Chest X-ray (CXR) is the most frequently ordered imaging test, supporting diverse clinical tasks from thoracic disease detection to postoperative monitoring. However, task-specific classification models are limited in scope, require costly labeled data, and lack generalizability to out-of-distribution datasets. To address these challenges, we introduce CheXFound, a self-supervised vision foundation model that learns robust CXR representations and generalizes effectively across a wide range of downstream tasks. We pretrained CheXFound on a curated CXR-987K dataset, comprising over approximately 987K unique CXRs from 12 publicly available sources. We propose a Global and Local Representations Integration (GLoRI) head for downstream adaptations, by incorporating fine- and coarse-grained disease-specific local features with global image features for enhanced performance in multilabel classification. Our experimental results showed that CheXFound outperformed state-of-the-art models in classifying 40 disease findings across different prevalence levels on the CXR-LT 24 dataset and exhibited superior label efficiency on downstream tasks with limited training data. Additionally, CheXFound achieved significant improvements on downstream tasks with out-of-distribution datasets, including opportunistic cardiovascular disease risk estimation, mortality prediction, malpositioned tube detection, and anatomical structure segmentation. The above results demonstrate CheXFound's strong generalization capabilities, which will enable diverse downstream adaptations with improved label efficiency in future applications. The project source code is publicly available at https://github.com/RPIDIAL/CheXFound.

Automatic segmentation of abdominal organs in CT is of essential importance for radiation therapy and image-guided surgery. However, the development of such automatic solutions remains challenging due to complicated structures and low tissue contrast in abdominal CT images. To address these issues, we propose a novel deep neural network equipped with an edge detection (ED) module and a graph-based regional feature enhancing (GRFE) module for better organ segmentation, by enhancing the long-range representation power of regional features. Specifically, the proposed ED module learns an edge representation by leveraging both fine-grained and structural information. The edge representation is then fused with the segmentation features to provide constraint guidance for better prediction. Our GRFE module propagates features to capture contextual information via graphic voxel-by-voxel connections. The GRFE module leverages the edge representation to highlight the features of boundaries to build strong contextual dependencies between the features of organs' boundaries and central areas. We evaluate the efficacy of the proposed network on two challenging abdominal multi-organ datasets. Experimental results demonstrate that our network outperforms several state-of-the-art methods. The code is publicly available at https://github.com/zefanyang/organseg_dags.

Ground penetrating radar (GPR) is a recognized nondestructive testing technique, which has been commonly applied to detect the steel bars (rebars) in concrete. However, most commercial GPR systems can only record reflection signals in single polarization, making the inspection of rebar corrosion difficult. In this paper, we employ a full polarization GPR system to record polarimetric information and use H-Alpha polarization decomposition to evaluate the rebar corrosion process. A yearly long corrosion process of one rebar was accelerated within 15 days by applying a constant current density of 0.3 mA/cm2on the embedded rebar. The preliminary experimental results of polarization decomposition show that the scattering characteristics of rebars change to low entropy surface scattering after corrosion. It is concluded that full-polarimetric GPR has a potential for characterization of the early-stage corrosion of concrete rebar.