Orthopantomogram (OPGs) and Cone-Beam Computed Tomography (CBCT) are vital for dentistry, but creating large datasets for automated tooth segmentation is hindered by the labor-intensive process of manual instance-level annotation. This research aimed to benchmark and advance semi-supervised learning (SSL) as a solution for this data scarcity problem. We organized the 2nd Semi-supervised Teeth Segmentation (STS 2024) Challenge at MICCAI 2024. We provided a large-scale dataset comprising over 90,000 2D images and 3D axial slices, which includes 2380 OPG images and 330 CBCT scans, all featuring detailed instance-level FDI annotations on part of the data. The challenge attracted 114 (OPG) and 106 (CBCT) registered teams. To ensure algorithmic excellence and full transparency, we rigorously evaluated the valid, open-source submissions from the top 10 (OPG) and top 5 (CBCT) teams, respectively. All successful submissions were deep learning-based SSL methods. The winning semi-supervised models demonstrated impressive performance gains over a fully-supervised nnU-Net baseline trained only on the labeled data. For the 2D OPG track, the top method improved the Instance Affinity (IA) score by over 44 percentage points. For the 3D CBCT track, the winning approach boosted the Instance Dice score by 61 percentage points. This challenge demonstrates the potential benefit benefit of SSL for complex, instance-level medical image segmentation tasks where labeled data is scarce. The most effective approaches consistently leveraged hybrid semi-supervised frameworks that combined knowledge from foundational models like SAM with multi-stage, coarse-to-fine refinement pipelines. Both the challenge dataset and the participants' submitted code have been made publicly available on GitHub (https://github.com/ricoleehduu/STS-Challenge-2024), ensuring transparency and reproducibility.

Computer-aided diagnosis greatly enhances personalized treatment planning and diagnostic efficiency by providing accurate dental anatomy through teeth segmentation. However, it still constrained by the scarcity of high-quality annotated dental datasets. To address this issue, this paper presents a dataset combining both 2D panoramic X-rays with over 6,500 images and 3D CBCT with over 580 volumes (88,500+ slices) to support the Semi-supervised Teeth Segmentation (STS) Challenge, which includes partially meticulous annotations and covers all age groups. Moreover, multi-phase semi-supervised teeth segmentation algorithms and high-confidence pseudo-labels refinement strategies were proposed by competitors during this challenge. Algorithms were verified on this proposed dataset and good segmentation performance were achieved, over 93+ and 80+ Dice score were obtained for top three 2D and 3D participants, demonstrating the high quality of this proposed dataset. This paper also summarizes the diverse methods employed by the top-ranking teams in the MICCAI 2023 STS Challenge. Our dataset is publicly accessible through Zenodo ( https://zenodo.org/records/10597292 ), and the participants’ code is hosted on GitHub ( https://github.com/ricoleehduu/STS-Challenge ).

With the increasing demand for higher bandwidth and frequency in high-speed digital systems, the interference of common-mode (CM) noise in differential signal transmission has become more severe. Common-mode suppression filters (CMF) were proposed to solve this issue, but their design process typically relies on empirical parameter tuning with extensive electromagnetic simulations, which not only increases design costs but also limits efficiency and flexibility. In this paper, a transformer-based inverse design method for CMFs is proposed for the first time, and it can eliminate the need for empirical parameter adjustments by automatically predicting the targeted geometric parameters, thereby improving the design efficiency. In addition, to address the problem of imbalanced data distribution, the multilabel synthetic minority over-sampling technique (MLSMOTE), which can enhance the data representation in sparse sample regions, is implemented. Further validation on tunable CMFs confirms that the proposed inverse design method has broad applicability. The experimental results demonstrate that the proposed inverse design method can accurately predict the geometric parameters and improve the efficiency, thereby providing an innovative solution for the design and applications of CMFs.

The layout method for Radio Frequency/Monolithic Microwave Integrated Circuit (RF/MMIC) is one of the key components in achieving MMIC design automation. Our work proposes a multi-stage progressive automated layout framework addressing MMIC layout challenges under 0.25 μmGaAs pHEMT technology. This framework employs dimensionality reduction from unsupervised learning, integrates practical layout design rules, and achieves automated layout through analytical optimization methods. Applied to multiple cases including filters, low-noise amplifiers (LNAs), and power amplifiers (PAs), the method successfully generated manufacturable layouts; the process achieves end-to-end automation from placement to routing, producing layouts that are verified to be DRC-clean. Simulation-verified results demonstrate comparable performance to manual designs while achieving ≥10× speedup over conventional methods. We believe this work contributes valuable attempts toward automated MMIC layout generation and advances progress in MMIC design automation.

This paper proposes a novel neural network architecture combining convolutional and transposed convolutional neural networks to accurately and efficiently modelS-parameter of interconnects for 3D integration. The network incorporates physical consistency constraints, specifically causality and passivity, into its design to ensure the physical effectiveness of the output. The transposed convolutional network serves as a sub-network to map the relationship between the geometrical parameters andS-parameter for sub-structures. Then, theS-parameters of individual sub-structures are cascaded for dealing with a complex structure composed of sub-structures. A coupling neural network, with causality and passivity constraints, is developed to map the coarse cascadedS-parameters to the fine accurateS-parameters. With the help of this high-dimensional space mapping, a small amount of electromagnetic simulation data of complex interconnect structures is sufficient to learn the relationship between cascaded and realS-parameters. To ensure the completeness of the training set distribution when training CONN on small datasets, a sensitivity analysis-based training set screening method is proposed to enhance the training performance of CONN. The proposed algorithm is demonstrated in two different assemble structure applications. The results highlight the effectiveness, flexibility and versatility of the proposed architecture in modeling complex structures with small costly simulation data while maintaining accuracy and physical consistency.

Artificial intelligence (AI) techniques are increasingly being used for the inverse design of microwave devices. However, several challenges, including intensive computation costs for training samples, high-dimensional data, nonuniformity, and low-quality samples in the design space, can negatively impact the final modeling performance. To alleviate these issues, a high-quality sampling inverse design scheme incorporating wavelet transform deep learning (HQS-WTDL) is proposed to achieve customized, automated microwave filter design. In the forward simulation-based sampling, particle swarm optimization (PSO) is used to tentatively select rule-defined samples. Multilabel synthetic minority over-sampling technique (MLSMOTE) is then applied to enlarge the training samples and improve their uniformity in the design space. An inverse modeling approach using neural networks to map the nonlinear relationship between a given set of S-parameters and required filter structural parameters is presented. Given that the dimension of the S-parameters is much higher than that of the structure parameters, the corresponding neural network used in this approach is deep and complex, with multiple layers and neurons. To reduce the number of input variables, the S-parameters are subjected to wavelet transformation, allowing for more efficient representation by the neural network. The proposed method is validated using a band-pass microstrip hairpin filter as an example. Results demonstrate that the proposed approach achieves better modeling effectiveness and inverse design efficiency than conventional methods. In addition, the proposed method allows for fast customization of device parameters, such as center frequencies and bandwidths with good prediction accuracy.

Analog circuit topology synthesis suffers from weak synthesis capability and low-synthesis efficiency, which result in a bottleneck toward its practical industrial applications. This article presents a proximal-policy-optimization-based circuit topology synthesis framework, which features a superior convergence rate. To further promote its synthesis efficiency, we have improved a deterministic optimization method by incorporating a bias-aware scheme and group concept, which is applied as a filter to eliminate the undesirable topologies in the early evaluation stage. Moreover, a graph-based refinement scheme is proposed to perform deterministically on the generated circuit topologies, which can efficiently add frequency compensation circuits. Compared with the state-of-the-art approaches, our proposed method not only boosts the synthesis efficiency by at least 3 times but also enhances the synthesis capability with a deterministic compensation scheme, showcasing significant advancement of performance efficacy.

As an important technology in high-speed systems, equalizer (EQ) is used to mitigate inter-symbol interference (ISI) caused by inconsistent attenuation of high and low frequencies. The difficulty of signal integrity improvement increases the complexity of EQ design, making the existing algorithms inefficient in high-dimensional searching and constraint processing. In this article, a local multi-constraint modeling-Bayesian optimization (BO) with region partitioning is proposed, aiming to provide a general optimization solution for high-dimensional multi-constraint EQs and improve convergence accuracy and efficiency. The constraint filtering mechanism is used to exclude areas that violate simulation-independent constraints. Local modeling and region partitioning techniques complement each other, taking into account both the local accuracy of the model and the global search performance of the algorithm. The multi-constraint modeling strategy allows simulation-dependent constraints to be pre-judged through the surrogate model, overcoming the shortcomings of the traditional solution of adding the penalty term to the target value, which makes it difficult to balance the weights and can only judge the constraints after simulation, thereby reducing the waste of computing resources caused by simulating data that violates the constraints. The proposed algorithm is applied to EQ optimization in a 16 Gbps high-bandwidth memory channel and a 64 Gbps differential peripheral component interconnect express channel, respectively. The algorithm is developed based on PyTorch, and the eye diagrams are obtained using Keysight ADS software. Two applications are conducted on computer with Intel Core i5-13500 processor and 32 GB RAM. By utilizing the region partitioning and constraint filtering techniques, the actual number of simulations in the optimization can be significantly reduced. The experimental results demonstrate that the proposed algorithm has significant shorter computing time than traditional BO and genetic algorithm, implying its practical application potential for dealing with high-dimensional multi-constraint problems.

Compared with conventional analog circuit topology synthesis methods, the deep-reinforcement-learning (DRL)-based method features much higher synthesis efficiency while possessing the merit of strong generalization capability. However, this method cannot synthesize operational amplifiers that involve signal division. To address this critical limitation, this article presents new synthesis rules to guide the DRL-based synthesis process. In addition, to meet various design specifications requested by users, we further develop a smart circuit synthesis system, which can robustly return a solution (i.e., a feasible circuit topology with detailed device sizes) right away as long as the input design specifications are reasonable. A transfer learning (TL) scheme is proposed to reduce the computation overhead of training this system. The experimental results show the efficacy of our smart circuit synthesis system and TL scheme, confirming an advancement over the state-of-the-art approaches.

Coronavirus disease 2019 (COVID-19) is one of the most destructive pandemic after millennium, forcing the world to tackle a health crisis. Automated lung infections classification using chest X-ray (CXR) images could strengthen diagnostic capability when handling COVID-19. However, classifying COVID-19 from pneumonia cases using CXR image is a difficult task because of shared spatial characteristics, high feature variation and contrast diversity between cases. Moreover, massive data collection is impractical for a newly emerged disease, which limited the performance of data thirsty deep learning models. To address these challenges, Multiscale Attention Guided deep network with Soft Distance regularization (MAG-SD) is proposed to automatically classify COVID-19 from pneumonia CXR images. In MAG-SD, MA-Net is used to produce prediction vector and attention from multiscale feature maps. To improve the robustness of trained model and relieve the shortage of training data, attention guided augmentations along with a soft distance regularization are posed, which aims at generating meaningful augmentations and reduce noise. Our multiscale attention model achieves better classification performance on our pneumonia CXR image dataset. Plentiful experiments are proposed for MAG-SD which demonstrates its unique advantage in pneumonia classification over cutting-edge models. The code is available at https://github.com/JasonLeeGHub/MAG-SD.

XML data warehouses are becoming more popular as data is harvested from the web or as output from web services. As these warehouses tend to grow significantly over time, various techniques for expediting queries have been developed. One such technique is to materialize some or all of the queries in advance of query processing. These views are then subject to change either when underlying data changes or view definitions themselves are modified by users. The work in this paper focuses on changes to view definitions or view adaptation as it is known. Our approach is to segment the materialized view into fragments to minimize the effect of view changes. One crucial aspect to this approach is how to select the best fragments for materialization. In this paper, we introduce a new approach to selecting fragments based on heuristics derived from costs associated with the view graph. Copyright © 2013 John Wiley & Sons, Ltd.

This paper proposes an efficient PSP-based model for cross-coupled metal-oxide-semiconductor field-effect transistors (MOSFETs) with optimized layout in the voltage controlled oscillator (VCO). The model employs a PSP charge model to characterize the bias-dependent extrinsic capacitance instead of numerical functions with strong non-linearity. The simulation convergence is greatly improved by this method. An original scheme is developed to extract the parameters of the PSP charge model based on S -parameters measurement. The interconnection parasitics of the cross-coupled MOSFETs are modeled based on vector fitting. The model is verified with an LC VCO design, and exhibits excellent convergence during simulation. The results show improvements as high as 60.5% and 61.8% in simulation efficiency and accuracy, respectively, indicating that the proposed model better characterizes optimized cross-coupled MOSFETs in advanced radio frequency (RF) circuit design.

A new compact model for the substrate network of RF MOSFETs with deep n-well (DNW) implantation is presented. A novel test structure proposed in is employed to directly access the characteristics of the substrate in two-port measurements for the extraction of substrate network components. A method is developed to analytically extract the parameters for the substrate network from two-port measurements. The methodology is verified and validated by the excellent match between the measured and simulated output admittances for a 64-finger DNW n-MOSFET in common-source configuration.

Many of the emergent digital ecosystems will employ sensor networks to generate data. Using XML to introduce structure and semantics assists the ecosystem as standard XML query languages such as XPath and XQuery are used to extract information and results of analyses. However, the creation of XML digital archives is hindered by the performance of these query languages. Furthermore, the multi-disciplinary nature of digital ecosystems means that knowledge workers and end users will rarely be IT professionals and therefore, unable to express XPath or XQuery easily. In this work, we present a method for precomputing and storing query results, leading to far higher levels of performance and removing the requirement for non-IT users to learn complex query languages.