Search

With the continuous progress of remote sensing image object detection tasks in recent years, researchers in this field have gradually shifted the focus of their research from horizontal object detection to the study of object detection in arbitrary directions. It is worth noting that some properties are different from horizontal object detection during oriented object detection that researchers have yet to notice much. This article presents the design of a straightforward and efficient arbitrary-oriented detection system, leveraging the inherent properties of the orientation task, including the rotation angle and box aspect ratio. In the detection of low aspect ratio objects, the angle is of little importance to the orientation bounding box, and it is even difficult to define the angle information in extreme categories. Conversely, in the detection of objects with high aspect ratios, the angle information plays a crucial role and can have a decisive impact on the quality of the detection results. By exploiting the aspect ratio of different targets, this letter proposes a ratio-balanced angle loss that allows the model to make a better trade-off between low-aspect ratio objects and high-aspect ratio objects. The rotation angle of each oriented object, which we naturally embed into a two-dimensional Euclidean space for regression, thus avoids an overly redundant design and preserving the topological properties of the circular space. The performance of the UCAS-AOD, HRSC2016, and DLR-3K datasets show that the proposed model in this paper achieves a leading level in terms of both accuracy and speed. [ABSTRACT FROM AUTHOR]

The prediction of human-virus protein-protein interactions (human-virus PPIs) is significant for exploring the mechanisms of viral infection, making their prediction a necessary and practically valuable research topic. Since conventional methods for the determination of human-virus protein-protein interactions are very complex and expensive, the construction of models plays a crucial role. In this paper, we construct an interpretable model, ECA-PHV, to predict human-virus protein-protein interactions based on an effective channel attention mechanism. First, we utilize five coding modalities, namely AAC, DDE, MMI, CT, and GTPC, to extract the hidden biological information in protein sequences. Individual feature weights are then learned by using a differential evolutionary algorithm that employs weighted combinations to adequately represent various protein sequence information. Next, irrelevant features in multi-information fusion are removed by Group Lasso. Finally, the prediction model is constructed by combining effective channel attention, BiGRU, and 1D-CNN. Compared with existing models, the interpretability framework ECA-PHV proposed in this paper has competitive and stable predictive performance. This shows that our model can efficiently focus on important information about protein sequences. In conclusion, this study accelerates the exploration of human-virus protein-protein interactions and provides some insights of practical value for probing human-virus relationships. • A novel method (ECA-PHV) to predict human-virus protein-protein interactions (human-virus PPIs). • The AAC, DDE, CT, MMI, and GTPC methods are fused to extract protein sequence feature information. • We first constructed an interpretable deep learning framework that combines BiGRU, 1D-CNN, and efficient channel attention to predict human-virus protein-protein interactions. • ECA-PHV increases the prediction performance compared with other methods. [ABSTRACT FROM AUTHOR]

Motivation Phosphorylation is one of the most studied post-translational modifications, which plays a pivotal role in various cellular processes. Recently, deep learning methods have achieved great success in prediction of phosphorylation sites, but most of them are based on convolutional neural network that may not capture enough information about long-range dependencies between residues in a protein sequence. In addition, existing deep learning methods only make use of sequence information for predicting phosphorylation sites, and it is highly desirable to develop a deep learning architecture that can combine heterogeneous sequence and protein–protein interaction (PPI) information for more accurate phosphorylation site prediction. Results We present a novel integrated deep neural network named PhosIDN, for phosphorylation site prediction by extracting and combining sequence and PPI information. In PhosIDN, a sequence feature encoding sub-network is proposed to capture not only local patterns but also long-range dependencies from protein sequences. Meanwhile, useful PPI features are also extracted in PhosIDN by a PPI feature encoding sub-network adopting a multi-layer deep neural network. Moreover, to effectively combine sequence and PPI information, a heterogeneous feature combination sub-network is introduced to fully exploit the complex associations between sequence and PPI features, and their combined features are used for final prediction. Comprehensive experiment results demonstrate that the proposed PhosIDN significantly improves the prediction performance of phosphorylation sites and compares favorably with existing general and kinase-specific phosphorylation site prediction methods. Availability and implementation PhosIDN is freely available at https://github.com/ustchangyuanyang/PhosIDN. Supplementary information Supplementary data are available at Bioinformatics online. [ABSTRACT FROM AUTHOR]

Motivation Phosphorylation is the most studied post-translational modification, which is crucial for multiple biological processes. Recently, many efforts have been taken to develop computational predictors for phosphorylation site prediction, but most of them are based on feature selection and discriminative classification. Thus, it is useful to develop a novel and highly accurate predictor that can unveil intricate patterns automatically for protein phosphorylation sites. Results In this study we present DeepPhos, a novel deep learning architecture for prediction of protein phosphorylation. Unlike multi-layer convolutional neural networks, DeepPhos consists of densely connected convolutional neuron network blocks which can capture multiple representations of sequences to make final phosphorylation prediction by intra block concatenation layers and inter block concatenation layers. DeepPhos can also be used for kinase-specific prediction varying from group, family, subfamily and individual kinase level. The experimental results demonstrated that DeepPhos outperforms competitive predictors in general and kinase-specific phosphorylation site prediction. Availability and implementation The source code of DeepPhos is publicly deposited at https://github.com/USTCHIlab/DeepPhos. Supplementary information Supplementary data are available at Bioinformatics online. [ABSTRACT FROM AUTHOR]

Malonylation is a new protein post-translational modification and regulates a variety of cellular physiological processes. However, it is costly and time-consuming to identify malonylation sites through traditional experiments. Therefore, the prediction of malonylation sites by computational methods plays an important role in experimental design. In this paper, a new prediction model of malonylation sites, Malsite-Deep, is proposed. First, the seven feature extraction methods are used to extract feature information of protein sequences. Then, the under-sampling NearMiss-2 method is applied to handle imbalance data, and the update gate and reset gate of gated recurrent units (GRU) are used to select the optimal feature subset. Finally, the data from GRU layer is input into deep neural networks (DNN) to predict the malonylation sites, and the model performance is evaluated by 10-fold cross-validation and independent test sets. The 10-fold cross-validation shows that the AUC value on the training dataset reaches 0.99. The AUC values on the four independent test datasets all reach above 0.95. Results suggest that Malsite-Deep presented here facilitates the identification of protein malonylation sites. • A novel method (Malsite-Deep) is used to predict malonylation sites. • The BE, PWAA, BPB, EAAC, DC, EBGW and BLOSUM62 methods are fused to extract protein sequence features information. • NearMiss-2 algorithm is used to process the category imbalance data for the first time. • We firstly build a new deep learning framework which combines the GRU and DNN to predict the malonylation sites. [ABSTRACT FROM AUTHOR]

With the rapid development of deep learning, automatic lesion detection is used widely in clinical screening. To solve the problem that existing deep learning-based cervical precancerous lesion detection algorithms cannot meet high classification accuracy and fast running speed at the same time, a ShuffleNet-based cervical precancerous lesion classification method is proposed. By adding channel attention to the ShuffleNet, the network performance is improved. In this study, the image dataset is classified into five categories: normal, cervical cancer, LSIL (CIN1), HSIL (CIN2/CIN3), and cervical neoplasm. The colposcopy images are expanded to solve the problems of the lack of colposcopy images and the uneven distribution of images from each category. For the test dataset, the accuracy of the proposed CNN models is 81.23% and 81.38%. Our classifier achieved an AUC score of 0.99. The experimental results show that the colposcopy image classification network based on artificial intelligence has good performance in classification accuracy and model size, and it has high clinical applicability. [ABSTRACT FROM AUTHOR]

Lysine malonylation is one of the important post-translational modification (PTM) of proteins. Malonylated proteins can affect various cell functions of eukaryotes and prokaryotes, which play an important role in metabolic pathways, mitochondrial functions, fatty acid oxidation and other life activities. However, accurate identification of the malonylation sites is the key to further understand the molecular mechanism of malonylation. Currently, experimental identification is still a challenging task, which usually requires a large amount of laboratory work and considerable cost. Regarding this situation, there is an urgent need to establish useful calculation methods and develop effective predictors. We propose a new deep learning network model called DeepMal. Firstly, features are extracted by enhanced amino acid composition (EAAC), enhanced grouped amino acid composition (EGAAC), dipeptide deviation from the expected mean (DDE), K nearest neighbors (KNN) and BLOSUM62 matrix. Secondly, the linear convolutional neural network is used to extract the specific features of malonylation sites, select the relevant features and reduce the feature dimension through maximum pooling. Finally, malonylation sites and non-malonylation sites are classified by a multilayer neural network. And an independent dataset is used to assess the predictive ability of the model DeepMal. On the independent datasets Escherichia coli (E. coli), Homo sapiens (H. sapiens), Mus musculus (M. musculus), the AUC values are 0.974, 0.956 and 0.944, and the accuracies are 96.5%, 95.5% and 94.5%, respectively. Compared with other prediction models, the prediction accuracy is increased by 9.5%–18.5%, and this indicates the effectiveness of the prediction model DeepMal. Using deep learning network can enhance the robustness of DeepMal model for predicting malonylation sites. The source codes and all datasets are publicly available at https://github.com/QUST-AIBBDRC/DeepMal/. • A novel method (DeepMal) to predict the lysine malonylation sites of proteins. • Protein sequence information is extracted from sequence features, physicochemical properties, and evolutionary features. • Convolutional neural networks can learn to extract significant features associated with malonylation sites. • Deep neural networks can improve the accurate prediction of protein lysine malonylation sites of proteins. • The proposed method increases the prediction performance on independent test datasets compared with other methods. [ABSTRACT FROM AUTHOR]

Carbonylation is an irreversible post-translational modification of proteins and regulates various cellular physiological processes. Due to the limitations of experimental methods, it is necessary to predict carbonylation sites by computational methods. In this paper, a new prediction model of carbonylation, Precar_Deep, is proposed. First, six feature extraction methods are used to obtain the original feature space from the protein sequences. Then, the Group LASSO method is used to remove redundant information and the oversampling Borderline-SMOTE method is employed to balance the data to obtain a new feature space. Finally, the processed data is input into the deep learning framework constructed in this paper to predict the carbonylation sites, and the performance of the model is evaluated by using 10-fold cross-validation and independent test datasets. The AUC values of the four datasets are all more than 90%. The experimental results show that PreCar_Deep is superior to other existing models and is helpful to identify protein carbonylation sites. The source code and all datasets are available at https://github.com/QUST-SHULI/PreCar_Deep/. • A novel method (PreCar_Deep) is used to predict carbonylation sites. • The AAC, DC, AAindex, EGAAC, CT and KNN methods are fused to extract protein sequence features information. • Borderline-SMOTE algorithm is utilized to process the category imbalance data for the first time. • We firstly build a new deep learning framework which combines the CNN and BiLSTM to predict the carbonylation sites. [ABSTRACT FROM AUTHOR]

This paper designs a method that can generate anchors of various shapes for the object detection framework. This method has the characteristics of novelty and flexibility. Different from the previous anchors generated by a pre-defined manner, our anchors are generated dynamically by an anchor generator. Specially, the anchor generator is not fixed but learned from the hand-designed anchors, which means that our anchor generator is able to work well in various scenes. In the inference time, the weights of anchor generator are estimated by a simple network where the input is some hand-designed anchor. In addition, in order to make the difference between the number of positive and negative samples smaller, we use an adaptive IOU threshold related to the object size to solve this problem. At the same time, we proved that our proposed method is effective and conducted a lot of experiments on the COCO dataset. Experimental results show that after replacing the anchor generation method in the previous object detectors (such as SSD, mask RCNN, and Retinanet) with our proposed method, the detection performance of the model has been greatly improved compared to before the replacement, which proves our method is effective. [ABSTRACT FROM AUTHOR]

• Species-specific ubiquitination site prediction method superior to general method. • Studies species-specific ubiquitination site prediction by using transfer learning technique. • The proposed DeepTL-Ubi method is very excellent in species-specific ubiquitination site prediction work. Ubiquitination is one of the most important post-translational modifications which involves in many biological processes. Because mass spectrometry-based ubiquitination site identification methods are costly and time consuming, computational approaches provide alternative ways to the determination of ubiquitination sites. Although machine learning based methods can effectively predict ubiquitination sites, most of them rely on feature engineering, which may lead to bias or incomplete feature. Recently, deep learning has achieved great success in prediction of post-translational modification sites. However, deep learning method has not been explored in the prediction of species-specific ubiquitination sites. In this paper, we propose a novel transfer deep learning method, named DeepTL-Ubi, for predicting ubiquitination sites of multiple species. DeepTL-Ubi enhances the performance of species-specific ubiquitination site prediction by transferring common knowledge from the large amount of human data to other species, which effectively solves the problem of insufficient training data for other species. Besides, we train and test our model by collecting ubiquitination sites for multiple species from several sources. Experiment results show that our transfer learning technique can effectively improve the predictive performance of species with small sample size, and DeepTL-Ubi is superior to existing tools in many species. The source code and training data of DeepTL-Ubi are publicly deposited at https://github.com/USTC-HIlab/DeepTL-Ubi. [ABSTRACT FROM AUTHOR]

Background: Accurately segment the tumor region of MRI images is important for brain tumor diagnosis and radiotherapy planning. At present, manual segmentation is wildly adopted in clinical and there is a strong need for an automatic and objective system to alleviate the workload of radiologists.Methods: We propose a parallel multi-scale feature fusing architecture to generate rich feature representation for accurate brain tumor segmentation. It comprises two parts: (1) Feature Extraction Network (FEN) for brain tumor feature extraction at different levels and (2) Multi-scale Feature Fusing Network (MSFFN) for merge all different scale features in a parallel manner. In addition, we use two hybrid loss functions to optimize the proposed network for the class imbalance issue.Results: We validate our method on BRATS 2015, with 0.86, 0.73 and 0.61 in Dice for the three tumor regions (complete, core and enhancing), and the model parameter size is only 6.3 MB. Without any post-processing operations, our method still outperforms published state-of-the-arts methods on the segmentation results of complete tumor regions and obtains competitive performance in another two regions.Conclusions: The proposed parallel structure can effectively fuse multi-level features to generate rich feature representation for high-resolution results. Moreover, the hybrid loss functions can alleviate the class imbalance issue and guide the training process. The proposed method can be used in other medical segmentation tasks. [ABSTRACT FROM AUTHOR]

Mitosis counting of biopsies is an important biomarker for breast cancer patients, which supports disease prognostication and treatment planning. Developing a robust mitotic cell detection model is highly challenging due to its complex growth pattern and high similarities with non-mitotic cells. Most mitosis detection algorithms have poor generalizability across image domains and lack reproducibility and validation in multicenter settings. To overcome these issues, we propose a generalizable and robust mitosis detection algorithm (called FMDet), which is independently tested on multicenter breast histopathological images. To capture more refined morphological features of cells, we convert the object detection task as a semantic segmentation problem. The pixel-level annotations for mitotic nuclei are obtained by taking the intersection of the masks generated from a well-trained nuclear segmentation model and the bounding boxes provided by the MIDOG 2021 challenge. In our segmentation framework, a robust feature extractor is developed to capture the appearance variations of mitotic cells, which is constructed by integrating a channel-wise multi-scale attention mechanism into a fully convolutional network structure. Benefiting from the fact that the changes in the low-level spectrum do not affect the high-level semantic perception, we employ a Fourier-based data augmentation method to reduce domain discrepancies by exchanging the low-frequency spectrum between two domains. Our FMDet algorithm has been tested in the MIDOG 2021 challenge and ranked first place. Further, our algorithm is also externally validated on four independent datasets for mitosis detection, which exhibits state-of-the-art performance in comparison with previously published results. These results demonstrate that our algorithm has the potential to be deployed as an assistant decision support tool in clinical practice. Our code has been released at https://github.com/Xiyue-Wang/1st-in-MICCAI-MIDOG-2021-challenge. [Display omitted] • A generalizable and effective mitosis detection algorithm to fully consider the mismatch across domains. • We transform the mitosis detection task into a semantic segmentation one to achieve precise mitotic figure identification. • An attention mechanism is utilized to generate channel-wise multi-scale features, which helps the network to extract more robust features. • An unsupervised Fourier-based data augmentation is used to align the domain discrepancies. • Our method won first place in the 2021 MIDOG challenge and showed successful generalization to unseen independent multicenter datasets. [ABSTRACT FROM AUTHOR]

Automated nuclei detection is crucial prerequisites for a number of histopathology related image analysis such as cancer diagnosis. Although existing deep learning based nuclei detection methods have achieved promising results, they cannot effectively deal with domain shift problem caused by different staining procedures and organ specific nuclear morphology. To handle this problem, in this paper a novel adversarial feature alignment method is proposed for domain adaptive nuclei detection, which includes both global alignment and local attentional alignment components to transfer the knowledge from source domain to target domain. Specifically, in local attentional alignment component, by using nuclei locations as guidance we extract local features and perform adversarial alignment. Furthermore, to address the issue that these local features from nuclei regions often contain insufficient information because of the small size of nuclei, we introduce an efficient location-aware self-attention (LocSA) module to refine local features by utilizing cues from all nuclei for obtaining discriminative features to perform successful feature alignment. Extensive experimental results are provided on two adaptation scenarios and our method demonstrates favorable performance against existing domain adaptation methods, which highlights the effectiveness of the proposed method for domain adaptive nuclei detection. [ABSTRACT FROM AUTHOR]

A large-scale and well-annotated dataset is a key factor for the success of deep learning in medical image analysis. However, assembling such large annotations is very challenging, especially for histopathological images with unique characteristics (e.g., gigapixel image size, multiple cancer types, and wide staining variations). To alleviate this issue, self-supervised learning (SSL) could be a promising solution that relies only on unlabeled data to generate informative representations and generalizes well to various downstream tasks even with limited annotations. In this work, we propose a novel SSL strategy called semantically-relevant contrastive learning (SRCL), which compares relevance between instances to mine more positive pairs. Compared to the two views from an instance in traditional contrastive learning, our SRCL aligns multiple positive instances with similar visual concepts, which increases the diversity of positives and then results in more informative representations. We employ a hybrid model (CTransPath) as the backbone, which is designed by integrating a convolutional neural network (CNN) and a multi-scale Swin Transformer architecture. The CTransPath is pretrained on massively unlabeled histopathological images that could serve as a collaborative local–global feature extractor to learn universal feature representations more suitable for tasks in the histopathology image domain. The effectiveness of our SRCL-pretrained CTransPath is investigated on five types of downstream tasks (patch retrieval, patch classification, weakly-supervised whole-slide image classification, mitosis detection, and colorectal adenocarcinoma gland segmentation), covering nine public datasets. The results show that our SRCL-based visual representations not only achieve state-of-the-art performance in each dataset, but are also more robust and transferable than other SSL methods and ImageNet pretraining (both supervised and self-supervised methods). Our code and pretrained model are available at https://github.com/Xiyue-Wang/TransPath. [Display omitted] • The first Transformer-based unsupervised feature extractor for histopathological images. • A semantically-relevant contrastive learning approach is proposed. • A hybrid backbone (CTransPath) is constructed by combining CNN and Transformer. • CTransPath could be a general-purpose feature extractor for histopathological images. [ABSTRACT FROM AUTHOR]

The analysis of hand–object poses from RGB images is important for understanding and imitating human behavior and acts as a key factor in various applications. In this paper, we propose a novel coarse-to-fine two-stage framework for hand–object pose estimation, which explicitly models hand–object relations in 3D pose refinement rather than in the process of converting 2D poses to 3D poses. Specifically, in the coarse stage, 2D heatmaps of hand and object keypoints are obtained from RGB image and subsequently fed into pose regressor to derive coarse 3D poses. As for the fine stage, an interaction-aware graph convolutional network called InterGCN is introduced to perform pose refinement by fully leveraging the hand–object relations in 3D context. One major challenge in 3D pose refinement lies in the fact that relations between hand and object change dynamically according to different HOI scenarios. In response to this issue, we leverage both general and interaction-specific relation graphs to significantly enhance the capacity of the network to cover variations of HOI scenarios for successful 3D pose refinement. Extensive experiments demonstrate state-of-the-art performance of our approach on benchmark hand–object datasets. [ABSTRACT FROM AUTHOR]

It is an essential capability of indoor mobile robots to avoid various kinds of obstacles. Recently, multimodal deep reinforcement learning (DRL) methods have demonstrated great capability for learning control policies in robotics by using different sensors. However, due to the complexity of indoor environment and the heterogeneity of different sensor modalities, it remains an open challenge to obtain reliable and robust multimodal information for obstacle avoidance. In this work, we propose a novel multimodal DRL method with auxiliary task (MDRLAT) for obstacle avoidance of indoor mobile robot. In MDRLAT, a powerful bilinear fusion module is proposed to fully capture the complementary information from two-dimensional (2D) laser range findings and depth images, and the generated multimodal representation is subsequently fed into dueling double deep Q-network to output control commands for mobile robot. In addition, an auxiliary task of velocity estimation is introduced to further improve representation learning in DRL. Experimental results show that MDRLAT achieves remarkable performance in terms of average accumulated reward, convergence speed, and success rate. Moreover, experiments in both virtual and real-world testing environments further demonstrate the outstanding generalization capability of our method. [ABSTRACT FROM AUTHOR]