Interpretable Deep Learning System for Identifying Critical Patients Through the Prediction of Triage Level, Hospitalization, and Length of Stay: Prospective Study.

BACKGROUND: Triage is the process of accurately assessing patients' symptoms and providing them with proper clinical treatment in the emergency department (ED). While many countries have developed their triage process to stratify patients' clinical severity and thus distribute medical resources, there are still some limitations of the current triage process. Since the triage level is mainly identified by experienced nurses based on a mix of subjective and objective criteria, mis-triage often occurs in the ED. It can not only cause adverse effects on patients, but also impose an undue burden on the health care delivery system. OBJECTIVE: Our study aimed to design a prediction system based on triage information, including demographics, vital signs, and chief complaints. The proposed system can not only handle heterogeneous data, including tabular data and free-text data, but also provide interpretability for better acceptance by the ED staff in the hospital. METHODS: In this study, we proposed a system comprising 3 subsystems, with each of them handling a single task, including triage level prediction, hospitalization prediction, and length of stay prediction. We used a large amount of retrospective data to pretrain the model, and then, we fine-tuned the model on a prospective data set with a golden label. The proposed deep learning framework was built with TabNet and MacBERT (Chinese version of bidirectional encoder representations from transformers [BERT]). RESULTS: The performance of our proposed model was evaluated on data collected from the National Taiwan University Hospital (901 patients were included). The model achieved promising results on the collected data set, with accuracy values of 63%, 82%, and 71% for triage level prediction, hospitalization prediction, and length of stay prediction, respectively. CONCLUSIONS: Our system improved the prediction of 3 different medical outcomes when compared with other machine learning methods. With the pretrained vital sign encoder and repretrained mask language modeling MacBERT encoder, our multimodality model can provide a deeper insight into the characteristics of electronic health records. Additionally, by providing interpretability, we believe that the proposed system can assist nursing staff and physicians in taking appropriate medical decisions.

Anisotropic thermoelectric properties in hydrogenated nitrogen-doped porous graphene nanosheets.

By using density functional theory calculations combined with the nonequilibrium Green's function method and machine learning, we systematically studied the thermoelectric properties of four kinds of porous graphene nanosheets (PGNS) before and after nitrogen doping. The results show that the thermoelectric performance of porous graphene nanosheets along the armchair or zigzag chiral direction is improved due to the dramatically enhanced power factor caused by nitrogen doping. The calculated ZT values of nitrogen-doped porous graphene nanosheets are boosted by about one order of magnitude compared with those of undoped porous graphene nanosheets at room temperature. More importantly, an anisotropic thermoelectric transport is found in the nitrogen-doped porous graphene nanosheets. The results show that the ZT values of nitrogen-doped porous graphene nanosheets along the zigzag transport direction are nearly 11 times larger than those of them along the armchair transport direction. These results reveal that the thermoelectric properties of porous graphene nanosheets can be well regulated by nitrogen doping, and provide a good theoretical guidance for their application in thermoelectric devices.

Recent progress of two-dimensional heterostructures for thermoelectric applications.

The rapid development of synthesis and fabrication techniques has opened up a research upsurge in two-dimensional (2D) material heterostructures, which have received extensive attention due to their superior physical and chemical properties. Currently, thermoelectric energy conversion is an effective means to deal with the energy crisis and increasingly serious environmental pollution. Therefore, an in-depth understanding of thermoelectric transport properties in 2D heterostructures is crucial for the development of micro-nano energy devices. In this review, the recent progress of 2D heterostructures for thermoelectric applications is summarized in detail. Firstly, we systematically introduce diverse theoretical simulations and experimental measurements of the thermoelectric properties of 2D heterostructures. Then, the thermoelectric applications and performance regulation of several common 2D materials, as well as in-plane heterostructures and van der Waals heterostructures, are also discussed. Finally, the challenges of improving the thermoelectric performance of 2D heterostructures materials are summarized, and related prospects are described.

Excellent Medium-Temperature Thermoelectric Performance of Monolayer BiOCl.

Recently, a ternary-layered material BiOCl has elicited intense interest in photocatalysis, environmental remediation, and ultraviolet light detection because of its unique band gap of around 3.6 eV, low toxicity, and earth abundance. In particular, Gibson et al. reported a measurement of the in-plane thermal conductivity of BiOCl experimentally using a four-point-probe method [Science, 373, 1017-1022 (2021)], which is only 1.25 W/m K at 300 K. Motivated by the work, we studied the thermoelectric property of monolayer BiOCl using first-principles calculations combined with the Boltzmann transport equation. The calculated phonon thermal conductivity of monolayer BiOCl is 3 W/m K at 300 K, which is far below that of other promising 2D thermoelectric materials like graphyne and MoS2. A comprehensive analysis of phonon modes is conducted to reveal the low thermal conductivity. Moreover, the maximal ZT value is as high as 1.8 at 300 K and 5.7 at 800 K for the p-type doping with the 2 × 1015 cm-2 concentration. More importantly, we found that the thermoelectric efficiency of such 2D materials is significantly enhanced to 8 at 800 K by applying 1.5% tensile strain, which clearly outperforms that of the reported 2D thermoelectric material SnSe. The results shed light on the promising application in medium-temperature (600-900 K) thermoelectric devices.

Perfect spin-filtering effect in molecular junctions based on half-metallic penta-hexa-graphene nanoribbons.

The design and control of spintronic devices is a research hotspot in the field of electronics, and pure carbon-based materials provide new opportunities for the construction of electronic devices with excellent performance. Using density functional theory in combination with nonequilibrium Green's functions method, we design spin filter devices based on Penta-hexa-graphene (PHG) nanoribbons-a carbon nanomaterial in which the intrinsic magnetic moments combines with edge effects leading to a half-metallic property. Spin-resolved electronic transport studies show that such carbon-based devices can achieve nearly 100% spin filtering effect at low bias voltages. Such SEF can resist the influence of hydrogen passivation at different positions, but hardly survive under a hydrogen-rich environment. Our analysis show that the perfect SEF transport properties are caused by the magnetic and electronic properties of PHG nanoribbons, especially the magnetic moments on the quasi-sp3carbons. These interesting results indicate that PHG nanomaterials have very prominent application prospects in future spintronic devices.

Penta-Hexa-Graphene Nanoribbons: Intrinsic Magnetism and Edge Effect Induce Spin-Gapless Semiconducting and Half-Metallic Properties.

Two-dimensional materials with intrinsic long-range ordered magnetic moments have drawn a lot of attention. However, for practical applications, whether or not the magnetism is stable in their nanostructures has not been revealed. Here, based on the recently proposed magnetic penta-hexa-graphene, we study the electronic and magnetic properties of its nanoribbons (named PHGNRs). The results show that the PHGNRs have intrinsic robust magnetic moments that are different from zigzag graphene nanoribbons, where the magnetic moments caused by the edge effect are vulnerable. Moreover, the magnetic ground states, namely, ferromagnetic (FM) or antiferromagnetic (AFM), can be transformed by changing the width of PHGNRs. Most interestingly, under the FM ground state, the spin-polarized electronic properties reveal that the zigzag PHGNRs transform from spin-gapless semiconductors (SGSs) to half-metals, as the width of nanoribbons increases, while all the armchair PHGNRs are magnetic semiconductors. Furthermore, by considering different edge effects caused by the residual carbon atoms on the edges, the PHGNRs can further derive different types of SGSs, as well as half-metals. Our work suggests that the PHGNRs possessing intrinsic robust magnetic moments have potential applications in the field of spintronic devices.