VISN: virus instance segmentation network for TEM images using deep attention transformer.

The identification of viruses from negative staining transmission electron microscopy (TEM) images has mainly depended on experienced experts. Recent advances in artificial intelligence have enabled virus recognition using deep learning techniques. However, most of the existing methods only perform virus classification or semantic segmentation, and few studies have addressed the challenge of virus instance segmentation in TEM images. In this paper, we focus on the instance segmentation of severe acute respiratory syndrome coronavirus type 2 (SARS-CoV-2) and other respiratory viruses and provide experts with more effective information about viruses. We propose an effective virus instance segmentation network based on the You Only Look At CoefficienTs backbone, which integrates the Swin Transformer, dense connections and the coordinate-spatial attention mechanism, to identify SARS-CoV-2, H1N1 influenza virus, respiratory syncytial virus, Herpes simplex virus-1, Human adenovirus type 5 and Vaccinia virus. We also provide a public TEM virus dataset and conduct extensive comparative experiments. Our method achieves a mean average precision score of 83.8 and F1 score of 0.920, outperforming other state-of-the-art instance segmentation algorithms. The proposed automated method provides virologists with an effective approach for recognizing and identifying SARS-CoV-2 and assisting in the diagnosis of viruses. Our dataset and code are accessible at https://github.com/xiaochiHNU/Virus-Instance-Segmentation-Transformer-Network.

Gradient Learning under Tilted Empirical Risk Minimization.

Gradient Learning (GL), aiming to estimate the gradient of target function, has attracted much attention in variable selection problems due to its mild structure requirements and wide applicability. Despite rapid progress, the majority of the existing GL works are based on the empirical risk minimization (ERM) principle, which may face the degraded performance under complex data environment, e.g., non-Gaussian noise. To alleviate this sensitiveness, we propose a new GL model with the help of the tilted ERM criterion, and establish its theoretical support from the function approximation viewpoint. Specifically, the operator approximation technique plays the crucial role in our analysis. To solve the proposed learning objective, a gradient descent method is proposed, and the convergence analysis is provided. Finally, simulated experimental results validate the effectiveness of our approach when the input variables are correlated.

Effective automated pipeline for 3D reconstruction of synapses based on deep learning.

BACKGROUND: The locations and shapes of synapses are important in reconstructing connectomes and analyzing synaptic plasticity. However, current synapse detection and segmentation methods are still not adequate for accurately acquiring the synaptic connectivity, and they cannot effectively alleviate the burden of synapse validation. RESULTS: We propose a fully automated method that relies on deep learning to realize the 3D reconstruction of synapses in electron microscopy (EM) images. The proposed method consists of three main parts: (1) training and employing the faster region convolutional neural networks (R-CNN) algorithm to detect synapses, (2) using the z-continuity of synapses to reduce false positives, and (3) combining the Dijkstra algorithm with the GrabCut algorithm to obtain the segmentation of synaptic clefts. Experimental results were validated by manual tracking, and the effectiveness of our proposed method was demonstrated. The experimental results in anisotropic and isotropic EM volumes demonstrate the effectiveness of our algorithm, and the average precision of our detection (92.8% in anisotropy, 93.5% in isotropy) and segmentation (88.6% in anisotropy, 93.0% in isotropy) suggests that our method achieves state-of-the-art results. CONCLUSIONS: Our fully automated approach contributes to the development of neuroscience, providing neurologists with a rapid approach for obtaining rich synaptic statistics.

Gradient Learning With the Mode-Induced Loss: Consistency Analysis and Applications.

Variable selection methods aim to select the key covariates related to the response variable for learning problems with high-dimensional data. Typical methods of variable selection are formulated in terms of sparse mean regression with a parametric hypothesis class, such as linear functions or additive functions. Despite rapid progress, the existing methods depend heavily on the chosen parametric function class and are incapable of handling variable selection for problems where the data noise is heavy-tailed or skewed. To circumvent these drawbacks, we propose sparse gradient learning with the mode-induced loss (SGLML) for robust model-free (MF) variable selection. The theoretical analysis is established for SGLML on the upper bound of excess risk and the consistency of variable selection, which guarantees its ability for gradient estimation from the lens of gradient risk and informative variable identification under mild conditions. Experimental analysis on the simulated and real data demonstrates the competitive performance of our method over the previous gradient learning (GL) methods.

Predicting and Visualizing Citrus Color Transformation Using a Deep Mask-Guided Generative Network.

Citrus rind color is a good indicator of fruit development, and methods to monitor and predict color transformation therefore help the decisions of crop management practices and harvest schedules. This work presents the complete workflow to predict and visualize citrus color transformation in the orchard featuring high accuracy and fidelity. A total of 107 sample Navel oranges were observed during the color transformation period, resulting in a dataset containing 7,535 citrus images. A framework is proposed that integrates visual saliency into deep learning, and it consists of a segmentation network, a deep mask-guided generative network, and a loss network with manually designed loss functions. Moreover, the fusion of image features and temporal information enables one single model to predict the rind color at different time intervals, thus effectively shrinking the number of model parameters. The semantic segmentation network of the framework achieves the mean intersection over a union score of 0.9694, and the generative network obtains a peak signal-to-noise ratio of 30.01 and a mean local style loss score of 2.710, which indicate both high quality and similarity of the generated images and are also consistent with human perception. To ease the applications in the real world, the model is ported to an Android-based application for mobile devices. The methods can be readily expanded to other fruit crops with a color transformation period. The dataset and the source code are publicly available at GitHub.

A multi-omics integrative network map of maize.

Networks are powerful tools to uncover functional roles of genes in phenotypic variation at a system-wide scale. Here, we constructed a maize network map that contains the genomic, transcriptomic, translatomic and proteomic networks across maize development. This map comprises over 2.8 million edges in more than 1,400 functional subnetworks, demonstrating an extensive network divergence of duplicated genes. We applied this map to identify factors regulating flowering time and identified 2,651 genes enriched in eight subnetworks. We validated the functions of 20 genes, including 18 with previously unknown connections to flowering time in maize. Furthermore, we uncovered a flowering pathway involving histone modification. The multi-omics integrative network map illustrates the principles of how molecular networks connect different types of genes and potential pathways to map a genome-wide functional landscape in maize, which should be applicable in a wide range of species.

Deep Transfer Learning for Ni-Based Superalloys Microstructure Recognition on γ' Phase.

Ni-based superalloys are widely used to manufacture the critical hot-end components of aviation jet engines and various industrial gas turbines. The analysis of Ni-based superalloys microstructures is an important research task during the design and development of superalloys. The material microstructure information can only be understood by experts in the long history. Image segmentation and recognition are developing techniques for accelerating the microstructure analysis automatically. Although deep learning techniques have achieved satisfactory performance, they usually suffer from generalization, i.e., performing worse on a new dataset. In this paper, a deep transfer learning method which just needs a small number of labeled images is proposed to perform the microstructure recognition on γ' phase. To evaluate the effectiveness, we homely prepare two Ni-based superalloys at temperatures 900 °C and 1000 °C, and manually annotate two datasets named as W-900 and W-1000. Experimental results demonstrate that the proposed method only needs 3 and 5 labeled images to achieve state-of-the-art segmentation accuracy during the transfer from W-900 to W-1000 and the transfer from W-1000 to W-900, while enjoying the advantage of fast convergence. In addition, a simple and effective software for the Ni-based superalloys microstructure recognition on γ' phase is developed to improve the efficiency of materials experts, which will greatly facilitate the design of new Ni-base superalloys and even other multicomponent alloys.

DeepBSA: A deep-learning algorithm improves bulked segregant analysis for dissecting complex traits.

Bulked segregant analysis (BSA) is a rapid, cost-effective method for mapping mutations and quantitative trait loci (QTLs) in animals and plants based on high-throughput sequencing. However, the algorithms currently used for BSA have not been systematically evaluated and are complex and fallible to operate. We developed a BSA method driven by deep learning, DeepBSA, for QTL mapping and functional gene cloning. DeepBSA is compatible with a variable number of bulked pools and performed well with various simulated and real datasets in both animals and plants. DeepBSA outperformed all other algorithms when comparing absolute bias and signal-to-noise ratio. Moreover, we applied DeepBSA to an F2 segregating maize population of 7160 individuals and uncovered five candidate QTLs, including three well-known plant-height genes. Finally, we developed a user-friendly graphical user interface for DeepBSA, by integrating five widely used BSA algorithms and our two newly developed algorithms, that is easy to operate and can quickly map QTLs and functional genes. The DeepBSA software is freely available to non-commercial users at http://zeasystemsbio.hzau.edu.cn/tools.html and https://github.com/lizhao007/DeepBSA.

Supervised Optimal Chemotherapy Regimen Based on Offline Reinforcement Learning.

In recent years, reinforcement learning (RL) has achieved a remarkable achievement and it has attracted researchers' attention in modeling real-life scenarios by expanding its research beyond conventional complex games. Prediction of optimal treatment regimens from observational real clinical data is being popularized, and more advanced versions of RL algorithms are being implemented in the literature. However, RL-generated medications still need careful supervision of expertise parties or doctors in healthcare. Hence, in this paper, a Supervised Optimal Chemotherapy Regimen (SOCR) approach to investigate optimal chemotherapy-dosing schedule for cancer patients was presented by using Offline Reinforcement Learning. The optimal policy suggested by the RL approach was supervised by incorporating previous treatment decisions of oncologists, which could add clinical expertise knowledge on algorithmic results. Presented SOCR approach followed a model-based architecture using conservative Q-Learning (CQL) algorithm. The developed model was tested using a manually constructed database of forty Stage-IV colon cancer patients, receiving line-1 chemotherapy treatments, who were clinically classified as 'Bevacizumab based patient' and 'Cetuximab based patient'. Experimental results revealed that the supervision from the oncologists has considered the effect to stabilize chemotherapy regimen and it was suggested that the proposed framework could be successfully used as a supportive model for oncologists in deciding their treatment decisions.

A common hub for sleep and motor control in the substantia nigra.

The arousal state of the brain covaries with the motor state of the animal. How these state changes are coordinated remains unclear. We discovered that sleep-wake brain states and motor behaviors are coregulated by shared neurons in the substantia nigra pars reticulata (SNr). Analysis of mouse home-cage behavior identified four states with different levels of brain arousal and motor activity: locomotion, nonlocomotor movement, quiet wakefulness, and sleep; transitions occurred not randomly but primarily between neighboring states. The glutamic acid decarboxylase 2 but not the parvalbumin subset of SNr γ-aminobutyric acid (GABA)-releasing (GABAergic) neurons was preferentially active in states of low motor activity and arousal. Their activation or inactivation biased the direction of natural behavioral transitions and promoted or suppressed sleep, respectively. These GABAergic neurons integrate wide-ranging inputs and innervate multiple arousal-promoting and motor-control circuits through extensive collateral projections.

Learning With Coefficient-Based Regularized Regression on Markov Resampling.

Big data research has become a globally hot topic in recent years. One of the core problems in big data learning is how to extract effective information from the huge data. In this paper, we propose a Markov resampling algorithm to draw useful samples for handling coefficient-based regularized regression (CBRR) problem. The proposed Markov resampling algorithm is a selective sampling method, which can automatically select uniformly ergodic Markov chain (u.e.M.c.) samples according to transition probabilities. Based on u.e.M.c. samples, we analyze the theoretical performance of CBRR algorithm and generalize the existing results on independent and identically distributed observations. To be specific, when the kernel is infinitely differentiable, the learning rate depending on the sample size $m$ can be arbitrarily close to $\mathcal {O}(m^{-1})$ under a mild regularity condition on the regression function. The good generalization ability of the proposed method is validated by experiments on simulated and real data sets.

A Novel 3D Connection Algorithm of Mitochondria From ATUM-SEM Stacks Based on Segmentation Information in Context.

Recent researches have shown that the relation between mitochondrial function and degenerative disorders is closely related to aging, such as Alzheimer's and Parkinson's diseases. Because these studies expose the need for detailed analysis of high-resolution physical alterations in mitochondria, three dimensional (3D) visualization of mitochondria from electron microscopy (EM) images is coming into prominence. To this end, how to develop suitable segmentation algorithms and connection algorithms has attracted our attentions. Since previous algorithms have shown preferable segmentation performance on mitochondria with different shapes and sizes. In this paper, we propose to utilize the segmentation information instead of detection information in context to obtain the mitochondrial connection relation in adjacent layers. Additionally, different from previous methods, we present a novel and effective connection approach by obtaining sparse matrixes and implementing a forward connection mode. Experiments on automated tape-collecting ultramicrotome scanning electron microscopy (ATUM-SEM) stacks demonstrate that our approach can effectively handle with the case of split and merge, and achieve a comparable connection quality measured by split error and merge error.

Automatic Detection and Segmentation of Mitochondria from SEM Images using Deep Neural Network.

Investigating the link between mitochondrial function and its physical structure is a hot topic in neurobiology research. With the rapid development of Scanning Electron Microscope (SEM), we can look closely into the fine mitochondrial structure with high resolution. Consequently, many meaningful researches have focused on how to detect and segment the mitochondria from EM images. Due to the complex background, hand-crafted features designed by traditional algorithms cannot provide satisfying results. In this paper, we propose an effective deep neural network improved from Mask R-CNN to produce the detection and segmentation results. On this base, we use the morphological processing and mitochondrial context information to rectify the local misleading results. The valuation was performed on two widely used datasets (FIB-SEM and ATUMSEM), and the results demonstrate that the proposed method has comparable performance than state-of-the-art methods.

A fast forward 3D connection algorithm for mitochondria and synapse segmentations from serial EM images.

BACKGROUND: It is becoming increasingly clear that the quantification of mitochondria and synapses is of great significance to understand the function of biological nervous systems. Electron microscopy (EM), with the necessary resolution in three directions, is the only available imaging method to look closely into these issues. Therefore, estimating the number of mitochondria and synapses from the serial EM images is coming into prominence. Since previous studies have achieved preferable 2D segmentation performance, it holds great promise to obtain the 3D connection relationship from the 2D segmentation results. RESULTS: In this paper, we improve upon Matlab's function bwconncomp and propose a fast forward 3D connection algorithm for mitochondria and synapse segmentations from serial EM images. To benchmark the performance of the proposed method, two EM datasets with the annotated ground truth are produced for mitochondria and synapses, respectively. Experimental results show that the proposed method can achieve the preferable connection performance that closely matches the ground truth. Moreover, it greatly reduces the computational burden and alleviates the memory requirements compared with the function bwconncomp. CONCLUSIONS: The proposed method can be deemed as an effective strategy to obtain the 3D connection relationship from serial mitochondria and synapse segmentations. It is helpful to accurately and quickly quantify the statistics of the numbers, volumes, surface areas, and lengths, which will greatly facilitate the data analysis of neurobiology research.

Automatic Mitochondria Segmentation for EM Data Using a 3D Supervised Convolutional Network.

Recent studies have supported the relation between mitochondrial functions and degenerative disorders related to ageing, such as Alzheimer's and Parkinson's diseases. Since these studies have exposed the need for detailed and high-resolution analysis of physical alterations in mitochondria, it is necessary to be able to perform segmentation and 3D reconstruction of mitochondria. However, due to the variety of mitochondrial structures, automated mitochondria segmentation and reconstruction in electron microscopy (EM) images have proven to be a difficult and challenging task. This paper puts forward an effective and automated pipeline based on deep learning to realize mitochondria segmentation in different EM images. The proposed pipeline consists of three parts: (1) utilizing image registration and histogram equalization as image pre-processing steps to maintain the consistency of the dataset; (2) proposing an effective approach for 3D mitochondria segmentation based on a volumetric, residual convolutional and deeply supervised network; and (3) employing a 3D connection method to obtain the relationship of mitochondria and displaying the 3D reconstruction results. To our knowledge, we are the first researchers to utilize a 3D fully residual convolutional network with a deeply supervised strategy to improve the accuracy of mitochondria segmentation. The experimental results on anisotropic and isotropic EM volumes demonstrate the effectiveness of our method, and the Jaccard index of our segmentation (91.8% in anisotropy, 90.0% in isotropy) and F1 score of detection (92.2% in anisotropy, 90.9% in isotropy) suggest that our approach achieved state-of-the-art results. Our fully automated pipeline contributes to the development of neuroscience by providing neurologists with a rapid approach for obtaining rich mitochondria statistics and helping them elucidate the mechanism and function of mitochondria.

An automated pipeline for mitochondrial segmentation on ATUM-SEM stacks.

It is possible now to look more closely into mitochondrial physical structures due to the rapid development of electron microscope (EM). Mitochondrial physical structures play important roles in both cellular physiology and neuronal functions. Unfortunately, the segmentation of mitochondria from EM images has proven to be a difficult and challenging task, due to the presence of various subcellular structures, as well as image distortions in the sophisticated background. Although the current state-of-the-art algorithms have achieved some promising results, they have demonstrated poor performances on these mitochondria which are in close proximity to vesicles or various membranes. In order to overcome these limitations, this study proposes explicitly modelling the mitochondrial double membrane structures, and acquiring the image edges by way of ridge detection rather than by image gradient. In addition, this study also utilizes group-similarity in context to further optimize the local misleading segmentation. Then, the experimental results determined from the images acquired by automated tape-collecting ultramicrotome scanning electron microscopy (ATUM-SEM) demonstrate the effectiveness of this study's proposed algorithm.