Candidate Regions Research Articles

Artificial Intelligence-Oriented User Interface Design and Human Behavior Recognition based on Human–Computer Nature Interaction

This work is to explore the application of intelligent algorithms based on deep learning in human–computer interaction systems, hoping to promote the development of human–computer interaction systems in the field of behavior recognition. Firstly, the design scheme of the human–computer interaction system is presented, and the establishment of the robot visual positioning system is emphasized. Then, the fast-region convolutional neural networks (fast-RCNN) algorithm is introduced, and it is combined with deep convolutional residual network (ResNet101). A candidate region extraction algorithm based on ResNet and long short-term memory network is proposed, and a residual network (ResNet) for spatial context memory is proposed. Both algorithms are employed in human–computer interaction systems. Finally, the performance of the algorithm and the human–computer interaction system are analyzed and characterized. The results show that the proposed candidate region extraction algorithm can significantly reduce the loss value of training set and test set after training. In addition, the corresponding accuracy, recall, and F-value of the model are all above 0.98, which proves that the model has a good detection accuracy. Spatial context memory ResNet shows good accuracy in speech expression detection. The detection accuracy of single attribute, double attribute, and multi-attribute speech expression is above 89%, and the detection accuracy is good. In summary, the human–computer interaction system shows good performance in capturing target objects, even for unlabeled objects, the corresponding grasping success rate is 95%. Therefore, this work provides a theoretical basis and reference for the application of intelligent optimization algorithm in human–computer interaction system.

Video flame recognition based on α-GMM and weight kernel sparse representation

Aiming at the problems of incomplete extraction on candidate flame region, a video flame recognition method is proposed based on Gaussian mixture model with adaptive learning rate α (α-GMM) and weight kernel sparse representation. Firstly, a Gaussian mixture model (GMM) for foreground extraction is constructed. Its learning rate α is adjusted adaptively according to the complexity of background change, and candidate flame region is extracted completely by combining the color features in the HSI space. Secondly, dynamic and static features of the flame are extracted from the candidate region, and a feature dictionary is constructed. Finally, a weighted kernel sparse representation classification model based on Mahalanobis distance (MD) is established to implement flame recognition. The kernel function is adopted to solve the nonlinear distribution problem effectively. In order to strengthen the discrimination between classes and improve the flame recognition rate, MD is employed to measure the similarity information between data and construct the weight matrix. The experiment results show that the candidate flame region obtained by the proposed α-GMM is more complete, and the segmentation accuracy is higher. Compared with other classifiers, the accuracy of the proposed weight kernel sparse representation classifier is improved by 10.94% on average, and the false positive rate is reduced by 55.19% on average, which indicates that the proposed method has a higher recognition rate and strong robustness.

Design and characterization of a high-resolution multiple-SNP capture array by target sequencing for sheep.

The efficiency of molecular breeding largely depends on inexpensive genotyping arrays. In this study, we aimed to develop an ovine high-resolution multiple-single-nucleotide polymorphism (SNP) capture array, based on genotyping by target sequencing (GBTS) system with capture-in-solution (liquid chip) technology. All the markers were from 40K captured regions, including genes located within selective sweep regions, breed-specific regions, quantitative trait loci (QTL), and the potential functional SNPs on the sheep genome. The results showed that a total of 210K high-quality SNPs were identified in the 40K regions, indicating a high average capture ratio (99.7%) for the target genomic regions. Using genotyped data (n = 317) from liquid chip technology, we further performed genome-wide association studies (GWAS) to detect the genetic loci affecting sheep hair types and teat number. A single significant association signal for hair types was identified on 6.7-7.1 Mb of chromosome 25. The IRF2BP2 gene (chr25: 7,067,974-7,071,785), which is located within this genomic region, has been previously known to be involved in hair/wool traits in sheep. The results further showed a new candidate region around 26.4 Mb of chromosome 13, between the ARHGAP21 and KIAA1217 genes, that was significantly related to teat number in sheep. The haplotype patterns of this region also showed differences in animals with 2, 3, or 4 teats. Advances in using the high-accuracy and low-cost liquid chip are expected to accelerate sheep genomic and breeding studies in the coming years.

A Method of Detecting Candidate Regions and Flames Based on Deep Learning Using Color-Based Pre-Processing

Recently, object detection methods using deep learning have made significant progress in terms of accuracy and speed. However, the requirements of a system to provide real-time detection are somewhat high, and current methods are still insufficient to accurately detect important factors directly related to life and safety, such as fires. Therefore, this study attempted to improve the detection rate by supplementing the existing research to reduce the false detection rate of flame detection in fire and to reduce the number of candidate regions extracted in advance. To this end, pre-processing based on the HSV and YCbCr color models was performed to filter the flame area simply and strongly, and a selective search was used to detect a valid candidate region for the filtered image. In addition, for the detected candidate region, a deep learning-based convolutional neural network (CNN) was used to infer whether the object was a flame. As a result, it was found that the flame-detection accuracy of the model proposed in this study was 7% higher than that of the other models presented for comparison, and the recall rate was increased by 6%.

Dissecting the loci underlying maturation timing in Atlantic salmon using haplotype and multi-SNP based association methods

Characterizing the role of different mutational effect sizes in the evolution of fitness-related traits has been a major goal in evolutionary biology for a century. Such characterization in a diversity of systems, both model and non-model, will help to understand the genetic processes underlying fitness variation. However, well-characterized genetic architectures of such traits in wild populations remain uncommon. In this study, we used haplotype-based and multi-SNP Bayesian association methods with sequencing data for 313 individuals from wild populations to test the mutational composition of known candidate regions for sea age at maturation in Atlantic salmon (Salmo salar). We detected an association at five loci out of 116 candidates previously identified in an aquaculture strain with maturation timing in wild Atlantic salmon. We found that at four of these five loci, variation explained by the locus was predominantly driven by a single SNP suggesting the genetic architecture of this trait includes multiple loci with simple, non-clustered alleles and a locus with potentially more complex alleles. This highlights the diversity of genetic architectures that can exist for fitness-related traits. Furthermore, this study provides a useful multi-SNP framework for future work using sequencing data to characterize genetic variation underlying phenotypes in wild populations.

A Semantic Segmentation Method for Three-Dimensional Image of Internet Data Center Machine Room Constructed by Intelligent Power Internet of Things Based on Fabric Architecture

In this paper, a three-dimensional image semantic segmentation method is studied. First, the data such as the equipment image in the computer room is obtained, and the data encryption and decentralized storage are completed through the consensus algorithm in the blockchain technology; Then transfer the data to the platform layer to realize the unified access, aggregation and distribution of the image data of the equipment in the machine room; Finally, the R-CNN method is used to introduce the FPN feature pyramid network to characterize the multi-scale image features of the equipment in the machine room. The candidate region generation network is used to obtain the region of interest of the equipment image, and the semantic segmentation of the three-dimensional image is realized through the classifier. The experimental results show that this method can accurately identify the equipment in the computer room and has good image data transmission performance.

TD-Net: Trans-Deformer network for automatic pancreas segmentation

Accurate and efficient pancreas segmentation is the basis for subsequent diagnosis and qualitative treatment of pancreatic cancer. Segmenting the pancreas from abdominal CT images is a challenging task because the morphology of the pancreas varies greatly among different individuals and may be affected by problems such as the unbalanced category and blurred boundaries. This paper proposes a two-stage Trans-Deformer network to solve these problems of pancreas segmentation. To be specific, we first use 2D Unet for coarse segmentation to generate candidate regions of the pancreas. In the fine segmentation stage, we propose to integrate deformable convolution into Vision Transformer (VIT) for solving the deformation problem of the pancreas. For the problem of blurred boundaries caused by low contrast in the pancreas, a multi-input module based on wavelet decomposition is proposed to make our network pay more attention to high-frequency texture information. In addition, we propose using the Scale Inter-active Fusion (SIF) module to merge local features and global features. Our method was evaluated on the public NIH dataset including 82 abdominal contrast-enhanced CT volumes and the public MSD dataset including 281 abdominal contrast-enhanced CT volumes via fourfold cross-validation. We have achieved the average Dice Similarity Coefficient (DSC) values of 89.89 ± 1.82 % on the NIH dataset, and 91.22 ± 1.37 % on the MSD dataset, outperforming other exiting state-of-the-art pancreas segmentation methods.

Lost pigs of Angola: Whole genome sequencing reveals unique regions of selection with emphasis on metabolism and feed efficiency.

Angola, in the western coast of Africa, has been through dramatic social events that have led to the near-disappearance of native swine populations, and the recent introduction of European exotic breeds has also contributed to the erosion of this native swine repertoire. In an effort to investigate the genetic basis of native pigs in Angola (ANG) we have generated whole genomes from animals of a remote local pig population in Huambo province, which we have compared with 78 genomes of European and Asian pig breeds as well as European and Asian wild boars that are currently in public domain. Analyses of population structure showed that ANG pigs grouped within the European cluster and were clearly separated from Asian pig breeds. Pairwise F ST ranged from 0.14 to 0.26, ANG pigs display lower levels of genetic differentiation towards European breeds. Finally, we have identified candidate regions for selection using a complementary approach based on various methods. All results suggest that selection towards feed efficiency and metabolism has occurred. Moreover, all analysis identified CDKAL1 gene, which is related with insulin and cholesterol metabolism, as a candidate gene overlapping signatures of selection unique to ANG pigs. This study presents the first assessment of the genetic relationship between ANG pigs and other world breeds and uncovers selection signatures that may indicate adaptation features unique to this important genetic resource.

Multiscale contrast enhancement method for small infrared target detection

Infrared search and track (IRST) systems have been applied extensively in the civil and military fields. However, fast, effective, and robust dim and small infrared target detection remains a challenging research topic. Dim and small targets are easily obscured in different backgrounds or noises, which results in a high false alarm rate. A novel strategy is proposed for small infrared target detection to achieve a high true positive rate and robust detection. First, the prominent edge structures of the original image are obtained by constructing a structure tensor. Subsequently, a Gaussian bandpass filter with multiscale windows is used to obtain an infrared image that contains complex edges and small targets. An infrared image with an enhanced signal-to-clutter ratio is achieved by combining with the two methods, whereby the noise can be effectively reduced. Thereafter, candidate regions that contain the target and a smaller number of edges with higher brightness than that of the target are obtained using relaxed threshold segmentation. Finally, a Gaussian gradient contrast method is proposed in which the highlighted edges are removed so that only small infrared targets remain. Extensive experimental results demonstrate that the proposed method can achieve a significant optimization of the true positive and false alarm rates, as well as improved efficiency and robustness.

Detecting Maritime Infrared Targets in Harsh Environment by Improved Visual Attention Model Preselector and Anti-Jitter Spatiotemporal Filter Discriminator

Without any prior knowledge, it has always been a serious challenge to accurately detect infrared targets under a maritime harsh environment (MHE). To solve this problem, the main contribution of this paper is to use the improved visual attention model (VAM) preselector and anti-jitter spatiotemporal filter (ASF) discriminator to detect infrared targets in the MHE. The proposed method consists of image preprocessing, the single frame detection of suspected targets and a multi-frame judgment of real targets. First, in the process of single-frame image processing, a combination of the Gaussian difference filtering and local minimum filtering is applied to overcome the uneven background brightness distribution and improve the saliency of the target. Second, an intensity standard deviation method is designed to determine the unevenness of the background. According to the difference in background smoothness, an appropriate “center-surround difference” operation is selected to suppress sea wave interference, and the single frame suspected target candidate region is obtained. Third, in order to “align” the multi-frame image sequence, a method for correcting the position of the inter-frame jitter is proposed. The inter-frame jitter is measured and compensated by the inter-frame block matching results. Finally, according to the three assumptions of multi-frame spatiotemporal filtering, single-frame false targets are filtered out and combined with the OTSU method to segment the real target area. To evaluate the performance index of the proposed method by comparing it with the four other state-of-the-art methods for dealing with the MHE. The experimental results show that the algorithm achieves the maximum detection rate (DR) on the premise of being far lower than the false alarm rate (FAR) of the comparison method. The final experimental results also confirmed that the proposed algorithm is more suitable for infrared target detection in diverse MHEs.

No evidence for intervention-associated DNA methylation changes in monocytes of patients with posttraumatic stress disorder

DNA methylation patterns can be responsive to environmental influences. This observation has sparked interest in the potential for psychological interventions to influence epigenetic processes. Recent studies have observed correlations between DNA methylation changes and therapy outcome. However, most did not control for changes in cell composition. This study had two aims: first, we sought to replicate therapy-associated changes in DNA methylation of commonly assessed candidate genes in isolated monocytes from 60 female patients with post-traumatic stress disorder (PTSD). Our second, exploratory goal was to identify novel genomic regions with substantial pre-to-post intervention DNA methylation changes by performing whole-genome bisulfite sequencing (WGBS) in two patients with PTSD. Equivalence testing and Bayesian analyses provided evidence against physiologically meaningful intervention-associated DNA methylation changes in monocytes of PTSD patients in commonly investigated target genes (NR3C1, FKBP5, SLC6A4, OXTR). Furthermore, WGBS yielded only a limited set of candidate regions with suggestive evidence of differential DNA methylation pre- to post-therapy. These differential DNA methylation patterns did not prove replicable when investigated in the entire cohort. We conclude that there is no evidence for major, recurrent intervention-associated DNA methylation changes in the investigated genes in monocytes of patients with PTSD.

Integrated genetic analysis of leaf blast resistance in upland rice: QTL mapping, bulked segregant analysis and transcriptome sequencing.

Elite upland rice cultivars have the advantages of less water requirement along with high yield but are usually susceptible to various diseases. Rice blast caused by Magnaporthe oryzae is the most devastating disease in rice. Identification of new sources of resistance and the introgression of major resistance genes into elite cultivars are required for sustainable rice production. In this study, an upland rice genotype UR0803 was considered an emerging source of blast resistance. An F2 mapping population was developed from a cross between UR0803 and a local susceptible cultivar Lijiang Xintuan Heigu. The individuals from the F2 population were evaluated for leaf blast resistance in three trials 7 days after inoculation. Bulked segregant analysis (BSA) by high-throughput sequencing and SNP-index algorithm was performed to map the candidate region related to disease resistance trait. A major quantitative trait locus (QTL) for leaf blast resistance was identified on chromosome 11 in an interval of 1.61-Mb genomic region. The candidate region was further shortened to a 108.9-kb genomic region by genotyping the 955 individuals with 14 SNP markers. Transcriptome analysis was further performed between the resistant and susceptible parents, yielding a total of 5044 differentially expressed genes (DEGs). There were four DEGs in the candidate QTL region, of which, two (Os11g0700900 and Os11g0704000) were upregulated and the remaining (Os11g0702400 and Os11g0703600) were downregulated in the susceptible parent after inoculation. These novel candidate genes were functionally annotated to catalytic response against disease stimulus in cellular membranes. The results were further validated by a quantitative real-time PCR analysis. The fine-mapping of a novel QTL for blast resistance by integrative BSA mapping and transcriptome sequencing enhanced the genetic understanding of the mechanism of blast resistance in upland rice. The most suitable genotypes with resistance alleles would be useful genetic resources in rice blast resistance breeding.

Identification of adenylate cyclase 2 methylation in bladder cancer with implications for prognosis and immunosuppressive microenvironment.

BackgroundThe incidence and mortality of bladder cancer (BCa) are increasing, while the existing diagnostic methods have limitations. Therefore, for early detection and response prediction, it is crucial to improve the prognosis and treatment strategies. However, with existing diagnostic methods, detecting BCa in the early stage is challenging. Hence, novel biomarkers are urgently needed to improve early diagnosis and treatment efficiency.MethodsThe gene expression profile and gene methylation profile dataset were downloaded from the Gene Expression Omnibus (GEO) database. Differentially expressed genes (DEGs), differentially methylated genes (DMGs), and methylation-regulated differentially expressed genes (MeDEGs) were gradually identified. A cancer genome map was obtained using online gene expression profile interaction analysis, and survival implications were produced using Kaplan-Meier survival analysis. GSEA was employed to predict the marker pathways where DEGs were significantly involved. The study used bisulfite PCR amplification combined with bisulfite amplicon sequencing (BSAS) to screen for methylation analysis of multiple candidate regions of the adenylate cyclase 2 (ADCY2) based on the sequence design of specific gene regions and CpG islands.ResultsIn this study, DEGs and DMGs with significantly up- or down-regulated expression were selected. The intersection method was used to screen the MeDEGs. The interaction network group in STRING was then visualized using Cytoscape, and the PPI network was constructed to identify the key genes. The key genes were then analyzed using functional enrichment. To compare the relationship between key genes and the prognosis of BCa patients, we further investigated ADCY2 and found that ADCY2 can be a potential clinical biomarker in BCa prognosis and immunotherapy response prediction. In human BCa 5637 and MGH1 cells, we developed and verified the effectiveness of ADCY2 primers using BSAS technology. The findings revealed that the expression of ADCY2 is highly regulated by the methylation of the promoter regions.ConclusionThis study revealed that increased expression of ADCY2 was significantly correlated with increased tumor heterogeneity, predicting worse survival and immunotherapy response in BCa patients.

RFfiller: a robust and fast statistical algorithm for gap filling in draft genomes.

Numerous published genomes contain gaps or unknown sequences. Gap filling is a critical final step in de novo genome assembly, particularly for large genomes. While certain computational approaches partially address the problem, others have shortcomings regarding the draft genome’s dependability and correctness (high rates of mis-assembly at gap-closing sites and high error rates). While it is well established that genomic repeats result in gaps, many sequence reads originating from repeat-related gaps are typically missed by existing approaches. A fast and reliable statistical algorithm for closing gaps in a draft genome is presented in this paper. It utilizes the alignment statistics between scaffolds, contigs, and paired-end reads to generate a Markov chain that appropriately assigns contigs or long reads to scaffold gap regions (only corrects candidate regions), resulting in accurate and efficient gap closure. To reconstruct the missing component between the two ends of the same insert, the RFfiller meticulously searches for valid overlaps (in repeat regions) and generates transition tables for similar reads, allowing it to make a statistical guess at the missing sequence. Finally, in our experiments, we show that the RFfiller’s gap-closing accuracy is better than that of other publicly available tools when sequence data from various organisms are used. Assembly benchmarks were used to validate RFfiller. Our findings show that RFfiller efficiently fills gaps and that it is especially effective when the gap length is longer. We also show that the RFfiller outperforms other gap closing tools currently on the market.

Genome-wide association analysis of the primary feather growth traits of duck: identification of potential Loci for growth regulation

The feather is an important epidermal appendage, plays an important role in the life activities of avian specie, and has important economic value. Revealing the molecular regulation mechanism of feather growth has a significant meaning in studying adaptive evolution, physiology, and mating of avian species and also provides a theoretical reference for poultry breeding. In this study, the genome-wide association analysis (GWAS) of 358 ducks was based on primary feather length phenotypic data (28-60 d), length growth rates (LGRs), and maturity scores (60 d) to explore the genetic basis affecting feather growth and maturation. The results showed that, among the primary feather 1 to 5 in ducks, the mean LGR of primary feather 2 was the fastest, with the longest length. The primary feathers in males grew and matured slightly faster than in females. The mean maturity scores of primary feather 10∼7 were higher than primary feather 1 to 3 in ducks. GWAS further showed 116 SNPs associated with feather length traits. In addition, 2 candidate regions (Chr1: 127,407,230-127,524,879 bp and Chr21: 182,061,707-183,616,298 bp) were associated with LGR, which contain total 13 candidate genes (The extremely significant SNPs were mainly located in 2 genes: Chr1: REPS2 and Chr21: PTPRT). Four candidate regions (Chr1: 29,113,036-28,675,018 bp, Chr2: 18,253,612-149,111,290 bp, Chr15: 6,489,774 to 12,138,221 bp and Chr21: 6,578,021-8,472,904 bp) were associated with feather maturity, which contain total 24 candidate genes (The extremely significant SNPs were mainly located in 4 genes: Chr1: IMMP2L, DOCK4 and DDX10, Chr2: LDLRAD4). In conclusion, sex factors influence feather growth and maturity, and the genetic basis of the growth /maturity trait between different feathers is similar. REPS2, PTPRT genes, and IMMP2L, DOCK4, DDX10, and LDLRAD4 are important candidate genes that influence feather growth and maturity, respectively.

Candidate genes for litter size in Xinjiang sheep identified by Specific Locus Amplified Fragment (SLAF) sequencing

The aim of this study was to investigate the selection signatures at a genome-wide level in ‘Pishan’ sheep using Specific Locus Amplified Fragment (SLAF)-seq. Blood samples from 126 ewes were sequenced using SLAF tags, and the ovarian tissues from 8 ewes (Bashbay sheep, a single litter size group (SG group); ‘Pishan’ sheep, double litter size group (DG group)) were collected to detect expression levels by quantitative real-time reverse transcription polymerase chain reaction (qRT-PCR). Selection signature analysis was performed using global fixation index (Fst) and nucleotide diversity (π) ratio. A total of 1,192,168 high-quality SLAFs were identified. Notably, 2380 candidate regions under selection using two approaches were identified. A total of 2069 genes were identified, which were involved in dopaminergic synapses, thyroid hormone synthesis, ovarian steroidogenesis and thyroid hormone signalling pathways. Furthermore, Growth Differentiation Factor 9 (GDF9), Period Circadian Regulator 2 (PER2), Thyroid Stimulating Hormone Receptor (TSHR), and Nuclear Receptor Coactivator 1 (NCOA1) reside within these regions and pathways. The expression levels of GDF9 and PER2 genes in sheep tissue of the DG group were significantly higher than those in the SG group. These genes are interesting candidates for litter size and provide a starting point for further identification of conservation strategies for ‘Pishan’ sheep.

Imputation to whole-genome sequence and its use in genome-wide association studies for pork colour traits in crossbred and purebred pigs.

Imputed whole-genome sequence (WGS) has been proposed to improve genome-wide association studies (GWAS), since all causative mutations responsible for phenotypic variation are expected to be present in the data. This approach was applied on a large number of purebred (PB) and crossbred (CB) pigs for 18 pork color traits to evaluate the impact of using imputed WGS relative to medium-density marker panels. The traits included Minolta A*, B*, and L* for fat (FCOL), quadriceps femoris muscle (QFCOL), thawed loin muscle (TMCOL), fresh ham gluteus medius (GMCOL), ham iliopsoas muscle (ICOL), and longissimus dorsi muscle on the fresh loin (FMCOL). Sequence variants were imputed from a medium-density marker panel (61K for CBs and 50K for PBs) in all genotyped pigs using BeagleV5.0. We obtained high imputation accuracy (average of 0.97 for PBs and 0.91 for CBs). GWAS were conducted for three datasets: 954 CBs and 891 PBs, and the combined CBs and PBs. For most traits, no significant associations were detected, regardless of panel density or population type. However, quantitative trait loci (QTL) regions were only found for a few traits including TMCOL Minolta A* and GMCOL Minolta B* (CBs), FMCOL Minolta B*, FMCOL Minolta L*, and ICOL Minolta B* (PBs) and FMCOL Minolta A*, FMCOL Minolta B*, GMCOL Minolta B*, and ICOL Minolta B* (Combined dataset). More QTL regions were identified with WGS (n = 58) relative to medium-density marker panels (n = 22). Most of the QTL were linked to previously reported QTLs or candidate genes that have been previously reported to be associated with meat quality, pH and pork color; e.g., VIL1, PRKAG3, TTLL4, and SLC11A1, USP37. CTDSP1 gene on SSC15 has not been previously associated with meat color traits in pigs. The findings suggest any added value of WGS was only for detecting novel QTL regions when the sample size is sufficiently large as with the Combined dataset in this study. The percentage of phenotypic variance explained by the most significant SNPs also increased with WGS compared with medium-density panels. The results provide additional insights into identification of a number of candidate regions and genes for pork color traits in different pig populations.

A Novel Method of Small Object Detection in UAV Remote Sensing Images Based on Feature Alignment of Candidate Regions

To solve the problem of low detection accuracy of small objects in UAV optical remote sensing images due to low contrast, dense distribution, and weak features, this paper proposes a small object detection method based on feature alignment of candidate regions is proposed for remote sensing images. Firstly, AFA-FPN (Attention-based Feature Alignment FPN) defines the corresponding relationship between feature mappings, solves the misregistration of features between adjacent levels, and improves the recognition ability of small objects by aligning and fusing shallow spatial features and deep semantic features. Secondly, the PHDA (Polarization Hybrid Domain Attention) module captures local areas containing small object features through parallel channel domain attention and spatial domain attention. It assigns a larger weight to these areas to alleviate the interference of background noise. Then, the rotation branch uses RROI to rotate the horizontal frame obtained by RPN, which avoids missing detection of small objects with dense distribution and arbitrary direction. Next, the rotation branch uses RROI to rotate the horizontal box obtained by RPN. It solves the problem of missing detection of small objects with dense distribution and arbitrary direction and prevents feature mismatch between the object and candidate regions. Finally, the loss function is improved to better reflect the difference between the predicted value and the ground truth. Experiments are conducted on a self-made dataset. The experimental results show that the mAP of the proposed method reaches 82.04% and the detection speed reaches 24.3 FPS, which is significantly higher than that of the state-of-the-art methods. Meanwhile, the ablation experiment verifies the rationality of each module.

Unsupervised contrastive learning based transformer for lung nodule detection

Objective. Early detection of lung nodules with computed tomography (CT) is critical for the longer survival of lung cancer patients and better quality of life. Computer-aided detection/diagnosis (CAD) is proven valuable as a second or concurrent reader in this context. However, accurate detection of lung nodules remains a challenge for such CAD systems and even radiologists due to not only the variability in size, location, and appearance of lung nodules but also the complexity of lung structures. This leads to a high false-positive rate with CAD, compromising its clinical efficacy. Approach. Motivated by recent computer vision techniques, here we present a self-supervised region-based 3D transformer model to identify lung nodules among a set of candidate regions. Specifically, a 3D vision transformer is developed that divides a CT volume into a sequence of non-overlap cubes, extracts embedding features from each cube with an embedding layer, and analyzes all embedding features with a self-attention mechanism for the prediction. To effectively train the transformer model on a relatively small dataset, the region-based contrastive learning method is used to boost the performance by pre-training the 3D transformer with public CT images. Results. Our experiments show that the proposed method can significantly improve the performance of lung nodule screening in comparison with the commonly used 3D convolutional neural networks. Significance. This study demonstrates a promising direction to improve the performance of current CAD systems for lung nodule detection.

Hybridization, missing wild ancestors and the domestication of cultivated diploid bananas.

Hybridization and introgressions are important evolutionary forces in plants. They contribute to the domestication of many species, including understudied clonal crops. Here, we examine their role in the domestication of a clonal crop of outmost importance, banana (Musa ssp.). We used genome-wide SNPs generated for 154 diploid banana cultivars and 68 samples of the wild M. acuminata to estimate and geo-localize the contribution of the different subspecies of M. acuminata to cultivated banana. We further investigated the wild to domesticate transition in New Guinea, an important domestication center. We found high levels of admixture in many cultivars and confirmed the existence of unknown wild ancestors with unequal contributions to cultivated diploid. In New Guinea, cultivated accessions exhibited higher diversity than their direct wild ancestor, the latter recovering from a bottleneck. Introgressions, balancing selection and positive selection were identified as important mechanisms for banana domestication. Our results shed new lights on the radiation of M. acuminata subspecies and on how they shaped banana domestication. They point candidate regions of origin for two unknown ancestors and suggest another contributor in New Guinea. This work feed research on the evolution of clonal crops and has direct implications for conservation, collection, and breeding.