Background and Aims: Clinical hepatology research often faces limited data availability, underrepresentation of minority groups, and complex data-sharing regulations. Synthetic data—artificially generated patient records designed to mirror real-world distributions— offers a potential solution. We hypothesized that diffusion models, a state-of-the-art generative technique, could produce synthetic liver transplant waitlist data from the United Network for Organ Sharing (UNOS) database that maintains statistical fidelity, replicates clinical correlations and survival patterns, and ensures robust privacy protection. Methods: Diffusion models were used to generate synthetic patient cohorts mirroring the UNOS liver transplant waitlist database between years 2019 and 2023. Statistical fidelity was assessed using Maximum Mean Discrepancy (MMD) and Wasserstein distance, correlation analysis, and variable-level metrics. Clinical utility was evaluated by comparing transplant-free survival via Kaplan-Meier curves and the MELD score performance. Privacy was quantified using the Distance to Closest Record (DCR) and attribute disclosure risk assessments. Results: The synthetic dataset was nearly indistinguishable from the original dataset (MMD=0.002, standardized Wasserstein distance<1.0), preserving clinically relevant correlations and survival patterns as evidenced by similar median survival times (110 vs. 101 days) and 5-year survival rates (22.2% vs. 22.8%). MELD-based 90-day mortality prediction was maintained (original AUC=0.839 vs. synthetic AUC=0.844). Privacy metrics indicated no identifiable patient matches, and mean DCR values ensured that synthetic individuals were not direct replicas of real patients. Conclusion: AI-generated synthetic data derived from diffusion models can faithfully replicate complex hepatology datasets, maintain key clinical signals, and ensure strong privacy safeguards. This approach can help address data scarcity, enhance model generalizability, foster multi-institutional collaboration, and accelerate progress in hepatology research. Export
5 November 2017 in International Journal of Differential Equations
This paper proposes the morbidity of the multivariable grey prediction MGM model. Based on the morbidity of the differential equations, properties of matrix, and Gerschgorin Panel Theorem, we analyze the factors that affect the morbidity of the multivariable grey model and give a criterion to justify the morbidity of MGM. Finally, an example is presented to illustrate the practicality of our results.
In recent years, data-driven approaches have been developed to predict Alzheimer's Disease. The goal of our report is to identify the factors that are most associated with memory loss. Data from ADNI (Alzheimer's Disease Neuro-imaging Initiative) with 8 variables from socioeconomic factors and biomarkers are analyzed, and will be used to propose a predictive model for the onset of AD. On the mathematical side, we propose to evaluate the data and the results for stability and reproducibility. Principal component analysis was first adopted to directly visualize the data structure and gain intuitive understanding. PCA shows a clustering of biomarkers TAUAB42, APOE4 and the current diagnosis DXCURREN, another clustering of characteristic variables including sex, age and education level, and the response AVLTot by itself. Furthermore, in the second principal component, which signifies a contrast between the characteristic descriptors and the biomarkers APOE4 and TAUAB42. The sex of the respondents played a significant effect: the contrast is larger for males and smaller for females. Linear regression model using backwards elimination further reached the result that the degree of memory loss indicated by AVLTot0 is positively related to education, but negatively related to gender, age, APOE4 and TAUAB420. In addition, higher level of TAUAB420 is related to lower memory function, and this decline effect is more dramatic for people who are younger.
In recent decades, data-driven approaches have been developed to analyze demographic and economic surveys on a large scale. Despite advances in multivariate techniques and learning methods, in practice the analysis and interpretations are often focused on a small portion of available data and limited to a single perspective. This paper aims to utilize a selected array of multivariate statistical learning methods in the analysis of income and expenditure patterns of households in the United States using the Public-Use Microdata from the Bureau of Labor Statistics Consumer Expenditure Survey (CE). The objective is to propose an effective data pipeline that provides visualizations and comprehensive interpretations for applications in governmental regulations and economic research, using thirty-five original survey variables covering the categories of demographics, income and expenditure. Details on feature extraction not only showcase CE as a unique publicly-shared big data resource with high potential for in-depth analysis, but also assist interested researchers with pre-processing. Challenges from missing values and categorical variables are treated in the exploratory analysis, while statistical learning methods are comprehensively employed to address multiple economic perspectives. Principal component analysis suggests that after-tax income, wage/salary income, and the quarterly expenditure in food, housing and overall as the five most important of the selected variables, while cluster analysis identifies and visualizes the implicit structure between variables. Based on this, canonical correlation analysis reveals high correlation between two selected groups of variables, one of income and the other of expenditure.
I have carefully reviewed your submission and have several concerns that need to be addressed before the manuscript can be considered for publication. While the manuscript makes an important contribution to the literature on disaster vulnerability, I...
Background: Prognostic models for alcohol-associated hepatitis (AH) are typically single-centre and generalise poorly, yet the data sharing needed to improve them is limited by privacy regulation. We assessed whether diffusion-generated synthetic data can reproduce the structure, clinical associations, and predictive utility of pooled multi-site AH data while limiting disclosure risk, enabling collaborative modelling without exchanging real patient records.Methods: We assembled four geographically and sociodemographically distinct Mayo Clinic severe AH cohorts (Rochester, Arizona, Florida, and the Mayo Clinic Health System; 2016–2024; N=1477). Pooled real training data (n=1032) trained a tabular denoising diffusion probabilistic model (TabDDPM), generating 10 000 synthetic patients. We evaluated fidelity, survival associations, privacy, and predictive utility; four algorithms trained on synthetic versus real data were tested on a common held-out real set (n=445) for 30-day and 90-day death or liver transplantation. Non-inferiority was prespecified as the upper 95% CI bound of the AUC difference (real minus synthetic) below 0·05.Findings: Baseline characteristics differed across sites (eg, 30-day composite event rate 12·0–21·1%; all p<0·05). Synthetic data reproduced every marginal distribution (p=0·54–1·00) and the correlation structure (Pearson r=0·998), and preserved survival stratification across five risk factors (p≤0·001). Synthetic-trained models were non-inferior to real-trained models for all four algorithms at both horizons in the combined cohort (AUC difference −0·007 to +0·024). Site-level non-inferiority was met in 13/20 and 12/20 comparisons (small per-site samples). Calibration was comparable and empirical disclosure risk was low (membership-inference AUC 0·499; no exact matches).Interpretation: Diffusion-generated synthetic data preserved the structure, clinical associations, and predictive utility of multi-site AH cohorts with low empirical disclosure risk, substituting for—not merely resembling—the real data. Such surrogates are a reusable complement to federated learning, pending validation across independent institutions.