Polygonal Approximation Learning for Convex Object Segmentation in Biomedical Images with Bounding Box Supervision.

As a common and critical medical image analysis task, deep learning based biomedical image segmentation is hindered by the dependence on costly fine-grained annotations. To alleviate this data dependence, in this paper, a novel approach, called Polygonal Approximation Learning (PAL), is proposed for convex object instance segmentation with only bounding-box supervision. The key idea behind PAL is that the detection model for convex objects already contains the necessary information for segmenting them since their convex hulls, which can be generated approximately by the intersection of bounding boxes, are equivalent to the masks representing the objects. To extract the essential information from the detection model, a repeated detection approach is employed on biomedical images where various rotation angles are applied and a dice loss with the projection of the rotated detection results is utilized as a supervised signal in training our segmentation model. In biomedical imaging tasks involving convex objects, such as nuclei instance segmentation, PAL outperforms the known models (e.g., BoxInst) that rely solely on box supervision. Furthermore, PAL achieves comparable performance with mask-supervised models including Mask R-CNN and Cascade Mask R-CNN. Interestingly, PAL also demonstrates remarkable performance on non-convex object instance segmentation tasks, for example, surgical instrument and organ instance segmentation. Our code is available at https://github.com/shenmishajing/PAL.

First-principles simulation of monolayer hydrogen passivated Bi2O2S2-metal interfaces.

Monolayer (ML) MoS2 is one of the most extensively studied two-dimensional (2D) semiconductors. However, it suffers from low carrier mobility and pervasive Schottky contact with metal electrodes. 2D semiconductor Bi2O2S, a sulfur analogue of 2D Bi2O2Se, has been prepared recently. ML fully hydrogen-passivated Bi2O2S2 (Bi2O2S2H2) posseses a comparable band gap (1.92 eV) with ML MoS2 (1.8 eV), but probably has a better device performance than ML MoS2. Based on the density functional theory, the electron and hole mobilities of ML Bi2O2S2H2 at 300 K are calculated to be 16 447-26 699 and 264-968 cm2 V-1 s-1, respectively. Then we firstly characterize the contact properties of ML half hydrogen-passivated Bi2O2S2 (Bi2O2S2H) with four bulk metal electrodes (Ti, Sc, Pd, and Pt) based on ab initio quantum transport simulation. In the lateral direction, a p-type Schottky contact is found in Pd and Pt electrodes, and the corresponding hole Schottky barrier heights (SBHs) are 0.54 and 0.99 eV, respectively. Remarkably, a coveted n-type Ohmic contact appears in Sc and Ti electrodes. Finally, the current on-off ratio of the ML hydrogen-passivated Bi2O2S2 field effect transistor with a Ti electrode reaches 105. Hence, the good intrinsic properties, contact properties, and large switching ability put ML hydrogen-passivated Bi2O2S2 in the rank of potential channel candidates for post-silicon era field effect transistors.

n-Type Ohmic contact and p-type Schottky contact of monolayer InSe transistors.

Owing to their few lateral dangling bonds and enhanced gate electrostatics, two-dimensional semiconductors have attracted much attention for the fabrication of channels in next-generation field-effect transistors (FETs). Herein, combining first-principle band structure calculations with more precise quantum transport simulations, we systematically explore the interface properties between monolayer (ML) indium selenide (InSe) and a sequence of common electrodes in an FET. The ML InSe band structure is damaged by Sc, Au, Cr, Pt, and Pd electrodes but identifiable in contact with Ag, Cu, In, graphene and ML O-terminated Cr2C electrodes. A lateral n-type Schottky contact is generated with Sc, Au, Cr, Pt, Pd, and ML graphene electrodes owing to Fermi level pinning originating from the metal-induced gap states, which feature a pinning factor of 0.32. Luckily, a highly desirable lateral n-type Ohmic contact is generated with the Ag, Cu, and In electrodes. The calculated contact polarity is in agreement with the available experimental results using Au, Cr, ML graphene, Ag, and In as electrodes. Remarkably, a lateral p-type Schottky contact is generated with ML O-terminated Cr2C despite the very high work function of ML InSe. Therefore, this study offers a deeper understanding of ML InSe device interfaces and instructions for the design of ML InSe transistors.

n- and p-type ohmic contacts at monolayer gallium nitride-metal interfaces.

Recently, two-dimensional (2D) gallium nitride (GaN) was experimentally fabricated, and has promising applications in next-generation electronic and optoelectronic devices. A direct contact with metals to inject the carrier is often required for potential 2D GaN devices. Herein, the first systematic study on the interface properties of monolayer (ML) planar and buckled GaN with different metal electrodes (Al, Ti, Ag, Au, Sc, and Pt) in a field-effect transistor framework is presented using first-principles energy band calculations and quantum transport simulations. Because of moderate Fermi level pinning (electron pinning factor S = 0.63), ML planar GaN and the Ag electrode form an n-type lateral Schottky contact, while ML planar GaN and Ti, Al, and Au electrodes form a p-type lateral Schottky contact. The ML buckled GaN, Ag, Al, Ti, and Sc electrodes form a p-type lateral Schottky contact as a result of Fermi level pinning with a hole pinning factor of S = 0.75. Notably, a highly desirable n-type/p-type lateral ohmic contact is formed at the lateral interface of the ML planar GaN and Sc/Pt electrodes, and a p-type lateral ohmic contact is formed at the lateral interface of the ML buckled GaN and Pt/Au electrodes. Therefore, a low resistance contact can be realized in ML planar and buckled GaN devices.