Search

Objectives: Leveraging artificial intelligence (AI) in conjunction with electronic health records (EHRs) holds transformative potential to improve healthcare. Yet, addressing bias in AI, which risks worsening healthcare disparities, cannot be overlooked. This study reviews methods to detect and mitigate diverse forms of bias in AI models developed using EHR data. Methods: We conducted a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines, analyzing articles from PubMed, Web of Science, and IEEE published between January 1, 2010, and Dec 17, 2023. The review identified key biases, outlined strategies for detecting and mitigating bias throughout the AI model development process, and analyzed metrics for bias assessment. Results: Of the 450 articles retrieved, 20 met our criteria, revealing six major bias types: algorithmic, confounding, implicit, measurement, selection, and temporal. The AI models were primarily developed for predictive tasks in healthcare settings. Four studies concentrated on the detection of implicit and algorithmic biases employing fairness metrics like statistical parity, equal opportunity, and predictive equity. Sixty proposed various strategies for mitigating biases, especially targeting implicit and selection biases. These strategies, evaluated through both performance (e.g., accuracy, AUROC) and fairness metrics, predominantly involved data collection and preprocessing techniques like resampling, reweighting, and transformation. Discussion: This review highlights the varied and evolving nature of strategies to address bias in EHR-based AI models, emphasizing the urgent needs for the establishment of standardized, generalizable, and interpretable methodologies to foster the creation of ethical AI systems that promote fairness and equity in healthcare., Comment: Published in JAMIA Volume 31, Issue 5, May 2024

When using machine learning (ML) to aid decision-making, it is critical to ensure that an algorithmic decision is fair and does not discriminate against specific individuals/groups, particularly those from underprivileged populations. Existing group fairness methods aim to ensure equal outcomes (such as loan approval rates) across groups delineated by protected variables like race or gender. However, these methods overlook the intricate, inherent differences among these groups that could influence outcomes. The confounding factors, which are non-protected variables but manifest systematic differences, can significantly affect fairness evaluation. Therefore, we recommend a more refined and comprehensive approach that accounts for both the systematic differences within groups and the multifaceted, intertwined confounding effects. We proposed a fairness metric based on counterparts (i.e., individuals who are similar with respect to the task of interest) from different groups, whose group identities cannot be distinguished algorithmically by exploring confounding factors. We developed a propensity-score-based method for identifying counterparts, avoiding the issue of comparing "oranges" with "apples". In addition, we introduced a counterpart-based statistical fairness index, called Counterpart-Fairness (CFair), to assess the fairness of ML models. Various empirical studies were conducted to validate the effectiveness of CFair., Comment: 24 pages, 9 figures, 14 tables

This presentation will discuss the results of an attempt to use laboratory test data and empirically derived models to quantify the degree of surface damage and associated light scattering that might be expected from hypervelocity particle impacts in space environment. Micrometeoroid sizes are predicted to be predominantly in the mass range 10(exp -5) g or less, with most having diameters near 1 micrometer, but some larger than 120 micrometers, with velocities near 20 km/s. In a laboratory test, latex (p = 1.1 g/cu cm) and iron (7.9 g/cu cm) particles with sizes ranging from 0.75 micrometer to 1.60 micrometer and with velocities ranging from 2.0 km/s to 18.5 km/s, were shot at a dielectric coated gold mirror. Scanning electron and atomic force microscopy were used to measure crater dimensions that were then associated with particle impact energies. These data were then fitted to empirical models derived from solar cell and other spacecraft surface components returned from orbit, as well as studies of impact craters studied on glassy materials returned from the lunar surface, to establish a link between particle energy and impact crater dimension. Published estimates of the Martian and Saturnian meteoroid environments were used as the sources of particle flux estimates. From these data, an estimate of total expected damaged area was computed and this result produced an estimate of expected surface scatter from the modeled environment.

This paper discusses the results of an attempt to use laboratory test data and empirically derived models to quantify the degree of surface damage and associated light scattering that might be expected from hypervelocity particle impacts in space environment. Published descriptions of the interplanetary dust environment were used as the sources of particle mass, size, and velocity estimates. Micrometeoroid sizes are predicted to be predominantly in the mass range 10(exp -5) g or less, with most having diameters near 1 micrometer, but some larger than I20 micrometers, with velocities near 20 kilometers per second. In a laboratory test, latex ( p = 1.1. grams per cubic centimeter) and iron (7.9 grams per cubic centimeter) particles with diameters ranging from 0.75 micrometers to 1.60 micrometers and with velocities ranging from 2.0 kilometers per second to 18.5 kilometers per second, were shot at a Be substrate mirror that had a dielectric coated gold reflecting surface. Scanning electron and atomic force microscopy were used to measure crater dimensions that were then associated with particle impact energies. These data were then fitted to empirical models derived from solar cell and other spacecraft surface components returned from orbit, as well as studies of impact craters studied on glassy materials returned from the lunar surface, to establish a link between particle energy and impact crater dimension. From these data, an estimate of total expected damaged area was computed and this result produced an estimate of expected surface scatter from the modeled environment.

Vacuum brazing is a viable process to achieve strong permanent and hermetic joints in space. As in any joining process, brazed Joints have various imperfections and defects. It is important to understand the impact that flaws have on the load carrying capacity and performance of the brazed joints. This study focuses on the behavior of lap shear joints due to their engineering importance in brazed aerospace structures. In Part 1 an average shear strength capabilitY and failure modes of the single lap joints are explored. Specimens comprised of 0.090 inch thick 347 stainless steel sheet brazed with pure silver are tested in accordance with the AWS C3.2 standard. Comparison of the measured loads and average shear stresses at failure with the Finite Element Analysis (FEA) of the same specimens as a function of the overlap lengths shows excellent correlation between the experimental and calculated values for the defect-free lap joints. In Part 2, various flaws will be introduced in the test specimens to simulate lack of braze coverage in the lap joints. Mechanical testing and FEA will be performed on these joints to verify if behavior of ductile ]appoints containing flaws is similar to the joints with the reduced braze area. Finally, in Part 3, the results obtained in Part 1 and 2 will be applied to the brazed assembly to evaluate a load carrying capability of the structural lap joint containing defects.

High sensitivity is a basic requirement for a new generation of thermal detectors. To meet the requirement, close-packed, two-dimensional silicon detector arrays have been developed in NASA Goddard Space Flight Center. The goal of the task is to fabricate detector arrays configured with thermal detectors such as infrared bolometers and x-ray calorimeters to use in space fliGht missions. This paper focuses on the fabrication and the mechanical testing of detector arrays in a 0.2 mm pixel size, the smallest pop-up detectors being developed so far. These array structures, nicknamed "PUDS" for "Pop-Up Detectors", are fabricated on I pm thick, single-crystal, silicon membranes. Their designs have been refined so we can utilize the flexibility of thin silicon films by actually folding the silicon membranes to 90 degrees in order to obtain close-packed two-dimensional arrays. The PUD elements consist of a detector platform and two legs for mechanical support while also serving as electrical and thermal paths. Torsion bars and cantilevers connecting the detector platform to the legs provide additional flexures for strain relief. Using micro-electromechanical structure (MEMS) fabrication techniques, including photolithography, anisotropic chemical etching, reactive-ion etching, and laser dicing, we have fabricated PLTD detector arrays of fourteen designs with a variation of four parameters including cantilever length, torsion bar length and width, and leg length. Folding tests were conducted to test mechanical stress distribution for the array structures. We obtained folding yields and selected optimum design parameters to reach minimal stress levels. Computer simulation was also employed to verify mechanical behaviors of PUDs in the folding process. In addition, scanning electron microscopy was utilized to examine the flatness of detectors and the alignment of detector pixels in arrays. The fabrication of thermistors and heaters on the pop-up detectors is under way, preparing us for the next step of the experiment, the thermal test.