion Improved spectral quality with higher downfield signal, shorter TE, lower nuisance signal, reduced artifacts, and narrower peaks was realized at 7 T. These methodological improvements allowed for previously unachievable detection of NAD + and tryptophan in human brain at 3 T in under 5 min.
Year:
2024
Background Thoracic aortic dissection is a life-threatening condition that often occurs in the presence of aortic dilation. Despite a known association between ascending aortic diameter (AscAoD) and dissection risk, predicting dissection risk remains challenging. Objectives Determine whether common variant genetics can be used to improve identification of individuals most at risk for dissection. Methods A genome wide association study (GWAS)-by-subtraction was performed to characterize the diameter-independent genetics of thoracic aortic dissection by subtracting a GWAS of aortic diameter (AoD) from a GWAS of thoracic aortic aneurysm and dissection (TAAD). A polygenic risk score (PRS) was calculated using the PRS-Continuous Shrinkage statistical package and evaluated for its ability to predict aortic dissection. The primary analytic cohort was Penn Medicine Biobank (PMBB) which comprises volunteers consenting to linkage of health records with biospecimens, including DNA which has undergone genome-wide genotyping; additional analyses were performed in the National Institutes of Health All of Us (AoU) cohort. Results We identified 43 significant genetic risk loci in our GWAS-by-subtraction and derived a “Dissection-PRS.” In the PMBB, the Dissection-PRS associated with an increased risk of prevalent dissection (odds ratio [OR]=2.13 per 1 standard deviation [sd] increase in Dissection-PRS, 95% confidence interval [CI] 1.91 to 2.39, P <0.001), These results were consistent when excluding individuals with pathogenic or likely pathogenic variants in established aortopathy genes. When adjusting for clinical risk factors including ascending aortic diameter, the association of the Dissection-PRS with prevalent dissection attenuated but remained significant (OR=1.62 per 1 sd increase in PRS, 95% CI 1.36 to 1.94, P <0.001). The addition of the PRS to a model containing age, sex, clinical risk factors, and ascending aortic diameter substantially improved model discrimination (base model area under the receiver operator characteristic curve [AUROC]=0.676, 95% CI 0.651 to 0.702; with addition of PRS AUROC=0.723, 95% CI 0.702 to 0.744). Analysis in AoU demonstrated similar findings. Conclusions A common-variant PRS can predict aortic dissection in a diverse population.
Year:
2024
Purpose The purpose of this study was to determine the effect of acute nicotinamide riboside (NR) supplementation on cerebral nicotinamide adenine dinucleotide (NAD + ) levels in the human brain in vivo by means of downfield proton MRS (DF 1 H MRS). Methods DF 1 H MRS was performed on 10 healthy volunteers in a 7.0 T MRI scanner with spectrally selective excitation and spatially selective localization to determine cerebral NAD + levels on two back‐to‐back days: once after an overnight fast (baseline) and once 4 h after oral ingestion of nicotinamide riboside (900 mg). Additionally, two more baseline scans were performed following the same paradigm to assess test–retest reliability of the NAD + levels in the absence of NR. Results NR supplementation increased mean NAD + concentration compared to the baseline (0.458 ± 0.053 vs. 0.392 ± 0.058 mM; p < 0.001). The additional two baseline scans demonstrated no differences in mean NAD + concentrations (0.425 ± 0.118 vs. 0.405 ± 0.082 mM; p = 0.45), and no difference from the first baseline scan (F(2, 16) = 0.907; p = 0.424). Conclusion These preliminary results confirm that acute NR supplementation increases cerebral NAD + levels in healthy human volunteers and shows the promise of DF 1 H MRS utility for robust detection of NAD + in humans in vivo.
Year:
2024
Background Aortic structure impacts cardiovascular health through multiple mechanisms. Aortic structural degeneration occurs with aging, increasing left ventricular afterload and promoting increased arterial pulsatility and target organ damage. Despite the impact of aortic structure on cardiovascular health, three-dimensional (3D) aortic geometry has not been comprehensively characterized in large populations. Methods We segmented the complete thoracic aorta using a deep learning architecture and used morphological image operations to extract multiple aortic geometric phenotypes (AGPs, including diameter, length, curvature, and tortuosity) across various subsegments of the thoracic aorta. We deployed our segmentation approach on imaging scans from 54,241 participants in the UK Biobank and 8,456 participants in the Penn Medicine Biobank. Conclusion Our method provides a fully automated approach towards quantifying the three-dimensional structural parameters of the aorta. This approach expands the available phenotypes in two large representative biobanks and will allow large-scale studies to elucidate the biology and clinical consequences of aortic degeneration related to aging and disease states.
Year:
2024
Purpose To determine the clinical relevance of reduced order model (ROM) aortic hemodynamic imaging-derived phenotypes (IDPs) for a range of flow conditions applied to computed tomography (CT) scan data in the Penn Medicine Biobank (PMBB). Methods The human thoracic aorta was automatically segmented in 3,204 chest CT scans from patients in the Penn Medicine Biobank (PMBB) patients using deep learning. Thoracic aorta anatomic IDPs such as aortic diameter and length were computed. Resistance, and flow boundary conditions, were varied, resulting in 125,000 ROM simulations, producing a fingerprint of aortic hemodynamics IDPs for a range of flow conditions. To determine the clinical relevance of the aortic hemodynamic fingerprint, untargeted phenome wide association studies (PheWAS) for disease conditions were performed using aortic geometries and pulse pressure as IDPs. Results By utilizing patient metadata from the PMBB, the human aortic radius for different age groups over a normalized radius was visualized, showing how the vessel deforms with age, as well as other characteristic geometric information. The average radius of the ascending thoracic aortic data set was 26.6 ± 3.1 mm, with an average length of 310 ± 37 mm. A combination of pathology codes (phecodes) and hemodynamic simulations were utilized to develop a relationship between them, showing a strong relationship between the resulting pulse pressure and diseases relating to aortic aneurysms and heart valve disorders. The average pulse pressure calculated by the model was 22.5 ± 8.5 mmHg, with the maximum pressure modeled by the system being 201 mmHg, with the minimum being 63.6 mmHg. The pulse pressures of the most significant phecodes were examined for patients with and without the condition, showing a slight separation between the two cases. The pulse pressure was also slightly negatively correlated with the calculated tapering angle of the ascending thoracic aorta. Conclusions ROM hemodynamic simulations can be applied to aortic imaging traits from thoracic imaging data in a medical biobank. The derived hemodynamic fingerprint, describing the response of the aorta to a range of flow conditions, shows clinically relevant associations with disease.
Year:
2024
Year:
2024
Year:
2024
Year:
2023
Collaborators (9)
Daniel Rader
Chair
University of Pennsylvania
James Gee
Professor
University of Pennsylvania
Charlene Compher
Professor, Clinician-Educator
University of Pennsylvania
Paul M Titchenell
University of Pennsylvania
Scott M Damrauer
Assistant Professor
University of Pennsylvania Perelman School of Medicine
Michael Levin
Cardiologist
Corporal Michael J. Crescenz VA Medical Center
James Pirruccello
University of California, San Francisco
Hersh Sagreiya
Assistant Professor
University of Pennsylvania
Gregory Tasian
University of Pennsylvania
Make your next move.
Prepare your application
Review drafts before you send them.