SWTRU: Star-shaped Window Transformer Reinforced U-Net for medical image segmentation.

In the last decade, deep neural networks have been widely applied to medical image segmentation, achieving good results in computer-aided diagnosis tasks etc. However, the task of segmenting highly complex, low-contrast images of organs and tissues with high accuracy still faces great challenges. To better address this challenge, this paper proposes a novel model SWTRU (Star-shaped Window Transformer Reinforced U-Net) by combining the U-Net network which plays well in the image segmentation field, and the Transformer which possesses a powerful ability to capture global contexts. Unlike the previous methods that import the Transformer into U-Net, an improved Star-shaped Window Transformer is introduced into the decoder of the SWTRU to enhance the decision-making capability of the whole method. The SWTRU uses a redesigned multi-scale skip-connection model, which retains the inductive bias of the original FCN structure for images while obtaining fine-grained features and coarse-grained semantic information. Our method also presents the FFIM (Filtering Feature Integration Mechanism) to integration and dimensionality reduction of the fused multi-layered features, which reduces the computation. Our SWTRU yields 0.972 DICE on CHLISC for liver and tumor segmentation, 0.897 DICE on LGG for glioma segmentation, and 0.904 DICE on ISIC2018 for skin diseases' segmentation, achieves substantial improvements over the current SoTA across 9 different medical image segment methods. SWTRU can combine feature mapping from different scales, high-level semantics, and global contextual relationships, this architecture is effective in the medical image segmentation. The experimental findings indicate that SWTRU produces superior performance on the medical image segmentation tasks.

Differences in actionable genomic alterations between brain metastases and non­brain metastases in patients with non­small cell lung cancer.

Brain metastases (BM) have been closely associated with increased morbidity and poor survival outcomes in patients with non­small cell lung cancer (NSCLC). Excluding risk factors in histological subtypes, genomic alterations, including epidermal growth factor receptor mutations and anaplastic lymphoma kinase (ALK) rearrangements have been also regarded as greater risk factors for BM in the aspect of molecular subtypes. In the present study, 69 tumor tissues and 51 peripheral blood samples from patients with NSCLC were analyzed using a hybridization capture­based next­generation sequencing (NGS) panel, including 95 known cancer genes. Among the 90 patients with stage IV NSCLC, 26 cases suffered from BM and 64 cases did not. In total, 174 somatic mutations in 35 mutated genes were identified, and 12 of these genes were concurrently present in the BM group and the non­BM group. Importantly, five mutated genes including ALK, cytidine deaminase (CDA), SMAD family member 4 (SMAD4), superoxide dismutase 2 (SOD2) and Von Hippel­Lindau tumor suppressor (VHL) genes were uniquely detected in the BM group, and they were enriched in the Hippo signaling pathway, pyrimidine metabolism and pantothenate and co­enzyme A (CoA) biosynthesis, as demonstrated using Kyoto Encyclopedia of Genes and Genomes enrichment analysis. RNA polymerase II transcription regulator complex and promyelocytic leukemia nuclear body were the top functional categories according to the Gene Ontology enrichment analysis in the BM group and non­BM group, respectively. Furthermore, 43.33% (13/30) of mutated genes were detected by both tumor tissue deoxyribonucleic acid (DNA) and plasma­derived circulating tumor DNA (ctDNA) in the non­BM group, while this percentage was only limited to 29.41% (5/17) in the BM group. To summarize, significant differences in somatic mutations, somatic interactions, key signaling pathways, functional biological information, and clinical actionability for the therapy of targeted agents were founded between the BM group and the non­BM group, and ctDNA analysis may by applied as a more credible alternative for genomic profiling in patients with advanced NSCLC without BM, due to its higher consistency for genomic profiling between ctDNA analysis and tissue DNA analysis.

EraseNet: End-to-End Text Removal in the Wild.

Scene text removal has attracted increasing research interests owing to its valuable applications in privacy protection, camera-based virtual reality translation, and image editing. However, existing approaches, which fall short on real applications, are mainly because they were evaluated on synthetic or unrepresentative datasets. To fill this gap and facilitate this research direction, this paper proposes a real-world dataset called SCUT-EnsText that consists of 3,562 diverse images selected from public scene text reading benchmarks, and each image is scrupulously annotated to provide visually plausible erasure targets. With SCUT-EnsText, we design a novel GANbased model termed EraseNet that can automatically remove text located on the natural images. The model is a two-stage network that consists of a coarse-erasure sub-network and a refinement sub-network. The refinement sub-network targets improvement in the feature representation and refinement of the coarse outputs to enhance the removal performance. Additionally, EraseNet contains a segmentation head for text perception and a local-global SN-Patch-GAN with spectral normalization (SN) on both the generator and discriminator for maintaining the training stability and the congruity of the erased regions. A sufficient number of experiments are conducted on both the previous public dataset and the brand-new SCUT-EnsText. Our EraseNet significantly outperforms the existing state-of-the-art methods in terms of all metrics, with remarkably superior higherquality results. The dataset and code will be made available at https://github.com/HCIILAB/SCUT-EnsText.

HL2 on chromosome 7D of wheat (Triticum aestivum L.) regulates both head length and spikelet number.

KEY MESSAGE: A major QTL QSpl.nau-7D, named HL2, was validated for its effects on head length and kernel number per spike using NIL, and mapped to a 0.2 cM interval using recombinants. Improvement in wheat inflorescence traits such as spike or head length and spikelet number provides an important avenue to increase grain yield potential. In a previous study, QSpl.nau-7D, the major QTL for head length on chromosome 7D, was identified in the recombinant inbred lines derived from Nanda2419 and Wangshuibai. To validate and precisely map this QTL, the Wangshuibai allele was transferred to elite cultivar Yangmai15 through marker-assisted selection. Compared with the recurrent parent, the resultant near-isogenic line (NIL) yielded not only 28% longer spikes on the average but also more spikelets and kernels per spike. Moreover, the NIL had a lower spikelet density and did not show significant kernel weight change. In the F2 population derived from the NIL, QSpl.nau-7D acted like a single semi-dominant gene controlling head length and was therefore designated as Head Length 2 (HL2). With this population, a high-density genetic map was constructed mainly using newly developed markers, and 100 homozygous recombinants including 17 genotypes were obtained. Field experiments showed that the recombinants carrying the 0.2-cM interval flanked by Xwgrb1588 and Xwgrb1902 from Wangshuibai produced longer spikes than those without this Wangshuibai allele. Comparative mapping of this interval revealed a conserved synteny among cereal grasses. HL2 is beneficial to wheat breeding for more kernels per spike at a lower spikelet density, which is a favored morphological trait for Fusarium head blight resistance.

TextField: Learning a Deep Direction Field for Irregular Scene Text Detection.

Scene text detection is an important step in the scene text reading system. The main challenges lie in significantly varied sizes and aspect ratios, arbitrary orientations, and shapes. Driven by the recent progress in deep learning, impressive performances have been achieved for multi-oriented text detection. Yet, the performance drops dramatically in detecting the curved texts due to the limited text representation (e.g., horizontal bounding boxes, rotated rectangles, or quadrilaterals). It is of great interest to detect the curved texts, which are actually very common in natural scenes. In this paper, we present a novel text detector named TextField for detecting irregular scene texts. Specifically, we learn a direction field pointing away from the nearest text boundary to each text point. This direction field is represented by an image of 2D vectors and learned via a fully convolutional neural network. It encodes both binary text mask and direction information used to separate adjacent text instances, which is challenging for the classical segmentation-based approaches. Based on the learned direction field, we apply a simple yet effective morphological-based post-processing to achieve the final detection. The experimental results show that the proposed TextField outperforms the state-of-the-art methods by a large margin (28% and 8%) on two curved text datasets: Total-Text and SCUT-CTW1500, respectively; TextField also achieves very competitive performance on multi-oriented datasets: ICDAR 2015 and MSRA-TD500. Furthermore, TextField is robust in generalizing unseen datasets.

Association between rs120963, rs152451, rs249935, rs447529, rs8053188, and rs16940342 Polymorphisms in the PALB2 Gene and Breast Cancer Susceptibility: A Meta-Analysis.

BACKGROUND: The aim of this study was to explore the association between single nucleotide polymorphisms (SNPs) in the rs120963, rs152451, rs249935, rs447529, rs8053188, and rs16940342 loci in the PALB2 gene and breast cancer risk. METHODS: Studies investigating the association between SNPs in the PALB2 gene and breast cancer susceptibility were retrieved from the PubMed, Embase, Web of Science, CNKI (Chinese National Knowledge Infrastructure), WanFang, and CBM (China Biology Medicine) databases. Eligible studies were screened according to inclusion/exclusion criteria and principles of quality evaluation. Meta-analysis was performed using Stata 14.0 software. Odds ratios with their corresponding 95% confidence intervals were pooled to assess the association between SNPs in the PALB2 gene loci rs249935, rs447529, rs8053188, rs16940342, rs152451, and rs120963 and breast cancer susceptibility. RESULTS: A total of 9 case-control studies were eligible for this meta-analysis. SNPs in the PALB2 gene loci rs120963, rs249935, and rs447529 were significantly associated with an increased or decreased risk of breast cancer. No significant association was detected for rs152451, rs8053188, and rs16940342 under 4 genetic models. CONCLUSION: The results of this study suggest that SNPs in the PALB2 loci rs120963/rs249935/rs447529, but not in the other 3 loci (rs152451/rs8053188/rs16940342), may contribute to breast cancer susceptibility.