Segment anything model (SAM), a foundation model with superior versatility and generalization across diverse segmentation tasks, has attracted widespread attention in medical imaging. However, it has been proved that SAM would encounter severe performance degradation due to the lack of medical knowledge in training and local feature encoding. Though several SAM-based models have been proposed for tuning SAM in medical imaging, they still suffer from insufficient feature extraction and highly rely on high-quality prompts. In this paper, we propose a powerful foundation model SAMCT allowing labor-free prompts and train it on a collected large CT dataset consisting of 1.1M CT images and 5M masks from public datasets. Specifically, based on SAM, SAMCT is further equipped with a U-shaped CNN image encoder, a cross-branch interaction module, and a task-indicator prompt encoder. The U-shaped CNN image encoder works in parallel with the ViT image encoder in SAM to supplement local features. Cross-branch interaction enhances the feature expression capability of the CNN image encoder and the ViT image encoder by exchanging global perception and local features from one to the other. The task-indicator prompt encoder is a plug-and-play component to effortlessly encode task-related indicators into prompt embeddings. In this way, SAMCT can work in an automatic manner in addition to the semi-automatic interactive strategy in SAM. Extensive experiments demonstrate the superiority of SAMCT against the state-of-the-art task-specific and SAM-based medical foundation models on various tasks. The code, data, and model checkpoints are available at https://github.com/xianlin7/SAMCT.

Despite the great potential in capturing long-range dependency, one rarely-explored underlying issue of transformer in medical image segmentation is attention collapse, making it often degenerate into a bypass module in CNN-Transformer hybrid architectures. This is due to the high computational complexity of vision transformers requiring extensive training data while well-annotated medical image data is relatively limited, resulting in poor convergence. In this paper, we propose a plug-n-play transformer block with dynamic token merging, named DTMFormer, to avoid building long-range dependency on redundant and duplicated tokens and thus pursue better convergence. Specifically, DTMFormer consists of an attention-guided token merging (ATM) module to adaptively cluster tokens into fewer semantic tokens based on feature and dependency similarity and a light token reconstruction module to fuse ordinary and semantic tokens. In this way, as self-attention in ATM is calculated based on fewer tokens, DTMFormer is of lower complexity and more friendly to converge. Extensive experiments on publicly-available datasets demonstrate the effectiveness of DTMFormer working as a plug-n-play module for simultaneous complexity reduction and performance improvement. We believe it will inspire future work on rethinking transformers in medical image segmentation. Code: https://github.com/iam-nacl/DTMFormer.

Designers often create visualizations to achieve specific high-level analytical or communication goals. These goals require people to naturally extract complex, contextualized, and interconnected patterns in data. While limited prior work has studied general high-level interpretation, prevailing perceptual studies of visualization effectiveness primarily focus on isolated, predefined, low-level tasks, such as estimating statistical quantities. This study more holistically explores visualization interpretation to examine the alignment between designers’ communicative goals and what their audience sees in a visualization, which we refer to as their comprehension. We found that statistics people effectively estimate from visualizations in classical graphical perception studies may differ from the patterns people intuitively comprehend in a visualization. We conducted a qualitative study on three types of visualizations—line graphs, bar graphs, and scatterplots—to investigate the high-level patterns people naturally draw from a visualization. Participants described a series of graphs using natural language and think-aloud protocols. We found that comprehension varies with a range of factors, including graph complexity and data distribution. Specifically, 1) a visualization’s stated objective often does not align with people’s comprehension, 2) results from traditional experiments may not predict the knowledge people build with a graph, and 3) chart type alone is insufficient to predict the information people extract from a graph. Our study confirms the importance of defining visualization effectiveness from multiple perspectives to assess and inform visualization practices.

The advent of social networks, mobile sensing, and the Internet of Things herald an age of data overload, where the amount of data generated and stored by various data services exceeds application consumption needs. In such an age, an increasingly important need of data clients will be one of data sub-sampling. This need calls for novel data dissemination protocols that allow clients to request from the network a representative sampling of data that matches a query. In this paper, we present the design of a new transport-layer dissemination protocol, called InfoMax, that allows applications to request such a data sampling. InfoMax exploits the recently proposed named-data-networking (NDN) stack that makes networks aware of hierarchical data names, as opposed to IP addresses. Assuming that named objects with longer prefixes are semantically more similar, InfoMax has the property of minimizing semantic redundancy among delivered data items, hence offering the best coverage of the requested topic with the fewest bytes. The paper discusses the design of InfoMax, its experimental evaluation, and example applications.

With the powerful sensing capability of mobile smart devices, users can easily obtained the crowd sensing services with smart devices in the Internet of Things (IoT). However, credible interaction issues between mobile users are still the hard problems in the past. In this paper, we focus on how to assign the crowdsourcing sensing tasks based on the credible interaction between users. First, a novel credible crowdsourcing assignment model is proposed based on social relationship cognition and community detection. Second, the service quality factor (SQF), link reliability factor (LRF), and region heat factor (RHF) are introduced to scientifically evaluate the user crowdsourcing preferences. Then, a crowdsourcing algorithm based on analytic hierarchy process (AHP) theory is proposed. Finally, the simulation experiments prove the correctness, effectiveness, and robustness of our method.