Automatic Classification of Thyroid Findings Using Static and Contextualized Ensemble Natural Language Processing Systems: Development Study.

BACKGROUND: In the case of Korean institutions and enterprises that collect nonstandardized and nonunified formats of electronic medical examination results from multiple medical institutions, a group of experienced nurses who can understand the results and related contexts initially classified the reports manually. The classification guidelines were established by years of workers' clinical experiences and there were attempts to automate the classification work. However, there have been problems in which rule-based algorithms or human labor-intensive efforts can be time-consuming or limited owing to high potential errors. We investigated natural language processing (NLP) architectures and proposed ensemble models to create automated classifiers. OBJECTIVE: This study aimed to develop practical deep learning models with electronic medical records from 284 health care institutions and open-source corpus data sets for automatically classifying 3 thyroid conditions: healthy, caution required, and critical. The primary goal is to increase the overall accuracy of the classification, yet there are practical and industrial needs to correctly predict healthy (negative) thyroid condition data, which are mostly medical examination results, and minimize false-negative rates under the prediction of healthy thyroid conditions. METHODS: The data sets included thyroid and comprehensive medical examination reports. The textual data are not only documented in fully complete sentences but also written in lists of words or phrases. Therefore, we propose static and contextualized ensemble NLP network (SCENT) systems to successfully reflect static and contextual information and handle incomplete sentences. We prepared each convolution neural network (CNN)-, long short-term memory (LSTM)-, and efficiently learning an encoder that classifies token replacements accurately (ELECTRA)-based ensemble model by training or fine-tuning them multiple times. Through comprehensive experiments, we propose 2 versions of ensemble models, SCENT-v1 and SCENT-v2, with the single-architecture-based CNN, LSTM, and ELECTRA ensemble models for the best classification performance and practical use, respectively. SCENT-v1 is an ensemble of CNN and ELECTRA ensemble models, and SCENT-v2 is a hierarchical ensemble of CNN, LSTM, and ELECTRA ensemble models. SCENT-v2 first classifies the 3 labels using an ELECTRA ensemble model and then reclassifies them using an ensemble model of CNN and LSTM if the ELECTRA ensemble model predicted them as "healthy" labels. RESULTS: SCENT-v1 outperformed all the suggested models, with the highest F1 score (92.56%). SCENT-v2 had the second-highest recall value (94.44%) and the fewest misclassifications for caution-required thyroid condition while maintaining 0 classification error for the critical thyroid condition under the prediction of the healthy thyroid condition. CONCLUSIONS: The proposed SCENT demonstrates good classification performance despite the unique characteristics of the Korean language and problems of data lack and imbalance, especially for the extremely low amount of critical condition data. The result of SCENT-v1 indicates that different perspectives of static and contextual input token representations can enhance classification performance. SCENT-v2 has a strong impact on the prediction of healthy thyroid conditions.

Reversible Ligand Exchange in Atomically Dispersed Catalysts for Modulating the Activity and Selectivity of the Oxygen Reduction Reaction.

Rational control of the coordination environment of atomically dispersed catalysts is pivotal to achieve desirable catalytic reactivity. We report the reversible control of coordination structure in atomically dispersed electrocatalysts via ligand exchange reactions to reversibly modulate their reactivity for oxygen reduction reaction (ORR). The CO-ligated atomically dispersed Rh catalyst exhibited ca. 30-fold higher ORR activity than the NHx -ligated catalyst, whereas the latter showed three times higher H2 O2 selectivity than the former. Post-treatments of the catalysts with CO or NH3 allowed the reversible exchange of CO and NHx ligands, which reversibly tuned oxidation state of metal centers and their ORR activity and selectivity. DFT calculations revealed that more reduced oxidation state of CO-ligated Rh site could further stabilize the *OOH intermediate, facilitating the two- and four-electron pathway ORR. The reversible ligand exchange reactions were generalized to Ir- and Pt-based catalysts.

Lattice Engineering to Simultaneously Control the Defect/Stacking Structures of Layered Double Hydroxide Nanosheets to Optimize Their Energy Functionalities.

An effective lattice engineering method to simultaneously control the defect structure and the porosity of layered double hydroxides (LDHs) was developed by adjusting the elastic deformation and chemical interactions of the nanosheets during the restacking process. The enlargement of the intercalant size and the lowering of the charge density were effective in increasing the content of oxygen vacancies and enhancing the porosity of the stacked nanosheets via layer thinning. The defect-rich Co-Al-LDH-NO3- nanohybrid with a small stacking number exhibited excellent performance as an oxygen evolution electrocatalyst and supercapacitor electrode with a large specific capacitance of ∼2230 F g-1 at 1 A g-1, which is the largest capacitance of carbon-free LDH-based electrodes reported to date. Combined with the results of density functional theory calculations, the observed excellent correlations between the overpotential/capacitance and the defect content/stacking number highlight the importance of defect/stacking structures in optimizing the energy functionalities. This was attributed to enhanced orbital interactions with water/hydroxide at an increased number of defect sites. The present cost-effective lattice engineering process can therefore provide an economically feasible methodology to explore high-performance electrocatalyst/electrode materials.

Identification of Single-Atom Ni Site Active toward Electrochemical CO2 Conversion to CO.

Electrocatalytic conversion of CO2 into value-added products offers a new paradigm for a sustainable carbon economy. For active CO2 electrolysis, the single-atom Ni catalyst has been proposed as promising from experiments, but an idealized Ni-N4 site shows an unfavorable energetics from theory, leading to many debates on the chemical nature responsible for high activity. To resolve this conundrum, here we investigated CO2 electrolysis of Ni sites with well-defined coordination, tetraphenylporphyrin (N4-TPP) and 21-oxatetraphenylporphyrin (N3O-TPP). Advanced spectroscopic and computational studies revealed that the broken ligand-field symmetry is the key for active CO2 electrolysis, which subordinates an increase in the Ni redox potential yielding NiI. Along with their importance in activity, ligand-field symmetry and strength are directly related to the stability of the Ni center. This suggests the next quest for an activity-stability map in the domain of ligand-field strength, toward a rational ligand-field engineering of single-atom Ni catalysts for efficient CO2 electrolysis.

A General Strategy to Atomically Dispersed Precious Metal Catalysts for Unravelling Their Catalytic Trends for Oxygen Reduction Reaction.

Atomically dispersed precious metal catalysts have emerged as a frontier in catalysis. However, a robust, generic synthetic strategy toward atomically dispersed catalysts is still lacking, which has limited systematic studies revealing their general catalytic trends distinct from those of conventional nanoparticle (NP)-based catalysts. Herein, we report a general synthetic strategy toward atomically dispersed precious metal catalysts, which consists of "trapping" precious metal precursors on a heteroatom-doped carbonaceous layer coated on a carbon support and "immobilizing" them with a SiO2 layer during thermal activation. Through the "trapping-and-immobilizing" method, five atomically dispersed precious metal catalysts (Os, Ru, Rh, Ir, and Pt) could be obtained and served as model catalysts for unravelling catalytic trends for the oxygen reduction reaction (ORR). Owing to their isolated geometry, the atomically dispersed precious metal catalysts generally showed higher selectivity for H2O2 production than their NP counterparts for the ORR. Among the atomically dispersed catalysts, the H2O2 selectivity was changed by the types of metals, with atomically dispersed Pt catalyst showing the highest selectivity. A combination of experimental results and density functional theory calculations revealed that the selectivity trend of atomically dispersed catalysts could be correlated to the binding energy difference between *OOH and *O species. In terms of 2 e- ORR activity, the atomically dispersed Rh catalyst showed the best activity. Our general approach to atomically dispersed precious metal catalysts may help in understanding their unique catalytic behaviors for the ORR.