Development and Validation of a Robust and Interpretable Early Triaging Support System for Patients Hospitalized With COVID-19: Predictive Algorithm Modeling and Interpretation Study.

BACKGROUND: Robust and accurate prediction of severity for patients with COVID-19 is crucial for patient triaging decisions. Many proposed models were prone to either high bias risk or low-to-moderate discrimination. Some also suffered from a lack of clinical interpretability and were developed based on early pandemic period data. Hence, there has been a compelling need for advancements in prediction models for better clinical applicability. OBJECTIVE: The primary objective of this study was to develop and validate a machine learning-based Robust and Interpretable Early Triaging Support (RIETS) system that predicts severity progression (involving any of the following events: intensive care unit admission, in-hospital death, mechanical ventilation required, or extracorporeal membrane oxygenation required) within 15 days upon hospitalization based on routinely available clinical and laboratory biomarkers. METHODS: We included data from 5945 hospitalized patients with COVID-19 from 19 hospitals in South Korea collected between January 2020 and August 2022. For model development and external validation, the whole data set was partitioned into 2 independent cohorts by stratified random cluster sampling according to hospital type (general and tertiary care) and geographical location (metropolitan and nonmetropolitan). Machine learning models were trained and internally validated through a cross-validation technique on the development cohort. They were externally validated using a bootstrapped sampling technique on the external validation cohort. The best-performing model was selected primarily based on the area under the receiver operating characteristic curve (AUROC), and its robustness was evaluated using bias risk assessment. For model interpretability, we used Shapley and patient clustering methods. RESULTS: Our final model, RIETS, was developed based on a deep neural network of 11 clinical and laboratory biomarkers that are readily available within the first day of hospitalization. The features predictive of severity included lactate dehydrogenase, age, absolute lymphocyte count, dyspnea, respiratory rate, diabetes mellitus, c-reactive protein, absolute neutrophil count, platelet count, white blood cell count, and saturation of peripheral oxygen. RIETS demonstrated excellent discrimination (AUROC=0.937; 95% CI 0.935-0.938) with high calibration (integrated calibration index=0.041), satisfied all the criteria of low bias risk in a risk assessment tool, and provided detailed interpretations of model parameters and patient clusters. In addition, RIETS showed potential for transportability across variant periods with its sustainable prediction on Omicron cases (AUROC=0.903, 95% CI 0.897-0.910). CONCLUSIONS: RIETS was developed and validated to assist early triaging by promptly predicting the severity of hospitalized patients with COVID-19. Its high performance with low bias risk ensures considerably reliable prediction. The use of a nationwide multicenter cohort in the model development and validation implicates generalizability. The use of routinely collected features may enable wide adaptability. Interpretations of model parameters and patients can promote clinical applicability. Together, we anticipate that RIETS will facilitate the patient triaging workflow and efficient resource allocation when incorporated into a routine clinical practice.

Combination of an inject-and-transfer system for serial femtosecond crystallography.

Serial femtosecond crystallography (SFX) enables the determination of room-temperature crystal structures of macromolecules with minimized radiation damage and provides time-resolved molecular dynamics by pump-probe or mix-and-inject experiments. In SFX, a variety of sample delivery methods with unique advantages have been developed and applied. The combination of existing sample delivery methods can enable a new approach to SFX data collection that combines the advantages of the individual methods. This study introduces a combined inject-and-transfer system (BITS) method for sample delivery in SFX experiments: a hybrid injection and fixed-target scanning method. BITS allows for solution samples to be reliably deposited on ultraviolet ozone (UVO)-treated polyimide films, at a minimum flow rate of 0.5 nl min-1, in both vertical and horizontal scanning modes. To utilize BITS in SFX experiments, lysozyme crystal samples were embedded in a viscous lard medium and injected at flow rates of 50-100 nl min-1 through a syringe needle onto a UVO-treated polyimide film, which was mounted on a fixed-target scan stage. The crystal samples deposited on the film were raster scanned with an X-ray free electron laser using a motion stage in both horizontal and vertical directions. Using the BITS method, the room-temperature structure of lysozyme was successfully determined at a resolution of 2.1 Å, and thus BITS could be utilized in future SFX experiments.

Photonic Encryption Platform via Dual-Band Vectorial Metaholograms in the Ultraviolet and Visible.

Metasurface-driven optical encryption devices have attracted much attention. Here, we propose a dual-band vectorial metahologram in the visible and ultraviolet (UV) regimes for optical encryption. Nine polarization-encoded vectorial holograms are observed under UV laser illumination, while another independent hologram appears under visible laser illumination. The proposed engineered silicon nitride, which is transparent in UV, is employed to demonstrate the UV hologram. Nine holographic images for different polarization states are encoded using a pixelated metasurface. The dual-band metahologram is experimentally implemented by stacking the individual metasurfaces that operate in the UV and visible. The visible hologram can be decrypted to provide the first key, a polarization state, which is used to decode the password hidden in the UV vectorial hologram through the use of an analyzer. Considering the property of UV to be invisible to the naked eye, the multiple polarization channels of the vectorial hologram, and the dual-band decoupling, the demonstrated dual-band vectorial hologram device could be applied in various high-security and anticounterfeiting applications.

Completely Hazy and Transparent Films by Embedding Air Gaps for Elimination of Angular Color Shift in Organic Light-Emitting Diodes.

Red, green, and blue top-emission organic light-emitting diodes (RGB TOLEDs) suffer from white color change with viewing angle due to the microcavity effect, called white angular dependence (WAD). Great efforts are devoted by applying various kinds of hazy films, but they suffer from poor mechanical stability and optical transmittance. Herein, we introduce an air-gap-embedded hazy film (AEHF) to solve these problems and suppress WAD in RGB TOLEDs. The AEHF is designed with optical simulation to realize high haze with transparency. By tuning geometries of the air gap inside the polymer, the AEHF realizes high haze of more than 90% in all RGB colors while maintaining high transparency. To experimentally demonstrate the AEHF, the O2 plasma is treated on a polymer film with AgCl as an etching mask to fabricate microstructures with high aspect ratios. Afterward, PDMS is coated on the patterned surface; air gaps develop spontaneously in the valleys between microstructures during the coating process. Using these processes, an air gap with 1.2 µm size and 400 nm period is formed inside the film and ∼100% haze is achieved while maintaining a high transmittance of 88%; these results agree well with rigorous coupled wave analysis results. By utilizing the AEHF into TOLEDs, the WAD can be drastically suppressed by 95.2% compared with that of a device without AEHF.

Photonic Multilayer Structure Induced High Near-Infrared (NIR) Blockage as Energy-Saving Window.

Energy-saving window that selectively blocks near-infrared (NIR) is a promising technology to save energy consumption. However, it is hard to achieve both high transmittance in visible light and high reflectance in NIR for the energy-saving windows. Here, a TiO2 /Ag/TiO2 /SiO2 /TiO2 multilayer is demonstrated on a glass substrate to selectively block NIR while maintaining high transmittance to visible light. The thickness of a TiO2 /Ag/TiO2 structure is first design and optimized; the metal layer reflects NIR and the dielectric layers increase transmittance of visible light with zero reflection condition. To further enhance NIR-blocking capability, a TiO2 back reflector is implemented with a SiO2 spacer to TiO2 /Ag/TiO2 structure. The back reflector can induce additional Fresnel reflection without sacrificing transmittance to visible light. The optimal TiO2 (32 nm)/Ag (22 nm)/TiO2 (30 nm)/SiO2 (100 nm)/TiO2 (110 nm)/glass shows solar energy rejection 89.2% (reflection 86.5%, absorption 2.7%) in NIR, visible transmittance 69.9% and high long-wave (3 ≤ λ ≤ 20 µm) reflectance > 95%. This proposed visible-transparent, near-infrared-reflecting multilayer film can be applied to the windows of buildings and automobiles to reduce the energy consumption.

Grain Boundary Engineering of Cu-Ag Thin-Film Catalysts for Selective (Photo)Electrochemical CO2 Reduction to CO and CH4.

We investigated the relationship between grain boundary (GB) oxidation of Cu-Ag thin-film catalysts and selectivity of the (photo)electrochemical CO2 reduction reaction (CO2 RR). The change in the thickness of the Cu thin film accompanies the variation of GB density, and the Ag layer (3 nm) has an island-like morphology on the Cu thin film. Therefore, oxygen from ambient air penetrates into the Cu thin film through the GB of Cu and binds with it because the uncoordinated Cu atoms at the GBs are unstable. It was found that the Cu thin film with a small grain size was susceptible to spontaneous oxidation and degraded the faradaic efficiency (FE) of CO and CH4. However, a relatively thick (≥80 nm) Cu layer was effective in preventing the GB oxidation and realized catalytic properties similar to those of bulk Cu-Ag catalysts. The optimized Cu (100 nm)-Ag (3 nm) thin film exhibited a unique bifunctional characteristic, which enables selective production of both CO (FECO = 79.8%) and CH4 (FECH4 = 59.3%) at a reductive potential of -1.0 and -1.4 VRHE, respectively. Moreover, the Cu-Ag thin film was used as a cocatalyst for photo-electrochemical CO2 reduction by patterning the Cu-Ag thin film and a SiO2 passivation layer on a p-type Si photocathode. This novel architecture improved the selectivity of CO and CH4 under light illumination (100 mW/cm2).

Highly transparent all-dielectric metasurface to block the near-infrared region of the solar spectrum.

Blocking the near-infrared region (NIR) is indispensable for saving energy consumed to maintain the interior temperature in buildings. However, simultaneously enhancing transmission in visible light and blocking in the NIR remains challenging. Here, we theoretically demonstrate a transparent all-dielectric metasurface selectively blocking the NIR by using TiO2 nanocylinders and an indium tin oxide (ITO) layer. The ITO layer is implemented as a back reflector because ITO is transparent in visible light, whereas the ITO becomes a reflective material in the long-wavelength region (λ > 1500 nm). The designed metasurface exhibits high average transmittance of 70% in visible light and high solar energy rejection (SER) of 90% in the NIR. Furthermore, the blocking capability in the NIR of the designed metasurface is maintained over a wide range of an incident angle and polarization angle of light. Therefore, the metasurface gives a guideline for designing energy-saving applications.

Application of a high-throughput microcrystal delivery system to serial femtosecond crystallography.

Microcrystal delivery methods are pivotal in the use of serial femtosecond crystallography (SFX) to resolve the macromolecular structures of proteins. Here, the development of a novel technique and instruments for efficiently delivering microcrystals for SFX are presented. The new method, which relies on a one-dimensional fixed-target system that includes a microcrystal container, consumes an extremely low amount of sample compared with conventional two-dimensional fixed-target techniques at ambient temperature. This novel system can deliver soluble microcrystals without highly viscous carrier media and, moreover, can be used as a microcrystal growth device for SFX. Diffraction data collection utilizing this advanced technique along with a real-time visual servo scan system has been successfully demonstrated for the structure determination of proteinase K microcrystals at 1.85 Šresolution.

Air-gap-embedded robust hazy films to reduce the screen-door effect in virtual reality displays.

We report a way to make an air-gap-embedded flexible film to reduce the screen-door effect (SDE) in virtual reality (VR) displays. Oxygen plasma was treated with a polyethylene terephthalate substrate to produce wavelength-scale micropatterns. These micropatterns induce an effective haze, but it is easily destroyed by a very small external scratch. Such a problem could be solved by coating the patterns with poly(dimethylsiloxane) (PDMS). The viscosity of PDMS, controlled by the ratio of the base and curing agents, plays a key role in determining the size of air-gaps at the valleys of micropatterns. As the ratio of base agent increases to 40, the average haze abruptly increased from 0.9% to 88.6% in visible wavelengths, while the average total transmittance maintained was between 89.8 and 91.7%. The origin of air-gap-induced haze is confirmed by numerical simulations. The hazy film remarkably reduced the SDE of the VR display from 30.27% to 4.83% for red color, from 21.82% to 2.58% for green, and from 26.02% to 3.38% for blue, as the size of air-gaps increases from 0 to 406 ± 91 nm. No defects were found after 10 000 bending cycles with a bending radius of 3 mm.

Nylon mesh-based sample holder for fixed-target serial femtosecond crystallography.

Fixed-target serial femtosecond crystallography (FT-SFX) was an important advance in crystallography by dramatically reducing sample consumption, while maintaining the benefits of SFX for obtaining crystal structures at room temperature without radiation damage. Despite a number of advantages, preparation of a sample holder for the sample delivery in FT-SFX with the use of many crystals in a single mount at ambient temperature is challenging as it can be complicated and costly, and thus, development of an efficient sample holder is essential. In this study, we introduced a nylon mesh-based sample holder enclosed by a polyimide film. This sample holder can be rapidly manufactured using a commercially available nylon mesh with pores of a desired size at a low cost without challenging technology. Furthermore, this simple device is highly efficient in data acquisition. We performed FT-SFX using a nylon mesh-based sample holder and collected over 130,000 images on a single sample holder using a 30 Hz X-ray pulse for 1.2 h. We determined the crystal structures of lysozyme and glucose isomerase using the nylon mesh at 1.65 and 1.75 Å, respectively. The nylon mesh exposed to X-rays produced very low levels of background scattering at 3.75 and 4.30 Å, which are negligible for data analysis. Our method provides a simple and rapid but highly efficient way to deliver samples for FT-SFX.

Nano-imprinting of refractive-index-matched indium tin oxide sol-gel in light-emitting diodes for eliminating total internal reflection.

Refractive-index (RI)-matched nanostructures are implemented in GaN-based light-emitting diodes (LEDs) for enhancing light output efficiency. The RI-matched indium tin oxide (ITO) nanostructures are successfully implemented in GaN-based lateral LEDs by using ITO sol-gel and nanoimprint lithography. The ITO sol-gel nanostructures annealed at 300 °C have RI of 1.95, showing high transparency of 90% and high diffused transmittance of 34%. Consequently, the light output power in LEDs with the RI-matched nanostructures increases by 8% in comparison with that in LEDs containing flat ITO. Ray tracing and finite-difference time-domain (FDTD) simulations show that the RI-matched nanostructures on the transparent current spreading layer dramatically reduce Fresnel reflection loss at the interface of the current spreading layer with the nanostructure and extract confined waveguide lights in LEDs.