Generative adversarial networks for modelling clinical biomarker profiles with race/ethnicity.

AIMS: Modelling biomarker profiles for under-represented race/ethnicity groups are challenging because the underlying studies frequently do not have sufficient participants from these groups. The aim was to investigate generative adversarial networks (GANs), an artificial intelligence technology that enables realistic simulations of complex patterns, for modelling clinical biomarker profiles of under-represented groups. METHODS: GANs consist of generator and discriminator neural networks that operate in tandem. GAN architectures were developed for modelling univariate and joint distributions of a panel of 16 diabetes-relevant biomarkers from the National Health and Nutrition Examination Survey, which contains laboratory and clinical biomarker data from a population-based sample of individuals of all ages, racial groups and ethnicities. Conditional GANs were used to model biomarker profiles for race/ethnicity categories. GAN performance was assessed by comparing GAN outputs to test data. RESULTS: The biomarkers exhibited non-normal distributions and varied in their bivariate correlation patterns. Univariate distributions were modelled with generator and discriminator neural networks consisting of 2 dense layers with rectified linear unit-activation. The distributions of GAN-generated biomarkers were similar to the test data distributions. The joint distributions of the biomarker panel in the GAN-generated data were dispersed and overlapped with the joint distribution of the test data as assessed by 3 multidimensional projection methods. Conditional GANs satisfactorily modelled the joint distribution of the biomarker panel in the Black, Hispanic, White and Other race/ethnicity categories. CONCLUSION: GAN is a promising artificial intelligence approach for generating virtual patient data with realistic biomarker distributions for under-represented race/ethnicity groups.

Generative models for age, race/ethnicity, and disease state dependence of physiological determinants of drug dosing.

Dosing requires consideration of diverse patient-specific factors affecting drug pharmacokinetics and pharmacodynamics. The available pharmacometric methods have limited capacity for modeling the inter-relationships and patterns of variability among physiological determinants of drug dosing (PDODD). To investigate whether generative adversarial networks (GANs) can learn a generative model from real-world data that recapitulates PDODD distributions. A GAN architecture was developed for modeling a PDODD panel comprised of: age, sex, race/ethnicity, body weight, body surface area, total body fat, lean body weight, albumin concentration, glomerular filtration rate (EGFR), urine flow rate, urinary albumin-to-creatinine ratio, alanine aminotransferase to alkaline phosphatase R-value, total bilirubin, active hepatitis B infection status, active hepatitis C infection status, red blood cell, white blood cell, and platelet counts. The panel variables were derived from National Health and Nutrition Examination Survey (NHANES) data sets. The dependence of GAN-generated PDODD on age, race, and active hepatitis infections was assessed. The continuous PDODD biomarkers had diverse non-normal univariate distributions and bivariate trend patterns. The univariate distributions of PDODD biomarkers from GAN simulations satisfactorily approximated those in test data. The joint distribution of the continuous variables was visualized using three 2-dimensional projection methods; for all three methods, the points from the GAN simulation random variate vectors were well dispersed amongst the test data. The age dependence trend patterns in GAN data were similar to those in test data. The histograms for R-values and EGFR from GAN simulations overlapped extensively with test data histograms for the Hispanic, White, African American, and Other race/ethnicity groups. The GAN-simulated data also mirrored the R-values and EGFR changes in active hepatitis C and hepatitis B infection. GANs are a promising approach for simulating the age, race/ethnicity and disease state dependencies of PDODD.

Exploring racial and gender disparities in voice biometrics.

Systemic inequity in biometrics systems based on racial and gender disparities has received a lot of attention recently. These disparities have been explored in existing biometrics systems such as facial biometrics (identifying individuals based on facial attributes). However, such ethical issues remain largely unexplored in voice biometric systems that are very popular and extensively used globally. Using a corpus of non-speech voice records featuring a diverse group of 300 speakers by race (75 each from White, Black, Asian, and Latinx subgroups) and gender (150 each from female and male subgroups), we explore and reveal that racial subgroup has a similar voice characteristic and gender subgroup has a significant different voice characteristic. Moreover, non-negligible racial and gender disparities exist in speaker identification accuracy by analyzing the performance of one commercial product and five research products. The average accuracy for Latinxs can be 12% lower than Whites (p < 0.05, 95% CI 1.58%, 14.15%) and can be significantly higher for female speakers than males (3.67% higher, p < 0.05, 95% CI 1.23%, 11.57%). We further discover that racial disparities primarily result from the neural network-based feature extraction within the voice biometric product and gender disparities primarily due to both voice inherent characteristic difference and neural network-based feature extraction. Finally, we point out strategies (e.g., feature extraction optimization) to incorporate fairness and inclusive consideration in biometrics technology.

Chart Mining: A Survey of Methods for Automated Chart Analysis.

Charts are useful communication tools for the presentation of data in a visually appealing format that facilitates comprehension. There have been many studies dedicated to chart mining, which refers to the process of automatic detection, extraction and analysis of charts to reproduce the tabular data that was originally used to create them. By allowing access to data which might not be available in other formats, chart mining facilitates the creation of many downstream applications. This paper presents a comprehensive survey of approaches across all components of the automated chart mining pipeline, such as (i) automated extraction of charts from documents; (ii) processing of multi-panel charts; (iii) automatic image classifiers to collect chart images at scale; (iv) automated extraction of data from each chart image, for popular chart types as well as selected specialized classes; (v) applications of chart mining; and (vi) datasets for training and evaluation, and the methods that were used to build them. Finally, we summarize the main trends found in the literature and provide pointers to areas for further research in chart mining.

Probabilistic Assessment of Glass Forming Ability Rules for Metallic Glasses Aided by Automated Analysis of Phase Diagrams.

The use of machine learning techniques to expedite the discovery and development of new materials is an essential step towards the acceleration of a new generation of domain-specific highly functional material systems. In this paper, we use the test case of bulk metallic glasses to highlight the key issues in the field of high throughput predictions and propose a new probabilistic analysis of rules for glass forming ability using rough set theory. This approach has been applied to a broad range of binary alloy compositions in order to predict new metallic glass compositions. Our data driven approach takes into account not only a broad variety of thermodynamic, structural and kinetic based criteria, but also incorporates qualitative and descriptive attributes associated with eutectic points in phase diagrams. For the latter, we demonstrate the use of automated machine learning methods that go far beyond text recognition approaches by also being able to interpret phase diagrams. When combined with structural descriptors, this approach provides the foundations to develop a hierarchical probabilistic predication tool that can rank the feasibility of glass formation.