Learning mapping by curve iteration estimation For real-time underwater image enhancement.

The degradation and attenuation of light in underwater images impose constraints on underwater vision tasks. However, the complexity and the low real-time performance of most current image enhancement algorithms make them challenging in practical applications. To address the above issues, we propose a new lightweight framework for underwater image enhancement. We adopt the curve estimation to learn the mapping between images rather than end-to-end networks, which greatly reduces the requirement for computing resources. Firstly, a designed iterative curve with parameters is used to simulate the mapping from the raw to the enhanced image. Then, the parameters of this curve are learned with a parameter estimation network called CieNet and a set of loss functions. Experimental results demonstrate that our proposed method is superior to existing algorithms in terms of evaluating indexes and visual perception quality. Furthermore, our highly lightweight network enables it to be easily integrated into small devices, making it highly applicable. The extremely short running-time of our method facilitates real-time underwater image enhancement.

MOCAT: multi-omics integration with auxiliary classifiers enhanced autoencoder.

BACKGROUND: Integrating multi-omics data is emerging as a critical approach in enhancing our understanding of complex diseases. Innovative computational methods capable of managing high-dimensional and heterogeneous datasets are required to unlock the full potential of such rich and diverse data. METHODS: We propose a Multi-Omics integration framework with auxiliary Classifiers-enhanced AuToencoders (MOCAT) to utilize intra- and inter-omics information comprehensively. Additionally, attention mechanisms with confidence learning are incorporated for enhanced feature representation and trustworthy prediction. RESULTS: Extensive experiments were conducted on four benchmark datasets to evaluate the effectiveness of our proposed model, including BRCA, ROSMAP, LGG, and KIPAN. Our model significantly improved most evaluation measurements and consistently surpassed the state-of-the-art methods. Ablation studies showed that the auxiliary classifiers significantly boosted classification accuracy in the ROSMAP and LGG datasets. Moreover, the attention mechanisms and confidence evaluation block contributed to improvements in the predictive accuracy and generalizability of our model. CONCLUSIONS: The proposed framework exhibits superior performance in disease classification and biomarker discovery, establishing itself as a robust and versatile tool for analyzing multi-layer biological data. This study highlights the significance of elaborated designed deep learning methodologies in dissecting complex disease phenotypes and improving the accuracy of disease predictions.

UIEOGP: an underwater image enhancement method based on optical geometric properties.

Due to the inconsistent absorption and scattering effects of different wavelengths of light, underwater images often suffer from color casts, blurred details, and low visibility. To address this image degradation problem, we propose a robust and efficient underwater image enhancement method named UIEOGP. It can be divided into the following three steps. First, according to the light attenuation effect presented by Lambert Beer's law, combined with the variance change after attenuation, we estimate the depth of field in the underwater image. Then, we propose a local-based color correction algorithm to address the color cast issue in underwater images, employing the statistical distribution law. Finally, drawing inspiration from the law of light propagation, we propose detail enhancement algorithms, each based on the geometric properties of circles and ellipses, respectively. The enhanced images produced by our method feature vibrant colors, improved contrast, and sharper detail. Extensive experiments show that our method outperforms current state-of-the-art methods. In further experiments, we found that our method is beneficial for downstream tasks of underwater image processing, such as the detection of keypoints and edges in underwater images.

Medical matting: Medical image segmentation with uncertainty from the matting perspective.

High-quality manual labeling of ambiguous and complex-shaped targets with binary masks can be challenging. The weakness of insufficient expression of binary masks is prominent in segmentation, especially in medical scenarios where blurring is prevalent. Thus, reaching a consensus among clinicians through binary masks is more difficult in multi-person labeling cases. These inconsistent or uncertain areas are related to the lesions' structure and may contain anatomical information conducive to providing an accurate diagnosis. However, recent research focuses on uncertainties of model training and data labeling. None of them has investigated the influence of the ambiguous nature of the lesion itself. Inspired by image matting, this paper introduces a soft mask called alpha matte to medical scenes. It can describe the lesions with more details better than a binary mask. Moreover, it can also be used as a new uncertainty quantification method to represent uncertain areas, filling the gap in research on the uncertainty of lesion structure. In this work, we introduce a multi-task framework to generate binary masks and alpha mattes, which outperforms all state-of-the-art matting algorithms compared. The uncertainty map is proposed to imitate the trimap in matting methods, which can highlight fuzzy areas and improve matting performance. We have created three medical datasets with alpha mattes to address the lack of available matting datasets in medical fields and evaluated the effectiveness of our proposed method on them comprehensively. Furthermore, experiments demonstrate that the alpha matte is a more effective labeling method than the binary mask from both qualitative and quantitative aspects.

Identify potential driver genes for PAX-FOXO1 fusion-negative rhabdomyosarcoma through frequent gene co-expression network mining.

Background: Rhabdomyosarcoma (RMS) is a soft tissue sarcoma usually originated from skeletal muscle. Currently, RMS classification based on PAX-FOXO1 fusion is widely adopted. However, compared to relatively clear understanding of the tumorigenesis in the fusion-positive RMS, little is known for that in fusion-negative RMS (FN-RMS). Methods: We explored the molecular mechanisms and the driver genes of FN-RMS through frequent gene co-expression network mining (fGCN), differential copy number (CN) and differential expression analyses on multiple RMS transcriptomic datasets. Results: We obtained 50 fGCN modules, among which five are differentially expressed between different fusion status. A closer look showed 23% of Module 2 genes are concentrated on several cytobands of chromosome 8. Upstream regulators such as MYC, YAP1, TWIST1 were identified for the fGCN modules. Using in a separate dataset we confirmed that, comparing to FP-RMS, 59 Module 2 genes show consistent CN amplification and mRNA overexpression, among which 28 are on the identified chr8 cytobands. Such CN amplification and nearby MYC (also resides on one of the above cytobands) and other upstream regulators (YAP1, TWIST1) may work together to drive FN-RMS tumorigenesis and progression. Up to 43.1% downstream targets of Yap1 and 45.8% of the targets of Myc are differentially expressed in FN-RMS vs. normal comparisons, which also confirmed the driving force of these regulators. Discussion: We discovered that copy number amplification of specific cytobands on chr8 and the upstream regulators MYC, YAP1 and TWIST1 work together to affect the downstream gene co-expression and promote FN-RMS tumorigenesis and progression. Our findings provide new insights for FN-RMS tumorigenesis and offer promising targets for precision therapy. Experimental investigation about the functions of identified potential drivers in FN-RMS are in progress.

A Distinguishable Pseudo-Feature Synthesis Method for Generalized Zero-Shot Learning.

Generalized zero-shot learning (GZSL) aims to classify seen classes and unseen classes that are disjoint simultaneously. Hybrid approaches based on pseudo-feature synthesis are currently the most popular among GZSL methods. However, they suffer from problems of negative transfer and low-quality class discriminability, causing poor classification accuracy. To address them, we propose a novel GZSL method of distinguishable pseudo-feature synthesis (DPFS). The DPFS model can provide high-quality distinguishable characteristics for both seen and unseen classes. Firstly, the model is pretrained by a distance prediction loss to avoid overfitting. Then, the model only selects attributes of similar seen classes and makes sparse representations based on attributes for unseen classes, thereby overcoming negative transfer. After the model synthesizes pseudo-features for unseen classes, it disposes of the pseudo-feature outliers to improve the class discriminability. The pseudo-features are fed into a classifier of the model together with features of seen classes for GZSL classification. Experimental results on four benchmark datasets verify that the proposed DPFS has GZSL classification performance better than that in existing methods.

3D matting: A benchmark study on soft segmentation method for pulmonary nodules applied in computed tomography.

Usually, lesions are not isolated but are associated with the surrounding tissues. For example, the growth of a tumour can depend on or infiltrate into the surrounding tissues. Due to the pathological nature of the lesions, it is challenging to distinguish their boundaries in medical imaging. However, these uncertain regions may contain diagnostic information. Therefore, the simple binarization of lesions by traditional binary segmentation can result in the loss of diagnostic information. In this work, we introduce the image matting into the 3D scenes and use the alpha matte, i.e., a soft mask, to describe lesions in a 3D medical image. The traditional soft mask acted as a training trick to compensate for the easily mislabelled or under-labelled ambiguous regions. In contrast, 3D matting uses soft segmentation to characterize the uncertain regions more finely, which means that it retains more structural information for subsequent diagnosis and treatment. The current study of image matting methods in 3D is limited. To address this issue, we conduct a comprehensive study of 3D matting, including both traditional and deep-learning-based methods. We adapt four state-of-the-art 2D image matting algorithms to 3D scenes and further customize the methods for CT images to calibrate the alpha matte with the radiodensity. Moreover, we propose the first end-to-end deep 3D matting network and implement a solid 3D medical image matting benchmark. Its efficient counterparts are also proposed to achieve a good performance-computation balance. Furthermore, there is no high-quality annotated dataset related to 3D matting, slowing down the development of data-driven deep-learning-based methods. To address this issue, we construct the first 3D medical matting dataset. The validity of the dataset was verified through clinicians' assessments and downstream experiments. The dataset and codes will be released to encourage further research.1.

SPCS: a spatial and pattern combined smoothing method for spatial transcriptomic expression.

High-dimensional, localized ribonucleic acid (RNA) sequencing is now possible owing to recent developments in spatial transcriptomics (ST). ST is based on highly multiplexed sequence analysis and uses barcodes to match the sequenced reads to their respective tissue locations. ST expression data suffer from high noise and dropout events; however, smoothing techniques have the promise to improve the data interpretability prior to performing downstream analyses. Single-cell RNA sequencing (scRNA-seq) data similarly suffer from these limitations, and smoothing methods developed for scRNA-seq can only utilize associations in transcriptome space (also known as one-factor smoothing methods). Since they do not account for spatial relationships, these one-factor smoothing methods cannot take full advantage of ST data. In this study, we present a novel two-factor smoothing technique, spatial and pattern combined smoothing (SPCS), that employs the k-nearest neighbor (kNN) technique to utilize information from transcriptome and spatial relationships. By performing SPCS on multiple ST slides from pancreatic ductal adenocarcinoma (PDAC), dorsolateral prefrontal cortex (DLPFC) and simulated high-grade serous ovarian cancer (HGSOC) datasets, smoothed ST slides have better separability, partition accuracy and biological interpretability than the ones smoothed by preexisting one-factor methods. Source code of SPCS is provided in Github (https://github.com/Usos/SPCS).

Distance correlation application to gene co-expression network analysis.

BACKGROUND: To construct gene co-expression networks, it is necessary to evaluate the correlation between different gene expression profiles. However, commonly used correlation metrics, including both linear (such as Pearson's correlation) and monotonic (such as Spearman's correlation) dependence metrics, are not enough to observe the nature of real biological systems. Hence, introducing a more informative correlation metric when constructing gene co-expression networks is still an interesting topic. RESULTS: In this paper, we test distance correlation, a correlation metric integrating both linear and non-linear dependence, with other three typical metrics (Pearson's correlation, Spearman's correlation, and maximal information coefficient) on four different arrays (macrophage and liver) and RNA-seq (cervical cancer and pancreatic cancer) datasets. Among all the metrics, distance correlation is distribution free and can provide better performance on complex relationships and anti-outlier. Furthermore, distance correlation is applied to Weighted Gene Co-expression Network Analysis (WGCNA) for constructing a gene co-expression network analysis method which we named Distance Correlation-based Weighted Gene Co-expression Network Analysis (DC-WGCNA). Compared with traditional WGCNA, DC-WGCNA can enhance the result of enrichment analysis and improve the module stability. CONCLUSIONS: Distance correlation is better at revealing complex biological relationships between gene profiles compared with other correlation metrics, which contribute to more meaningful modules when analyzing gene co-expression networks. However, due to the high time complexity of distance correlation, the implementation requires more computer memory.

The morphology of stigma of Asteraceae observed by scanning electron microscopy.

The morphology of stigma has taxonomic values. To further explore the taxonomy of family Asteraceae, the morphological characteristics of stigma of 28 genera, 32 species, and two varieties in the family were observed using scanning electron microscopy. The results indicated that the stigma morphology of these Asteraceae plants could be divided into 10 types, of which eight are reported for the first time. The morphological characteristics of stigma support the close relationship between genera Aster and Erigeron and among genera Sonchus, Taraxacum, and Youngia. Our results enriched the stigma type diversity data and provided a morphological basis for the study of the phylogenetic evolution of Asteraceae.

Model-free nonlinear robust control design via online critic learning.

Online critic learning or solving robust control problems of complex systems usually requires knowledge about system dynamics. In order to achieve these goals in data-driven method, a new performance index related to the decreasing rate of the conventional cost is designed. The corresponding optimal control policy can be approximated online using a new actor-critic scheme with three neural networks, without depending on initial stable control and knowledge about system dynamics. The learning process and the learned control policy show excellent robustness. Numerical simulations and an inverted pendulum experiment show that compared with benchmark methods, the proposed method relaxes the dependence on initial admissible control and exhibits better disturbance attenuation performance.

Morphological differences between anemophilous and entomophilous pollen.

In order to provide a palynological guide for the identification of insect-carrying pollen grains, we studied the pollen grains of 10 anemophilous species and 10 entomophilous species in the Beijing urban area using light and scanning electron microscopies. We found that anemophilous pollen grains are small, spheroidal, or oblate spheroidal, while entomophilous pollen grains are medium and oblate. Comparison of the exine thickness and surface ornamentation showed that anemophilous pollen grains have significantly thinner exine and smoother surface ornamentation than entomophilous pollen grains. The results also revealed pollen characteristics adaptive to different pollination types. Overall, our study indicated that pollen morphology might be helpful for preliminary identification of anemophilous and entomophilous pollen.

TPSC: a module detection method based on topology potential and spectral clustering in weighted networks and its application in gene co-expression module discovery.

BACKGROUND: Gene co-expression networks are widely studied in the biomedical field, with algorithms such as WGCNA and lmQCM having been developed to detect co-expressed modules. However, these algorithms have limitations such as insufficient granularity and unbalanced module size, which prevent full acquisition of knowledge from data mining. In addition, it is difficult to incorporate prior knowledge in current co-expression module detection algorithms. RESULTS: In this paper, we propose a novel module detection algorithm based on topology potential and spectral clustering algorithm to detect co-expressed modules in gene co-expression networks. By testing on TCGA data, our novel method can provide more complete coverage of genes, more balanced module size and finer granularity than current methods in detecting modules with significant overall survival difference. In addition, the proposed algorithm can identify modules by incorporating prior knowledge. CONCLUSION: In summary, we developed a method to obtain as much as possible information from networks with increased input coverage and the ability to detect more size-balanced and granular modules. In addition, our method can integrate data from different sources. Our proposed method performs better than current methods with complete coverage of input genes and finer granularity. Moreover, this method is designed not only for gene co-expression networks but can also be applied to any general fully connected weighted network.

TPQCI: A topology potential-based method to quantify functional influence of copy number variations.

Copy number variation (CNV) is a major type of chromosomal structural variation that play important roles in many diseases including cancers. Due to genome instability, a large number of CNV events can be detected in diseases such as cancer. Therefore, it is important to identify the functionally important CNVs in diseases, which currently still poses a challenge in genomics. One of the critical steps to solve the problem is to define the influence of CNV. In this paper, we provide a topology potential based method, TPQCI, to quantify this kind of influence by integrating statistics, gene regulatory associations, and biological function information. We used this metric to detect functionally enriched genes on genomic segments with CNV in breast cancer and multiple myeloma and discovered biological functions influenced by CNV. Our results demonstrate that, by using our proposed TPQCI metric, we can detect disease-specific genes that are influenced by CNVs. Source codes of TPQCI are provided in Github (https://github.com/usos/TPQCI).

Stratification of Estrogen Receptor-Negative Breast Cancer Patients by Integrating the Somatic Mutations and Transcriptomic Data.

Patients with estrogen receptor-negative breast cancer generally have a worse prognosis than estrogen receptor-positive patients. Nevertheless, a significant proportion of the estrogen receptor-negative cases have favorable outcomes. Identifying patients with a good prognosis, however, remains difficult, as recent studies are quite limited. The identification of molecular biomarkers is needed to better stratify patients. The significantly mutated genes may be potentially used as biomarkers to identify the subtype and to predict outcomes. To identify the biomarkers of receptor-negative breast cancer among the significantly mutated genes, we developed a workflow to screen significantly mutated genes associated with the estrogen receptor in breast cancer by a gene coexpression module. The similarity matrix was calculated with distance correlation to obtain gene modules through a weighted gene coexpression network analysis. The modules highly associated with the estrogen receptor, called important modules, were enriched for breast cancer-related pathways or disease. To screen significantly mutated genes, a new gene list was obtained through the overlap of the important module genes and the significantly mutated genes. The genes on this list can be used as biomarkers to predict survival of estrogen receptor-negative breast cancer patients. Furthermore, we selected six hub significantly mutated genes in the gene list which were also able to separate these patients. Our method provides a new and alternative method for integrating somatic gene mutations and expression data for patient stratification of estrogen receptor-negative breast cancers.

K-Module Algorithm: An Additional Step to Improve the Clustering Results of WGCNA Co-Expression Networks.

Among biological networks, co-expression networks have been widely studied. One of the most commonly used pipelines for the construction of co-expression networks is weighted gene co-expression network analysis (WGCNA), which can identify highly co-expressed clusters of genes (modules). WGCNA identifies gene modules using hierarchical clustering. The major drawback of hierarchical clustering is that once two objects are clustered together, it cannot be reversed; thus, re-adjustment of the unbefitting decision is impossible. In this paper, we calculate the similarity matrix with the distance correlation for WGCNA to construct a gene co-expression network, and present a new approach called the k-module algorithm to improve the WGCNA clustering results. This method can assign all genes to the module with the highest mean connectivity with these genes. This algorithm re-adjusts the results of hierarchical clustering while retaining the advantages of the dynamic tree cut method. The validity of the algorithm is verified using six datasets from microarray and RNA-seq data. The k-module algorithm has fewer iterations, which leads to lower complexity. We verify that the gene modules obtained by the k-module algorithm have high enrichment scores and strong stability. Our method improves upon hierarchical clustering, and can be applied to general clustering algorithms based on the similarity matrix, not limited to gene co-expression network analysis.

Integration of molecular features with clinical information for predicting outcomes for neuroblastoma patients.

BACKGROUND: Neuroblastoma is one of the most common types of pediatric cancer. In current neuroblastoma prognosis, patients can be stratified into high- and low-risk groups. Generally, more than 90% of the patients in the low-risk group will survive, while less than 50% for those with the high-risk disease will survive. Since the so-called "high-risk" patients still contain patients with mixed good and poor outcomes, more refined stratification needs to be established so that for the patients with poor outcome, they can receive prompt and individualized treatment to improve their long-term survival rate, while the patients with good outcome can avoid unnecessary over treatment. METHODS: We first mined co-expressed gene modules from microarray and RNA-seq data of neuroblastoma samples using the weighted network mining algorithm lmQCM, and summarize the resulted modules into eigengenes. Then patient similarity weight matrix was constructed with module eigengenes using two different approaches. At the last step, a consensus clustering method called Molecular Regularized Consensus Patient Stratification (MRCPS) was applied to aggregate both clinical information (clinical stage and clinical risk level) and multiple eigengene data for refined patient stratification. RESULTS: The integrative method MRCPS demonstrated superior performance to clinical staging or transcriptomic features alone for the NB cohort stratification. It successfully identified the worst prognosis group from the clinical high-risk group, with less than 40% survived in the first 50 months of diagnosis. It also identified highly differentially expressed genes between best prognosis group and worst prognosis group, which can be potential gene biomarkers for clinical testing. CONCLUSIONS: To address the need for better prognosis and facilitate personalized treatment on neuroblastoma, we modified the recently developed bioinformatics workflow MRCPS for refined patient prognosis. It integrates clinical information and molecular features such as gene co-expression for prognosis. This clustering workflow is flexible, allowing the integration of both categorical and numerical data. The results demonstrate the power of survival prognosis with this integrative analysis workflow, with superior prognostic performance to only using transcriptomic data or clinical staging/risk information alone. REVIEWERS: This article was reviewed by Lan Hu, Haibo Liu, Julie Zhu and Aleksandra Gruca.

Integrative analysis based on survival associated co-expression gene modules for predicting Neuroblastoma patients' survival time.

BACKGROUND: More than 90% of neuroblastoma patients are cured in the low-risk group while only less than 50% for those with high-risk disease can be cured. Since the high-risk patients still have poor outcomes, we need more accurate stratification to establish an individualized precise treatment plan for the patients to improve the long-term survival rate. RESULTS: We focus on extracting features and providing a workflow to improve survival prediction for neuroblastoma patients. With a workflow for gene co-expression network (GCN) mining in microarray and RNA-Seq datasets, we extracted molecular features from each co-expressed module and summarized them into eigengenes. Then we adopted the lasso-regularized Cox proportional hazards model to select the most informative eigengene features regarding association to the risk of metastasis. Nine eigengenes were selected which show strong association with patient survival prognosis. All of the nine corresponding gene modules also have highly enriched biological functions or cytoband locations. Three of them are unique modules to RNA-Seq data, which complement the modules from microarray data in terms of survival prognosis. We then merged all eigengenes from these unique modules and used an integrative method called Similarity Network Fusion to test the prognostic power of these eigengenes for prognosis. The prognostic accuracies are significantly improved as compared to using all eigengenes, and a subgroup of patients with very poor survival rate was identified. CONCLUSIONS: We first compared GCNs mined from microarray and RNA-seq data. We discovered that each data modality yields unique GCNs, which are enriched with clear biological functions. Then we do module unique analysis and use lasso-cox model to select survival-associated eigengenes. Integration of unique and survival-associated eigengenes from both data types provides complementary information that leads to more accurate survival prognosis. REVIEWERS: Reviewed by Susmita Datta, Marco Chierici and Dimitar Vassilev.

Integrative Analysis of Histopathological Images and Genomic Data Predicts Clear Cell Renal Cell Carcinoma Prognosis.

In cancer, both histopathologic images and genomic signatures are used for diagnosis, prognosis, and subtyping. However, combining histopathologic images with genomic data for predicting prognosis, as well as the relationships between them, has rarely been explored. In this study, we present an integrative genomics framework for constructing a prognostic model for clear cell renal cell carcinoma. We used patient data from The Cancer Genome Atlas (n = 410), extracting hundreds of cellular morphologic features from digitized whole-slide images and eigengenes from functional genomics data to predict patient outcome. The risk index generated by our model correlated strongly with survival, outperforming predictions based on considering morphologic features or eigengenes separately. The predicted risk index also effectively stratified patients in early-stage (stage I and stage II) tumors, whereas no significant survival difference was observed using staging alone. The prognostic value of our model was independent of other known clinical and molecular prognostic factors for patients with clear cell renal cell carcinoma. Overall, this workflow and the shared software code provide building blocks for applying similar approaches in other cancers. Cancer Res; 77(21); e91-100. ©2017 AACR.

A fast and stable vascular deformation scheme for interventional surgery training system.

BACKGROUND: The emergence and development of robot assistant interventional vascular surgery technologies have benefited many patients with cardiovascular or cerebrovascular diseases. Due to the absence of effective training measures, these new advanced technologies have not been fully utilized and only few experienced surgeons can perform such complicated surgeries so far. In order to solve such problems, virtual reality based vascular interventional surgery training system, a promising way to train young surgeons or assist experienced surgeons to perform surgery, has been widely studied. METHODS: In this paper, we mainly conduct a thorough study on both reliable deformation and high real-time performance of an interactive surgery training system. An efficient hybrid geometric blood vessel model which handles the collision detection query and vascular deformation calculation separately is employed to enhance the real-time performance of our surgery training system. In addition, a position-based dynamic approach with volume conservation constraint is used to improve the vascular deformation result. Finally, a hash table based spatial adaptive acceleration algorithm which makes the training system much more efficient and reliable is described. RESULTS: Several necessary experiments are conducted to validate the vascular deformation scheme presented in this paper. From the results we can see that the position-based dynamic modeling method with volume conservation constraint can prevent the vascular deformation from the issue of penetration. In addition, the deformation calculation with spatial acceleration algorithm has enhanced the real-time performance significantly. CONCLUSION: The corresponding experimental results indicate that both the hybrid geometric blood vessel model and the hash table based spatial adaptive acceleration algorithm can enhance the performance of our surgery training system greatly without losing the deformation accuracy.