In semi-supervised medical image segmentation, most previous works draw on the common assumption that higher entropy means higher uncertainty. In this paper, we investigate a novel method of estimating uncertainty. We observe that, when assigned different misclassification costs in a certain degree, if the segmentation result of a pixel becomes inconsistent, this pixel shows a relative uncertainty in its segmentation. Therefore, we present a new semi-supervised segmentation model, namely, conservative-radical network ( CoraNet in short) based on our uncertainty estimation and separate self-training strategy. In particular, our CoraNet model consists of three major components: a conservative-radical module (CRM), a certain region segmentation network (C-SN), and an uncertain region segmentation network (UC-SN) that could be alternatively trained in an end-to-end manner. We have extensively evaluated our method on various segmentation tasks with publicly available benchmark datasets, including CT pancreas, MR endocardium, and MR multi-structures segmentation on the ACDC dataset. Compared with the current state of the art, our CoraNet has demonstrated superior performance. In addition, we have also analyzed its connection with and difference from conventional methods of uncertainty estimation in semi-supervised medical image segmentation.

Rail surface defects (RSDs) are a major problem that reduces operation safety. Unfortunately, the existing RSD detection systems have very limited accuracy. Current image processing methods are not tailored for the railway track and many fully convolutional networks (FCN)-based methods suffer from the blurry rail edges (RE). This paper proposes a new rail boundary guidance network (RBGNet) for salient RS detection. First, a novel architecture is proposed to fully utilize the complementarity between the RS and the RE to accurately identify the RS with well-defined boundaries. The newly developed RBGNet injects high-level RS object information into shallow RS edge features by a progressive fused way for obtaining fine edge features. Then, the system integrates the refined edge features with RS features at different high-level layers to predict the RS precisely. Second, an innovative hybrid loss consisting of binary cross entropy (BCE), structural similarity index measure (SSIM), and intersection-over-union (IoU) is proposed and equipped into the RBGNet to supervise the network and learn the transformation between the input and ground truth. The input and ground truth then further refine the RS location and edges. Conveniently, an image-based model for RSD detection and quantification is also developed and integrated for an automatic inspection purpose. Finally, experiments conducted on the complex unmanned aerial vehicle (UAV) rail dataset indicate the system can achieve a high detection rate with good adaptation capability in complicated environments.

Due to the unpredictable location, fuzzy texture and diverse shape, accurate segmentation of the kidney tumor in CT images is an important yet challenging task. To this end, we in this paper present a cascaded trainable segmentation model termed as Crossbar-Net. Our method combines two novel schemes: (1) we originally proposed the crossbar patches, which consists of two orthogonal non-squared patches (i.e., the vertical patch and horizontal patch). The crossbar patches are able to capture both the global and local appearance information of the kidney tumors from both the vertical and horizontal directions simultaneously. (2) With the obtained crossbar patches, we iteratively train two sub-models (i.e., horizontal sub-model and vertical sub-model) in a cascaded training manner. During the training, the trained sub-models are encouraged to become more focus on the difficult parts of the tumor automatically (i.e., mis-segmented regions). Specifically, the vertical (horizontal) sub-model is required to help segment the mis-segmented regions for the horizontal (vertical) sub-model. Thus, the two sub-models could complement each other to achieve the self-improvement until convergence. In the experiment, we evaluate our method on a real CT kidney tumor dataset which is collected from 94 different patients including 3,500 CT slices. Compared with the state-of-the-art segmentation methods, the results demonstrate the superior performance of our method on the Dice similarity coefficient, true positive fraction, centroid distance and Hausdorff distance. Moreover, to exploit the generalization to other segmentation tasks, we also extend our Crossbar-Net to two related segmentation tasks: (1) cardiac segmentation in MR images and (2) breast mass segmentation in X-ray images, showing the promising results for these two tasks. Our implementation is released at https: //github.com/Qianyu1226/Crossbar-Net.

Unsupervised domain adaptation (UDA) methods have shown their promising performance in the cross-modality medical image segmentation tasks. These typical methods usually utilize a translation network to transform images from the source domain to target domain or train the pixel-level classifier merely using translated source images and original target images. However, when there exists a large domain shift between source and target domains, we argue that this asymmetric structure, to some extent, could not fully eliminate the domain gap. In this paper, we present a novel deep symmetric architecture of UDA for medical image segmentation, which consists of a segmentation sub-network, and two symmetric source and target domain translation sub-networks. To be specific, based on two translation sub-networks, we introduce a bidirectional alignment scheme via a shared encoder and two private decoders to simultaneously align features 1) from source to target domain and 2) from target to source domain, which is able to effectively mitigate the discrepancy between domains. Furthermore, for the segmentation sub-network, we train a pixel-level classifier using not only original target images and translated source images, but also original source images and translated target images, which could sufficiently leverage the semantic information from the images with different styles. Extensive experiments demonstrate that our method has remarkable advantages compared to the state-of-the-art methods in three segmentation tasks, such as cross-modality cardiac, BraTS, and abdominal multi-organ segmentation.

In the United States, to ensure railroad safety and keep its efficient operation, regular track inspections on track component defects are required by the Federal Railroad Administration (FRA). Various types of inspection equipment are applied, such as ground penetrating radar, laser, and LiDAR, but they are usually very expensive and require extensive training and rich experience to operate. To date, track inspections still rely heavily on manual inspections which are low‐efficiency, subjective, and not as accurate as desired, especially for missing and broken track components, such as spikes, clips, and tie plates. To address this issue, a real‐time pixel‐level rail components detection framework to inspect tracks timely and accurately is proposed in this study. The first public rail components image database, including rails, spikes, and clips, is built and released online. A real‐time pixel‐level detection framework with improved real‐time instance segmentation models is developed. The improved models leverage fast object detection and highly accurate instance segmentation. Backbones with more granular levels and receptive fields are implemented in the proposed models. Compared with the original YOLACT and Mask R‐CNN models, the proposed models are able to: (1) achieve 59.9 bbox mAP, and 63.6 mask mAP with the customized dataset, which are higher than the other models and (2) achieve a real‐time speed which is over 30 FPS processing a high‐resolution video (1,080 × 1,092) with a single GPU. The fast processing speed can quickly turn inspection videos into useful information to assist track maintenance. The railroad track components image dataset can be accessed at https://github.com/jonguo111/Rail_components_image_data

Geogrid (uniaxial or biaxial) is one type of geosynthetics that has been successfully used in slopes, walls, roads, and other applications. The main application of biaxial geogrid is to stabilize soft subgrade and reinforce weak base courses by providing lateral confinement. The confinement due to the interaction between aggregates and the ribs of biaxial geogrid depends on the geometry properties of the geogrid, such as rib shape and apertures size, the stiffness of the ribs, and the properties of aggregates. Research has shown that biaxial geogrid cannot provide uniform tensile resistance in all directions. To overcome this problem, a geogrid product with triangular apertures was developed and introduced into the market. Recent studies showed that the triangular aperture geogrid can provide nearly uniform tensile resistance in all directions and is more efficient in improving the performance of reinforced bases as compared with biaxial geogrid. However, the performance of triangular aperture geo...