Search

Laparoscopic surgery is a technique in which surgeons insert laparoscopic instruments into the body through a small incision and conduct surgical operations. It offers a shorter recovery period, less blood loss, and less postoperative pain than open surgeries. In order to maximize these benefits of this surgery technique, it is important to localize the lesion accurately without palpation, minimizing the surgical area. It is because surgeons cannot palpate the organ during laparoscopic surgeries. For accurate localization, we propose a lesion localization system (LLS). This adopts an endoscopic clip with a small-size magnet as a marker and an improved magnetometer as a probe for measuring the magnetic field from the magnet. The magnetometer, which consists of two Hall sensors, can cancel out the fluctuating magnetic baseline mainly caused by external environments so that it can detect a smaller change in the target magnetic field. As a result, the LLS improves the detection range up to 40 mm while using a weak magnet with a small volume of 17.1 mm3, which is compatible with the use of commercial clip appliers. An error of less than 0.1% is achieved at a distance of 40 mm, and the maximum error is kept below 6% even when the rotation angle is varied to ±90°. This LLS does not cause discomfort to the surgery operation because it uses a marker small enough and does not require external coils.

In this study, a hybrid bipolar arc plasma stimulation (BAPS) and dual pulse electroporation (DPEP) technique was used to form microchannels, and the effect of highly efficient transformation drug delivery (TDD) on the skin was analyzed. The BAPS technique was applied to pig skin to form a microchannel as a biological sample to measure and observe the plasma as a micropore. During the BAPS, an output voltage of 1 kVpp at 40 kHz was applied at duty rates of 20%, 40%, and 60%. The BAPS technique was applied to the keratin skin layer, and microchannels with an average size of $30\,\,\mu \text{m}\,\,\times 30\,\,\mu \text{m}$ and up to $77\,\,\mu \text{m}\,\,\times 55\,\,\mu \text{m}$ were formed only in the keratin layer without any damage to the dermal layer. Next, DPEP was stimulated to increase the permeability of the endothelial membrane in the structure with the microchannels. The proposed hybrid technique effectively introduces TDD deeper into the employed 20 female subjects skin test with average age of 44 years, and the results in the absorption rate increase about 418%.

In the successive cancellation list decoding of polar codes, the metric sorting dominates the overall decoding latency. To reduce the latency of metric sorting, this paper proposes a new sorting method, called interleaved local sorting, which divides the metrics to be sorted into several groups and locally sorts each group independently. In addition, an interleaving scheme is proposed to recover the performance degradation caused by the local sorting. A hardware architecture effective in reducing the overall latency as well as the hardware complexity is also proposed based on the proposed metric sorting. The evaluation results show that the proposed interleaved local sorting architecture outperforms the state-of-the-art metric sorting architectures in terms of latency and hardware complexity when the list size is not small.

The key to the depth estimation from a single image lies in inferring the distance of various objects without copying texture while maintaining clear object boundaries. In this paper, we propose depth estimation from a single image using edge extraction network and dark channel prior (DCP). We build an edge extraction network based on generative adversarial networks (GANs) to select valid depth edges from a number of edges in an image. We use DCP to generate a transmission map that is able to represent distance from the camera. Transmission map is generated by conducting minimum value filtering on DCP. First, we concatenate the transmission map with the original RGB image to form a tensor, i.e. RGB + T. Second, we generate an initial depth image from the tensor through the generator by inferring depth from stacked residual blocks. Third, we compare the edge map of the initial depth image with that of the input RGB image to select valid depth edges. Both edge maps are generated by the edge extraction network. Finally, we distinguish real and fake on the generated depth image using a discriminator, and enhances the performance of the generator. Various experiments on NYU, Make3D and MPI Sintel datasets demonstrate that the proposed network generates clear edges in depth images as well as outperforms state-of-the-art methods in terms of visual quality and quantitative measurements.

In recent industrial applications, machine learning technology is proving useful in preventing equipment failures in advance through early failure diagnosis. In particular, we show that different domains can be linked through adversarial learning with data available in different working conditions to facilitate the training of the model, as it is impractical to acquire data for all conditions in real-world applications. Nevertheless, the initial failure is a difficult problem to diagnose because it does not show a significant difference from the normal data between different conditions. Moreover, if only the domain discriminator is judged when adapting the domain, it tends to easily cause misclassification, so the reliability of the detection result needs to be improved. In this study, we propose a new learning method that improves classification performance by sharing the classification characteristics of the classifier for each task with the target domain characteristic generator. The proposed mechanism uses spatial attention to extract the focused partial information of the feature generator and discriminator, and further enhances task-specific features using the attention mechanism between the two extracted information types. Addresses the challenge of implementing both domain adaptation and classification. Extensive experimentation demonstrates efficiency and improved classification performance on benchmark and real-world application datasets. In real machine cases, the classification accuracy is improved by almost 4%. In addition, the negative impact on false alarms was lowered by increasing the classification accuracy of minimum failures. Convincingly demonstrate model effectiveness by performing an empirical analysis of the method through ablation analysis and visualization.

X-ray baggage inspection has been widely used for maintaining airport and transportation security. Towards automated inspection, recent deep learning-based methods have attempted to detect hazardous objects directly from X-ray images. Since it is challenging to collect a large number of training images from real-world environments, most previous learning-based methods rely on image synthesis for training data generation. However, these methods randomly combine foreground and background images, restricting the effectiveness of synthetic images for object detection. To solve this problem, in this paper, we propose a learning-based X-ray image synthesis method for object detection. Specifically, for each foreground object to be synthesized, we first estimate positions difficult to detect by the object detector. These positions and their corresponding confidence values are then used to construct a difficulty map, which is used for sampling the target foreground position for image synthesis. The performance analysis using various state-of-the-art object detectors shows that the proposed synthesis method can produce more useful training data compared with the conventional random synthesis method.

X-ray imaging is one of the most widely used security measures for maintaining airport and transportation security. Conventional X-ray imaging systems typically apply tone-mapping (TM) algorithms to visualize high-dynamic-range (HDR) X-ray images on a standard 8-bit display device. However, X-ray images obtained through traditional TM algorithms often suffer from halo artifacts or detail loss in inter-object overlapping regions, which makes it difficult for an inspector to detect unsafe or hazardous objects. To alleviate these problems, this article proposes a deep learning-based TM method for X-ray inspection. The proposed method consists of two networks called detail-recovery network (DR-Net) and TM network (TM-Net). The goal of DR-Net is to restore the details in the input HDR image, whereas TM-Net aims to compress the dynamic range while preserving the restored details and preventing halo artifacts. Since there are no standard ground-truth images available for the TM of X-ray images, we propose a novel loss function for unsupervised learning of TM-Net. We also introduce a dataset synthesis technique using the Beer-Lambert law for supervised learning of DR-Net. Extensive experiments comparing the performance of our proposed method with state-of-the-art TM methods demonstrate that the proposed method not only achieves visually compelling results but also improves the quantitative performance measures such as FSITM and HDR-VDP-2.2.

Manufacturing firms are facing the challenge of minimizing variety-induced complexity in assembly systems. One of the effective approaches to complexity management is to reconfigure the assembly system by rearranging its assembly sequence. In order to rearrange an assembly sequence, a design-oriented approach is necessary because assembly is an activity of connecting parts through an interface between them. It means that the assembly sequence in production is restricted by the structure of interface connections in design. In this vein, this paper introduces a new design-oriented approach called interface design approach to complexity management in assembly systems. First, the mechanism of how the structure of interface connections affects assembly system complexity is identified. Then, an interface design framework is proposed for effectively finding an optimal interface structure and its assembly sequence to minimize assembly system complexity. For evaluating the complexity, the operator choice complexity index is adopted and modified for the interface design problem. In the case study, the framework is applied to the interface design problem by using industrial data of a plasma display panel (PDP) family from LG Electronics. The result of the study demonstrates that the assembly system complexity is significantly reduced by the proposed interface design framework.

Log-Structured Merge-tree (LSM-tree) organizes write-friendly and hierarchical structure, which leads to inevitable disk I/O from data compaction occuring between layers. Previous research tried to reduce write bottlenecks of LSM-tree by using non-volatile memory (NVM) to LSM-tree, but write amplification on compaction does not resolved. In this paper, we present tiered LSM (TLSM), a persistent LSM-based key-value database tiering NVM, which is regarded as fast storage device. Our design aims to exploit multi-layered LSM system utilizing NVM and reduce redundant cost on compaction. We address 4 key points: (1) analysis of disk I/O overhead on current LSM-tree, (2) novel design of TLSM preserving the concept of LSM-tree and utilizing NVM, (3) persistent and byte-addressable data compaction minimizing disk write, and (4) tiering policy for increasing usability to TLSM. To implement TLSM-tree, we extended LevelDB, which is a simple LSM-based key-value store. On micro-benchmark DB Bench, the write and read latencies of TLSM are improved up to 34% and 57%, respectively, compared to those of LevelDB-NVM. In evaluation of YCSB, TLSM also reduces the write and read latencies of existing LevelDB up to 25%, 33%, respectively. Furthermore, frequency of write stall and total amount of data write are reduced significantly in TLSM.

In this paper, we propose an advanced three-dimensional visualization method for an integral imaging microscope system to simultaneously improve the resolution and quality of the reconstructed image. The main advance of the proposed method is that it generates a high-quality three-dimensional model without limitation of resolution by combining the high-resolution two-dimensional color image with depth data obtained through a fully convolutional neural network. First, the high-resolution two-dimensional image and an elemental image array for a specimen are captured, and the orthographic-view image is reconstructed from the elemental image array. Then, via a convolutional neural network-based depth estimation after the brightness of input images are uniformed, a more accurate and improved depth image is generated; and the noise of result depth image is filtered. Subsequently, the estimated depth data is combined with the high-resolution two-dimensional image and transformed into a high-quality three-dimensional model. In the experiment, it was confirmed that the displayed high-quality three-dimensional model could be visualized very similarly to the original image.