Applying XAI to an AI-based system for candidate management to mitigate bias and discrimination in hiring.

Assuming that potential biases of Artificial Intelligence (AI)-based systems can be identified and controlled for (e.g., by providing high quality training data), employing such systems to augment human resource (HR)-decision makers in candidate selection provides an opportunity to make selection processes more objective. However, as the final hiring decision is likely to remain with humans, prevalent human biases could still cause discrimination. This work investigates the impact of an AI-based system's candidate recommendations on humans' hiring decisions and how this relation could be moderated by an Explainable AI (XAI) approach. We used a self-developed platform and conducted an online experiment with 194 participants. Our quantitative and qualitative findings suggest that the recommendations of an AI-based system can reduce discrimination against older and female candidates but appear to cause fewer selections of foreign-race candidates. Contrary to our expectations, the same XAI approach moderated these effects differently depending on the context. Supplementary Information: The online version contains supplementary material available at 10.1007/s12525-022-00600-9.

Thermal, Mechanical and Biocompatibility Analyses of Photochemically Polymerized PEGDA250 for Photopolymerization-Based Manufacturing Processes.

Novel fabrication techniques based on photopolymerization enable the preparation of complex multi-material constructs for biomedical applications. This requires an understanding of the influence of the used reaction components on the properties of the generated copolymers. The identification of fundamental characteristics of these copolymers is necessary to evaluate their potential for biomaterial applications. Additionally, knowledge of the properties of the starting materials enables subsequent tailoring of the biomaterials to meet individual implantation needs. In our study, we have analyzed the biological, chemical, mechanical and thermal properties of photopolymerized poly(ethyleneglycol) diacrylate (PEGDA) and specific copolymers with different photoinitiator (PI) concentrations before and after applying a post treatment washing process. As comonomers, 1,3-butanediol diacrylate, pentaerythritol triacrylate and pentaerythritol tetraacrylate were used. The in vitro studies confirm the biocompatibility of all investigated copolymers. Uniaxial tensile tests show significantly lower tensile strength (82% decrease) and elongation at break (76% decrease) values for washed samples. Altered tensile strength is also observed for different PI concentrations: on average, 6.2 MPa for 1.25% PI and 3.1 MPa for 0.5% PI. The addition of comonomers lowers elongation at break on average by 45%. Moreover, our observations show glass transition temperatures (Tg) ranging from 27 °C to 56 °C, which significantly increase with higher comonomer content. These results confirm the ability to generate biocompatible PEGDA copolymers with specific thermal and mechanical properties. These can be considered as resins for various additive manufacturing-based applications to obtain personalized medical devices, such as drug delivery systems (DDS). Therefore, our study has advanced the understanding of PEGDA multi-materials and will contribute to the future development of tools ensuring safe and effective individual therapy for patients.

Establishment of a New Device for Electrical Stimulation of Non-Degenerative Cartilage Cells In Vitro.

In cell-based therapies for cartilage lesions, the main problem is still the formation of fibrous cartilage, caused by underlying de-differentiation processes ex vivo. Biophysical stimulation is a promising approach to optimize cell-based procedures and to adapt them more closely to physiological conditions. The occurrence of mechano-electrical transduction phenomena within cartilage tissue is physiological and based on streaming and diffusion potentials. The application of exogenous electric fields can be used to mimic endogenous fields and, thus, support the differentiation of chondrocytes in vitro. For this purpose, we have developed a new device for electrical stimulation of chondrocytes, which operates on the basis of capacitive coupling of alternating electric fields. The reusable and sterilizable stimulation device allows the simultaneous use of 12 cavities with independently applicable fields using only one main supply. The first parameter settings for the stimulation of human non-degenerative chondrocytes, seeded on collagen type I elastin-based scaffolds, were derived from numerical electric field simulations. Our first results suggest that applied alternating electric fields induce chondrogenic re-differentiation at the gene and especially at the protein level of human de-differentiated chondrocytes in a frequency-dependent manner. In future studies, further parameter optimizations will be performed to improve the differentiation capacity of human cartilage cells.

A Novel Hybrid Additive Manufacturing Process for Drug Delivery Systems with Locally Incorporated Drug Depots.

Here, we present a new hybrid additive manufacturing (AM) process to create drug delivery systems (DDSs) with selectively incorporated drug depots. The matrix of a DDS was generated by stereolithography (SLA), whereas the drug depots were loaded using inkjet printing. The novel AM process combining SLA with inkjet printing was successfully implemented in an existing SLA test setup. In the first studies, poly(ethylene glycol) diacrylate-based specimens with integrated depots were generated. As test liquids, blue and pink ink solutions were used. Furthermore, bovine serum albumin labeled with Coomassie blue dye as a model drug was successfully placed in a depot inside a DDS. The new hybrid AM process makes it possible to place several drugs independently of each other within the matrix. This allows adjustment of the release profiles of the drugs depending on the size as well as the position of the depots in the DDS.

How to perform indexing of extravascular lung water: a validation study.

BACKGROUND: Extravascular lung water is a quantitative marker of the amount of fluid in the thoracic cavity besides the vasculature. Indexing to both predicted and actual body weight have been proposed to compare different individuals and provide a uniform range of normal. OBJECTIVE: We explored extravascular lung water measured by single-indicator transpulmonary thermodilution in a large cohort of patients without cardiopulmonary instability, in order to evaluate current and alternative indexing methods. DESIGN: Prospective, observational. SETTING: Neurosurgical ICU in a tertiary referral academic teaching hospital. PATIENTS: One hundred and one consecutive patients requiring elective brain tumor surgery and postoperative ICU surveillance. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Indexed to predicted body weight, females had a mean extravascular lung water of 9.1 (SD=3.1, range: 5-23) mL/kg and males of 8.0 (SD=2.0, range: 4-19) mL/kg (p<0.001). Values indexed to predicted body weight were inversely correlated with the patient's height (p<0.001). Indexed to the traditionally used actual body weight, data showed a significant relationship to weight (p<0.001) and gender (p<0.05). In contrast, indexing to body height presented a method without dependencies on height, weight, or gender, yielding a uniform 95% confidence interval of 218-430 mL/m. Extravascular lung water increased with positive perioperative fluid balance (p=0.04). CONCLUSIONS: Using either predicted or actual body weight for indexing extravascular lung water does not lead to independence of height, weight, and gender of the patient. Specifying a fixed range of normal or a uniform upper threshold for all patients is misleading for either method, despite widespread use. Our data suggest that indexing extravascular lung water to height is superior to weight-based methods. As we are not aware of any abnormal hemodynamic profile for brain tumor patients, we propose our findings to be a close approximation to normal values.

Global end-diastolic volume acquired by transpulmonary thermodilution depends on age and gender in awake and spontaneously breathing patients.

INTRODUCTION: Volumetric parameters acquired by transpulmonary thermodilution had been repeatedly proven superior to filling pressures for estimation of cardiac preload. Up to now, the proposed normal ranges were never studied in detail. We investigated the relationship of the global end-diastolic volume (GEDV) acquired by transpulmonary thermodilution with age and gender in awake and spontaneously breathing patients. METHODS: Patients requiring brain tumor surgery were equipped prospectively with a transpulmonary thermodilution device. On postoperative day one, thermodilution measurements were performed in 101 patients ready for discharge from the ICU. All subjects were awake, spontaneously breathing, hemodynamically stable and free of catecholamines. RESULTS: Main finding was a dependence of GEDV on age and gender, height and weight of the patient. Age was a highly significant non-linear coefficient for GEDV with large inter-individual variance (p < 0.001). On average, GEDV was 131.1 ml higher in males (p = 0.027). Each cm body height accounted for 13.0 ml additional GEDV (p < 0.001). GEDV increased by 2.90 ml per kg actual body weight (p = 0.043). Each cofactor, including height and weight, remained significant after indexing GEDV to body surface area using predicted body weight. CONCLUSIONS: The volumetric parameter GEDV shows a large inter-individual variance and is dependent on age and gender. These dependencies persist after indexing GEDV to body surface area calculated with predicted body weight. Targeting resuscitation using fixed ranges of preload volumes acquired by transpulmonary thermodilution without concern to an individual patient's age and gender seems not to be appropriate.

Curcuminoids form reactive glucuronides in vitro.

Curcumin is of current interest because of its putative anti-inflammatory, anticarcinogenic, and anti-Alzheimer's activity, but its pharmacokinetic and metabolic fate is poorly understood. The present in vitro study has therefore been conducted on the glucuronidation of curcumin and its major phase I metabolite, hexahydro-curcumin, as well as of various natural and artificial analogs. The predominant glucuronide generated by rat and human liver microsomes from curcumin, hexahydro-curcumin, and other analogs with a phenolic hydroxyl group was a phenolic glucuronide according to LC-MS/MS analysis. However, a second glucuronide carrying the glucuronic acid moiety at the alcoholic hydroxyl group was formed from the same curcuminoids, but not hexahydro-curcuminoids, by human microsomes. Curcuminoids without a phenolic hydroxyl group gave rise to the aliphatic glucuronide only. The phenolic glucuronides of curcuminoids, but not of hexahydro-curcuminoids, were rather lipophilic and, in part, unstable in aqueous solution, their stability depending strongly on the type of aromatic substitution. The phenolic glucuronide of curcumin and of its natural congeners, but not the parent compounds, clearly inhibited the assembly of microtubule proteins under cell-free conditions, implying chemical reactivity of the glucuronides. These novel properties of the major phase II metabolites of curcuminoids deserve further investigation.