Role Exchange-Based Self-Training Semi-Supervision Framework for Complex Medical Image Segmentation.

Segmentation of complex medical images such as vascular network and pulmonary tracheal network requires segmentation of many tiny targets on each tomographic section of the 3-D medical image volume. Although semantic segmentation of medical images based on deep learning has made great progress, fully supervised models require a great amount of annotations, making such complex medical image segmentation a difficult problem. In this article, we propose a semi-supervised model for complex medical image segmentation, which innovatively proposes a bidirectional self-training paradigm, through dynamically exchanging the roles of teacher and student by estimating the reliability at the model level. The direction of information and knowledge transfer between the two networks can be controlled, and the probability distribution of the roles of teacher and student in the next stage will be jointly determined by the model's uncertainty and instability in the training process. We also resolve the problem that loosely coupled networks are prone to collapse when training on small-scale annotated data by proposing asymmetric supervision (AS) strategy and hierarchical dual student (HDS) structure. In particular, a bidirectional distillation loss combined with the role exchange (RE) strategy and a global-local-aware consistency loss are introduced to obtain stable mutual promotion and achieve matching of global and local features, respectively. We conduct detailed experiments on two public datasets and one private dataset and lead existing semi-supervised methods by a large margin, while achieving fully supervised performance at a labeling cost of 5%.

High-Performance 2D Ambipolar MoTe2 Lateral Memristors by Mild Oxidation.

2D transition metal dichalcogenides (TMDCs) have been intensively explored in memristors for brain-inspired computing. Oxidation, which is usually unavoidable and harmful in 2D TMDCs, could also be used to enhance their memristive performances. However, it is still unclear how oxidation affects the resistive switching behaviors of 2D ambipolar TMDCs. In this work, a mild oxidation strategy is developed to greatly enhance the resistive switching ratio of ambipolar 2H-MoTe2 lateral memristors by more than 10 times. Such an enhancement results from the amplified doping due to O2 and H2O adsorption and the optimization of effective gate voltage distribution by mild oxidation. Moreover, the ambipolarity of 2H-MoTe2 also enables a change of resistive switching direction, which is uncommon in 2D memristors. Consequently, as an artificial synapse, the MoTe2 device exhibits a large dynamic range (≈200) and a good linearity (1.01) in long-term potentiation and depression, as well as a high-accuracy handwritten digit recognition (>96%). This work not only provides a feasible and effective way to enhance the memristive performance of 2D ambipolar materials, but also deepens the understanding of hidden mechanisms for RS behaviors in oxidized 2D materials.

Controlled Fabrication of Metallic MoO2 Nanosheets towards High-Performance p-Type 2D Transistors.

Two-dimensional (2D) semiconducting transition metal dichalcogenides (TMDCs) are extensively employed as channel materials in advanced electronic devices. The electrical contacts between electrodes and 2D semiconductors play a crucial role in the development of high-performance transistors. While numerous strategies for electrode interface engineering have been proposed to enhance the performance of n-type 2D transistors, upgrading p-type ones in a similar manner remains a challenge. In this work, significant improvements in a p-type WSe2 transistor are demonstrated by utilizing metallic MoO2 nanosheets as the electrode contact, which are controllably fabricated through physical vapor deposition and subsequent annealing. The MoO2 nanosheets exhibit an exceptional electrical conductivity of 8.4 × 104 S m‒1 and a breakdown current density of 3.3 × 106 A cm‒2. The work function of MoO2 nanosheets is determined to be ≈5.1 eV, making them suitable for contacting p-type 2D semiconductors. Employing MoO2 nanosheets as the electrode contact in WSe2 transistors results in a notable increase in the field-effect mobility to 92.0 cm2 V‒1 s‒1, which is one order of magnitude higher than the counterpart devices with conventional electrodes. This study not only introduces an intriguing 2D metal oxide to improve the electrical contact in p-type 2D transistors, but also offers an effective approach to fabricating all-2D devices.

Dual-Limit Growth of Large-Area Monolayer Transition Metal Dichalcogenides.

The large-scale growth of monolayer transition metal dichalcogenide (TMDC) films is a determinant for the implementation of two-dimensional materials in industrial applications. However, the simultaneous realization of uniform monolayer thickness and large-area coverage is still a challenge, because it requires precise control of reaction kinetics in both space and time dimensions. Herein, we achieve a variety of large-area monolayer TMDCs films by a dual-limit growth (DLG) that is realized through nanoporous carbon nanotube (CNT) films. In the DLG, a precursor-loaded CNT film placed face-to-face with a substrate provides a space-limited environment facilitating the monolayer growth, while the byproducts formed in the CNT film timely limits the supply of precursors released from nanopores of the CNT film, inhibiting the growth of multilayer TMDCs on the substrate. Consequently, large-area monolayer TMDC films are grown in a wide range of reaction times and show good homogeneity in thickness, optical properties, and device performance over the entire substrate. The DLG strategy is widely applicable for the growth of a variety of TMDC films including WSe2, MoS2, MoSe2, WS2, and ReS2. Our work provides a universal strategy to attain large-area monolayer TMDC films that can be used in practical applications of integrated circuits.

Multi-site, Multi-domain Airway Tree Modeling.

Open international challenges are becoming the de facto standard for assessing computer vision and image analysis algorithms. In recent years, new methods have extended the reach of pulmonary airway segmentation that is closer to the limit of image resolution. Since EXACT'09 pulmonary airway segmentation, limited effort has been directed to the quantitative comparison of newly emerged algorithms driven by the maturity of deep learning based approaches and extensive clinical efforts for resolving finer details of distal airways for early intervention of pulmonary diseases. Thus far, public annotated datasets are extremely limited, hindering the development of data-driven methods and detailed performance evaluation of new algorithms. To provide a benchmark for the medical imaging community, we organized the Multi-site, Multi-domain Airway Tree Modeling (ATM'22), which was held as an official challenge event during the MICCAI 2022 conference. ATM'22 provides large-scale CT scans with detailed pulmonary airway annotation, including 500 CT scans (300 for training, 50 for validation, and 150 for testing). The dataset was collected from different sites and it further included a portion of noisy COVID-19 CTs with ground-glass opacity and consolidation. Twenty-three teams participated in the entire phase of the challenge and the algorithms for the top ten teams are reviewed in this paper. Both quantitative and qualitative results revealed that deep learning models embedded with the topological continuity enhancement achieved superior performance in general. ATM'22 challenge holds as an open-call design, the training data and the gold standard evaluation are available upon successful registration via its homepage (https://atm22.grand-challenge.org/).

Perforated Carbon Nanotube Film Assisted Growth of Uniform Monolayer MoS2.

Scaling up the chemical vapor deposition (CVD) of monolayer transition metal dichalcogenides (TMDCs) is in high demand for practical applications. However, for CVD-grown TMDCs on a large scale, there are many existing factors that result in their poor uniformity. In particular, gas flow, which usually leads to inhomogeneous distributions of precursor concentrations, has yet to be well controlled. In this work, the growth of uniform monolayer MoS2 on a large scale by the delicate control of gas flows of precursors, which is realized by vertically aligning a well-designed perforated carbon nanotube (p-CNT) film face-to-face with the substrate in a horizontal tube furnace, is achieved. The p-CNT film releases gaseous Mo precursor from the solid part and allows S vapor to pass through the hollow part, resulting in uniform distributions of both gas flow rate and precursor concentrations near the substrate. Simulation results further verify that the well-designed p-CNT film guarantees a steady gas flow and a uniform spatial distribution of precursors. Consequently, the as-grown monolayer MoS2 shows quite good uniformity in geometry, density, structure, and electrical properties. This work provides a universal pathway for the synthesis of large-scale uniform monolayer TMDCs, and will advance their applications in high-performance electronic devices.

Liquid Precursor-Guided Phase Engineering of Single-Crystal VO2 Beams.

The study of VO2 flourishes due to its rich competing phases induced by slight stoichiometry variations. However, the vague mechanism of stoichiometry manipulation makes the precise phase engineering of VO2 still challenging. Here, stoichiometry manipulation of single-crystal VO2 beams in liquid-assisted growth is systematically studied. Contrary to previous experience, oxygen-rich VO2 phases are abnormally synthesized under a reduced oxygen concentration, revealing the important function of liquid V2 O5 precursor: It submerges VO2 crystals and stabilizes their stoichiometric phase (M1) by isolating them from the reactive atmosphere, while the uncovered crystals are oxidized by the growth atmosphere. By varying the thickness of liquid V2 O5 precursor and thus the exposure time of VO2 to the atmosphere, various VO2 phases (M1, T, and M2) can be selectively stabilized. Furthermore, this liquid precursor-guided growth can be used to spatially manages multiphase structures in single VO2 beams, enriching their deformation modes for actuation applications.

All-Transfer Electrode Interface Engineering Toward Harsh-Environment-Resistant MoS2 Field-Effect Transistors.

Nanoscale electronic devices that can work in harsh environments are in high demand for wearable, automotive, and aerospace electronics. Clean and defect-free interfaces are of vital importance for building nanoscale harsh-environment-resistant devices. However, current nanoscale devices are subject to failure in these environments, especially at defective electrode-channel interfaces. Here, harsh-environment-resistant MoS2 transistors are developed by engineering electrode-channel interfaces with an all-transfer of van der Waals electrodes. The delivered defect-free, graphene-buffered electrodes keep the electrode-channel interfaces intact and robust. As a result, the as-fabricated MoS2 devices have reduced Schottky barrier heights, leading to a very large on-state current and high carrier mobility. More importantly, the defect-free, hydrophobic graphene buffer layer prevents metal diffusion from the electrodes to MoS2 and the intercalation of water molecules at the electrode-MoS2 interfaces. This enables high resistances of MoS2 devices with all-transfer electrodes to various harsh environments, including humid, oxidizing, and high-temperature environments, surpassing the devices with other kinds of electrodes. The work deepens the understanding of the roles of electrode-channel interfaces in nanoscale devices and provides a promising interface engineering strategy to build nanoscale harsh-environment-resistant devices.

Humidity-Controlled Dynamic Engineering of Buckling Dimensionality in MoS2 Thin Films.

Dynamic engineering of buckling deformation is of vital importance as it provides multiphase modulation of thin film devices. In particular, dynamic switch of buckles between one-dimensional (1D) and two-dimensional (2D) configurations in a single film system on rigid substrates is intriguing but very challenging. The current approach to changing buckling configuration is mainly achieved by varying the built-in stress at the film-substrate interface, but it is difficult to realize dynamic engineering on rigid substrates. Herein, we report a dynamic engineering of buckling deformation in MoS2 thin films by humidity-tuned interfacial adhesion. With the change of humidity, the MoS2 thin films deform from 1D telephone-cord buckles to 2D web-like buckles due to the hydrophilic nature of both MoS2 and substrate. Such 1D-to-2D evolution of buckles is attributed to the weakened interfacial adhesion of mixed deformation modes induced by humidity, which is verified by finite-element modeling. These buckled films further find potential applications as patterned templates for liquid condensation and sensing units for tactile sensors. Our work not only demonstrates the humidity-controlled dimensionality engineering of buckles in MoS2 thin films but also sheds light on the functional applications of buckled films based on their profile features.

Determining the interlayer shearing in twisted bilayer MoS2 by nanoindentation.

The rise of twistronics has increased the attention of the community to the twist-angle-dependent properties of two-dimensional van der Waals integrated architectures. Clarification of the relationship between twist angles and interlayer mechanical interactions is important in benefiting the design of two-dimensional twisted structures. However, current mechanical methods have critical limitations in quantitatively probing the twist-angle dependence of two-dimensional interlayer interactions in monolayer limits. Here we report a nanoindentation-based technique and a shearing-boundary model to determine the interlayer mechanical interactions of twisted bilayer MoS2. Both in-plane elastic moduli and interlayer shear stress are found to be independent of the twist angle, which is attributed to the long-range interaction of intermolecular van der Waals forces that homogenously spread over the interfaces of MoS2. Our work provides a universal approach to determining the interlayer shear stress and deepens the understanding of twist-angle-dependent behaviours of two-dimensional layered materials.

Monolayer MoS2 Synaptic Transistors for High-Temperature Neuromorphic Applications.

As essential units in an artificial neural network (ANN), artificial synapses have to adapt to various environments. In particular, the development of synaptic transistors that can work above 125 °C is desirable. However, it is challenging due to the failure of materials or mechanisms at high temperatures. Here, we report a synaptic transistor working at hundreds of degrees Celsius. It employs monolayer MoS2 as the channel and Na+-diffused SiO2 as the ionic gate medium. A large on/off ratio of 106 can be achieved at 350 °C, 5 orders of magnitude higher than that of a normal MoS2 transistor in the same range of gate voltage. The short-term plasticity has a synaptic transistor function as an excellent low-pass dynamic filter. Long-term potentiation/depression and spike-timing-dependent plasticity are demonstrated at 150 °C. An ANN can be simulated, with the recognition accuracy reaching 90%. Our work provides promising strategies for high-temperature neuromorphic applications.

Grain-Boundary Engineering of Monolayer MoS2 for Energy-Efficient Lateral Synaptic Devices.

Synaptic devices based on 2D-layered materials have emerged as high-efficiency electronic synapses and neurons for neuromorphic computing. Lateral 2D synaptic devices have the advantages of multiple functionalities by responding to diverse stimuli, but they consume large amounts of energy, far more than the human brain. Moreover, current lateral devices employ several mechanisms based on conductive filaments and grain boundaries (GBs), but their formation is random and difficult to control, also hindering their practical applications. Here, four-terminal, lateral synaptic devices with artificially engineered GBs are reported, which are made from monolayer MoS2 . With lithography-free, direct-laser-writing-controlled MoS2 /MoS2- x Oδ GBs, such synaptic devices exhibit short-term and long-term plasticity characteristics that are responsive to electric and light stimulation simultaneously. This enables detailed simulations of biological learning and cognitive processes as well as image perception and processing. In particular, the device exhibits low energy consumption, similar to that of the human brain and much lower than those of other lateral 2D synaptic devices. This work provides an effective way to fabricate lateral synaptic devices for practical application development and sheds light on controllable electrical state switching for neuromorphic computing.

Two-Dimensional Lateral Heterostructures Made by Selective Reaction on a Patterned Monolayer MoS2 Matrix.

Two-dimensional (2D) heterostructures have attracted widespread attention for their promising prospects in the fields of electronics and optoelectronics. However, in order to truly realize 2D-material-based integrated circuits, precisely controllable fabrication of 2D heterostructures is crucial and urgently needed. Here, we demonstrate an ex situ growth method of MoSe2/MoS2 lateral heterostructures by selective selenization of a laser-scanned, ultrathin oxidized region (MoOx) on a monolayer MoS2 matrix. In our method, monolayer MoS2 is scanned by a laser with a pre-designed pattern, where the laser-scanned MoS2 is totally oxidized into MoOx. The oxidized region is then selenized in a furnace, while the unoxidized MoS2 region remains unchanged, delivering a MoSe2/MoS2 heterostructure. Unlike in situ laser direct growth methods, our method separates the laser-scanned process from the selenization process, which avoids the long time of point-by-point selenization of MoS2 by laser, making the efficiency of the synthesis greatly improved. The formation process of the heterostructure is studied by Raman spectroscopy and Auger electron spectroscopy. This simple and controllable approach to lateral heterostructures with desired patterns paves the way for fast and mass integration of devices based on 2D heterostructures.

Ultrafast, Kinetically Limited, Ambient Synthesis of Vanadium Dioxides through Laser Direct Writing on Ultrathin Chalcogenide Matrix.

Vanadium dioxide (VO2) is a strongly correlated electronic material and has attracted significant attention due to its metal-to-insulator transition and diverse smart applications. Traditional synthesis of VO2 usually requires minutes or hours of global heating and low oxygen partial pressure to achieve thermodynamic control of the valence state. Further patterning of VO2 through a series of lithography and etching processes may inevitably change its surface valence, which poses a great challenge for the assembly of micro- and nanoscale VO2-based heterojunction devices. Herein, we report an ultrafast method to simultaneously synthesize and pattern VO2 on the time scale of seconds under ambient conditions through laser direct writing on a V5S8 "canvas". The successful ambient synthesis of VO2 is attributed to the ultrafast local heating and cooling process, resulting in controlled freezing of the intermediate oxidation phase during the relatively long kinetic reaction. A Mott memristor based on a V5S8-VO2-V5S8 lateral heterostructure can be fabricated and integrated with a MoS2 channel, delivering a transistor with abrupt switching transfer characteristics. The other device with a VSxOy channel exhibits a large negative temperature coefficient of approximately 4.5%/K, which is highly desirable for microbolometers. The proposed approach enables fast and efficient integration of VO2-based heterojunction devices and is applicable to other intriguing intermediate phases of oxides.