Deep Clustering: A Comprehensive Survey.

Cluster analysis plays an indispensable role in machine learning and data mining. Learning a good data representation is crucial for clustering algorithms. Recently, deep clustering (DC), which can learn clustering-friendly representations using deep neural networks (DNNs), has been broadly applied in a wide range of clustering tasks. Existing surveys for DC mainly focus on the single-view fields and the network architectures, ignoring the complex application scenarios of clustering. To address this issue, in this article, we provide a comprehensive survey for DC in views of data sources. With different data sources, we systematically distinguish the clustering methods in terms of methodology, prior knowledge, and architecture. Concretely, DC methods are introduced according to four categories, i.e., traditional single-view DC, semi-supervised DC, deep multiview clustering (MVC), and deep transfer clustering. Finally, we discuss the open challenges and potential future opportunities in different fields of DC.

CDDnet: Cross-domain denoising network for low-dose CT image via local and global information alignment.

The domain shift problem has emerged as a challenge in cross-domain low-dose CT (LDCT) image denoising task, where the acquisition of a sufficient number of medical images from multiple sources may be constrained by privacy concerns. In this study, we propose a novel cross-domain denoising network (CDDnet) that incorporates both local and global information of CT images. To address the local component, a local information alignment module has been proposed to regularize the similarity between extracted target and source features from selected patches. To align the general information of the semantic structure from a global perspective, an autoencoder is adopted to learn the latent correlation between the source label and the estimated target label generated by the pre-trained denoiser. Experimental results demonstrate that our proposed CDDnet effectively alleviates the domain shift problem, outperforming other deep learning-based and domain adaptation-based methods under cross-domain scenarios.

EEGDnet: Fusing non-local and local self-similarity for EEG signal denoising with transformer.

Electroencephalogram (EEG) has shown a useful approach to produce a brain-computer interface (BCI). One-dimensional (1-D) EEG signal is yet easily disturbed by certain artifacts (a.k.a. noise) due to the high temporal resolution. Thus, it is crucial to remove the noise in received EEG signal. Recently, deep learning-based EEG signal denoising approaches have achieved impressive performance compared with traditional ones. It is well known that the characteristics of self-similarity (including non-local and local ones) of data (e.g., natural images and time-domain signals) are widely leveraged for denoising. However, existing deep learning-based EEG signal denoising methods ignore either the non-local self-similarity (e.g., 1-D convolutional neural network) or local one (e.g., fully connected network and recurrent neural network). To address this issue, we propose a novel 1-D EEG signal denoising network with 2-D transformer, namely EEGDnet. Specifically, we comprehensively take into account the non-local and local self-similarity of EEG signal through the transformer module. By fusing non-local self-similarity in self-attention blocks and local self-similarity in feed forward blocks, the negative impact caused by noises and outliers can be reduced significantly. Extensive experiments show that, compared with other state-of-the-art models, EEGDnet achieves much better performance in terms of both quantitative and qualitative metrics. Specifically, EEGDnet can achieve 18% and 11% improvements in correlation coefficients when removing ocular artifacts and muscle artifacts, respectively.

BS-80K: The first large open-access dataset of bone scan images.

BACKGROUND: Radionuclide bone scanning is one of the most common tools in the inspection of bone metastasis. Conventionally, the analysis of bone scan image is derived from manual diagnosing. However, this task requires extensive subjective diagnostic experience and is extremely time-consuming. To this end, a series of studies concerning computer-aided diagnosis via machine learning tools have been proposed. Although some inspiring progress has been achieved, the implemented bone scan image datasets in these research areas are generally too small, private or non-general, which limits their practical significance and impedes the follow-up research. METHOD: To address this issue, we present a large, publicly available and general dataset consisting of 82544 bone scan images associated with 3247 patients from West China Hospital, named BS-80K. In BS-80K, each patient provides two whole bone scan images corresponding to the anterior view (ANT) and the posterior view (POST). For each view, there are 13 region-wise slices of the body parts susceptible to bone metastasis. Based on an authorized original labeling criterion, labels annotated by experienced specialists are offered with the images. Moreover, within each whole body image, multiple bounding boxes containing suspectable hot spots and their annotations are supplied as well. All images in BS-80K have been de-identified to protect patients' privacy. RESULTS: Based on 6 popular deep learning models for classification and object detection, we provide the benchmark for a number of computer-aided medical tasks, including general bone metastasis prediction and object detection for whole body images, and specific bone metastasis prediction for different body parts. According to extensive experiments, the adopted classification models achieve remarkable results in accuracy and specificity (around 95%) on most metastasis prediction tasks, which are approximate to the average ability of corresponding specialists. As for the object detection task, the best average precision of the adopted models reaches 0.2484 and the lowest is 0.1334. DISCUSSION: Through the comparison of metastasis prediction performance between the benchmark and related work, we observe that the widely used models trained by BS-80K achieve significantly better results than the elaborately designed models trained by smaller datasets. This indicates that with the large amount of data, BS-80K has great potential to galvanize the research about computer-aided analysis on bone scan image. CONCLUSION: To the best of our knowledge, BS-80K is the first large publicly available dataset of bone scanning, which favors a wide range of research on computer-aided bone metastasis diagnosis. The full dataset is now available at https://drive.google.com/drive/folders/1DOBkLXgQeREQjF-nQIGNBBzPCb5s7RNu?usp=sharing.

Machine learning based early mortality prediction in the emergency department.

BACKGROUND: It is a great challenge for emergency physicians to early detect the patient's deterioration and prevent unexpected death through a large amount of clinical data, which requires sufficient experience and keen insight. OBJECTIVE: To evaluate the performance of machine learning models in quantifying the severity of emergency department (ED) patients and identifying high-risk patients. METHODS: Using routinely-available demographics, vital signs and laboratory tests extracted from electronic health records (EHRs), a framework based on machine learning and feature engineering was proposed for mortality prediction. Patients who had one complete record of vital signs and laboratory tests in ED were included. The following patients were excluded: pediatric patients aged < 18 years, pregnant woman, and patients died or were discharged or hospitalized within 12 h after admission. Based on 76 original features extracted, 9 machine learning models were adopted to validate our proposed framework. Their optimal hyper-parameters were fine-tuned using the grid search method. The prediction results were evaluated on performance metrics (i.e., accuracy, area under the curve (AUC), recall and precision) with repeated 5-fold cross-validation (CV). The time window from patient admission to the prediction was analyzed at 12 h, 24 h, 48 h, and entire stay. RESULTS: We studied a total of 1114 ED patients with 71.54% (797/1114) survival and 28.46% (317/1114) death in the hospital. The results revealed a more complete time window leads to better prediction performance. Using the entire stay records, the LightGBM model with refined feature engineering demonstrated high discrimination and achieved 93.6% (±0.008) accuracy, 97.6% (±0.003) AUC, 97.1% (±0.008) recall, and 94.2% (±0.006) precision, even if no diagnostic information was utilized. CONCLUSIONS: This study quantifies the criticality of ED patients and appears to have significant potential as a clinical decision support tool in assisting physicians in their clinical routine. While the model requires validation before use elsewhere, the same methodology could be used to create a strong model for the new hospital.

Dual self-paced multi-view clustering.

By utilizing the complementary information from multiple views, multi-view clustering (MVC) algorithms typically achieve much better clustering performance than conventional single-view methods. Although in this field, great progresses have been made in past few years, most existing multi-view clustering methods still suffer the following shortcomings: (1) most MVC methods are non-convex and thus are easily stuck into suboptimal local minima; (2) the effectiveness of these methods is sensitive to the existence of noises or outliers; and (3) the qualities of different features and views are usually ignored, which can also influence the clustering result. To address these issues, we propose dual self-paced multi-view clustering (DSMVC) in this paper. Specifically, DSMVC takes advantage of self-paced learning to tackle the non-convex issue. By applying a soft-weighting scheme of self-paced learning for instances, the negative impact caused by noises and outliers can be significantly reduced. Moreover, to alleviate the feature and view quality issues, we develop a novel feature selection approach in a self-paced manner and a weighting term for views. Experimental results on real-world data sets demonstrate the effectiveness of the proposed method.

Differential effects of size-specific particulate matter on lower respiratory infections in children: A multi-city time-series analysis in Sichuan, China.

Evidence on the short-term effects of size-specific particulate matter with aerodynamic diameter ≤2.5 µm (PM2.5), ≤10 µm (PM10), and their difference (PMC) on children's Lower Respiratory Infections (LRI) is scare. This study aimed to estimate the differential effects of three size-specific PM on hospitalizations of children aged <18 years for pneumonia and bronchitis in 18 cities of southwestern China. The city-specific association was firstly estimated using the over-dispersed generalized additive model and then combined to obtain the regional average association. Further, to evaluate the robustness of the key findings, subgroup analyses and co-pollutant models were constructed. PM-related risks of LRI differed by PM fractions and cause-specific LRI. A 10 µg/m3 increment in PM2.5_lag03, PM10_lag06, and PMC_lag06 was associated with a 0.79% (95% CI: 0.29%, 1.29%), 0.77% (95% CI: 0.13%, 1.41%), and 2.33% (95% CI: 1.23%, 3.44%) increase in children's LRI hospitalizations, respectively. After adjustment for gaseous pollutants, adverse effects of the three types of size-specific PM on pneumonia hospitalizations were stable, ranging from 0.29% (95% CI: 0.05%, 0.54%) for PM2.5-2.50% (95% CI: 1.38%, 3.64%) for PMC. Additionally, PMC-related risk of bronchitis hospitalizations remained stable after adjustment for gaseous pollutants. Associations of pneumonia with PMC and PM10 in infants, bronchitis with PM2.5 in children aged 6-17 years, pneumonia and bronchitis with PM2.5, PMC, and PM10 in children aged 1-5 years were all statistical significant. Specifically, the effects of PM2.5 on LRI hospitalizations increased by age, with the highest effect of 1.72% (95%CI: 1.01%, 2.43%) in children aged 6-17 years. Our study provided evidence for short-term effects of different PM fractions on children LRI hospitalizations in Southwestern China, which will be useful for making and promoting policies on air quality standards in order to protect children's health.

Targeted and Combined TPCA-1-Gold Nanocage Therapy for In Vivo Treatment of Inflammatory Arthritis.

Rheumatoid arthritis (RA) is an autoimmune disease that is currently incurable. Inhibition of inflammation can prevent the deterioration of RA. 2-[(Aminocarbonyl)amino]-5-(4-fluorophenyl)-3-thiophenecarboxamide (TPCA-1) suppresses inflammation via the inhibition of nuclear factor-κ (NF-κB) signaling pathway. Gold-based therapies have been used to treat inflammatory arthritis since the 1940s. Hyaluronic acid (HA) is a targeting ligand for CD44 receptors overexpressed on activated macrophages. Therefore, a combined therapy based on TPCA-1, gold, and HA was explored for the treatment of RA in this study. We used gold nanocages (AuNCs) to load TPCA-1 and modified the TPCA-1 (T) loaded AuNCs with HA and peptides (P) to construct an anti-inflammatory nanoparticle (HA-AuNCs/T/P). An adjuvant-induced arthritis (AIA) mice model was used to investigate the in vivo anti-inflammatory efficacy of HA-AuNCs/T/P. In vivo distribution results showed that HA-AuNCs/T/P had increased and prolonged accumulation at the inflamed paws of AIA mice. Treatment by the HA-AuNCs/T/P suppressed joint swelling and alleviated cartilage and bone damage. By loading to HA-AuNCs/T/P, the effective concentration of TPCA-1 was greatly reduced from 20 to 0.016 mg/kg mice. This study demonstrated that HA-AuNCs/T/P could effectively suppress inflammation and alleviate the symptoms of AIA mice, suggesting a great potential of HA-AuNCs/T/P for the treatment of RA.

[Machine learning-based method for interpreting the guidelines of the diagnosis and treatment of COVID-19].

The outbreak of pneumonia caused by novel coronavirus (COVID-19) at the end of 2019 was a major public health emergency in human history. In a short period of time, Chinese medical workers have experienced the gradual understanding, evidence accumulation and clinical practice of the unknown virus. So far, National Health Commission of the People's Republic of China has issued seven trial versions of the "Guidelines for the Diagnosis and Treatment of COVID-19". However, it is difficult for clinicians and laymen to quickly and accurately distinguish the similarities and differences among the different versions and locate the key points of the new version. This paper reports a computer-aided intelligent analysis method based on machine learning, which can automatically analyze the similarities and differences of different treatment plans, present the focus of the new version to doctors, reduce the difficulty in interpreting the "diagnosis and treatment plan" for the professional, and help the general public better understand the professional knowledge of medicine. Experimental results show that this method can achieve the topic prediction and matching of the new version of the program text through unsupervised learning of the previous versions of the program topic with an accuracy of 100%. It enables the computer interpretation of "diagnosis and treatment plan" automatically and intelligently.

Synergistic Treatment of Tumor by Targeted Biotherapy and Chemotherapy via Site-Specific Anchoring of Aptamers on DNA Nanotubes.

BACKGROUND: Aptamers have been widely used as targeted therapeutic agents due to its relatively small physical size, flexible structure, high specificity, and selectivity. Aptamers functionalized nanomaterials, not only enhance the targeting of nanomaterials, but can also improve the stability of the aptamers. We developed aptamer C2NP (Apt) conjugated straight DNA nanotubes (S-DNT-Apt) and twisted DNA nanotubes (T-DNT-Apt) as nanocarriers for doxorubicin (DOX). METHODS: The twisted DNA nanotubes (T-DNT) and straight DNA nanotubes (S-DNT) were assembled with a scaffold and hundreds of staples. Apt was site-specifically anchored on DNA nanotubes with either different spatial distribution (3 or 6 nm) or varied stoichiometry (15Apt or 30Apt). The developed nanocarriers were characterized with agarose gel electrophoresis and transmission electron microscopy. The drug loading and release in vitro were evaluated by measuring the fluorescence intensity of DOX using a microplate reader. The stability of DNT in cell culture medium plus 10% of FBS was evaluated by agarose gel electrophoresis. The cytotoxicity of DNA nanostructures against K299 cells was tested with a standard CCK8 method. Cellular uptake, cell apoptosis, cell cycle and reactive oxygen species level were investigated by flow cytometry. The expression of p53 was examined by Western Blot. RESULTS: T-DNT-30Apt-6 exhibited the highest cytotoxicity when the concentration of Apt was 120 nM. After intercalation of DOX, the cytotoxicity of DOX@T-DNT-30Apt-6 was further enhanced due to the combination of chemotherapy of DOX and biotherapy of Apt. The enhanced cytotoxicity of DOX@T-DNT-30Apt-6 can be explained by the increase in the cellular uptake, cell apoptosis and intracellular ROS levels. Additionally, the interaction between Apt and its receptor CD30 could upregulate the expression of p53. CONCLUSION: These results demonstrate that both stoichiometry and spatial arrangement of Apt on T-DNT-Apt influence the anticancer activity. The developed twisted DNA nanotubes may be a solution for the synergistic treatment of cancer.

Attributable risk of hospital admissions for overall and specific mental disorders due to particulate matter pollution: A time-series study in Chengdu, China.

The associations of particulate matter (PM) pollution with the morbidity of overall and subtypes of mental disorders (MDs), as well as the corresponding morbidity burden, remain understudied, especially in developing countries. This study aimed to evaluate the short-term effects of PM2.5 (diameters ≤ 2.5 µm), PM10 (diameters ≤ 10 µm) and PMC (diameters between 2.5 and 10 µm) on hospital admissions (HAs) for MDs in Chengdu, China, during 2015-2016, and calculate corresponding attributable risks. A generalized additive model (GAM) with controlling for time trend, meteorological conditions, holidays and day of the week was used to estimate the associations. Stratified analyses were also performed by age, gender and season. We further estimated the burden of HAs for MDs attributable to PM exposure. During the study period, a total of 10,947 HAs for MDs were collected. PM2.5, PM10 and PMC were significantly associated with elevated risks of MDs hospitalizations. Each 10 µg/m3 increase in PM2.5, PM10 and PMC at lag06 corresponded to an increase of 2.89% (95% CI: 0.75-5.08%), 1.91% (95% CI: 0.57-3.28%) and 3.95% (95% CI: 0.84-7.15%) in daily HAs for MDs, respectively. The risk estimates of PM on MDs hospitalizations were generally robust after adjustment for gaseous pollutants in two-pollutant models. We found stronger associations between PM pollution and MDs in males and in cool seasons than in females and in warm seasons. For specific subtypes of MDs, significant associations of PM pollution with dementia,schizophrenia and depression were observed. Using WHO's air quality guidelines as the reference concentrations, 9.53% (95% CI: 2.67-15.58%), 9.17% (95% CI: 2.91-14.70%) and 6.10% (95% CI: 1.40-10.32%) of HAs for MDs could be attributable to PM2.5, PM10 and PMC, respectively. Our results suggested that PM exposure might be an important trigger of hospitalizations for MDs in Chengdu, China, and account for substantial morbidity burden.

The burden of overall and cause-specific respiratory morbidity due to ambient air pollution in Sichuan Basin, China: A multi-city time-series analysis.

Few studies have investigated the respiratory morbidity burden due to ambient air pollution in China, especially in a multi-city setting. This study aimed to estimate the short-term effects of ambient air pollutants (PM10, PM2.5, NO2 and SO2) on hospital admissions (HAs) for overall and cause-specific respiratory diseases, as well as the associated burden in 17 cities of Sichuan Basin, China during 2015-2016. Firstly, city-specific effect estimates for each pollutant on respiratory HAs were obtained using generalized additive model with quasi-Poisson link, and then random- or fixed-effects meta-analysis was applied to pool the effect estimates at the regional level. Subgroup analyses by sex, age, season and region were also performed. A total of 757,712 respiratory HAs were collected from all the tertiary and secondary hospitals located in the 17 cities. Risks of HAs for overall and cause-specific respiratory diseases were elevated following increased PM10, PM2.5, NO2 and SO2 exposure. An increase of 10 µg/m3 in PM10 at lag01, PM2.5 at lag01, NO2 at lag0 and SO2 at lag02 was associated with a 0.43% (95% CI: 0.33%, 0.53%), 0.53% (95% CI: 0.39%, 0.68%), 2.36% (95% CI: 1.75%, 2.98%) and 2.54% (95% CI: 1.51%, 3.59%) increases in total respiratory HAs, respectively. Children (≤ 14 years) and elderly (≥ 65 years) appeared to be more vulnerable to the effects of ambient air pollutants. Comparing to the WHO's air quality guidelines, we estimated that 1.84% (95%CI: 1.42%, 2.25%), 1.73% (95%CI: 1.27%, 2.19%) and 0.34% (95%CI: 0.21%, 0.48%) of respiratory HAs were due to PM10, PM2.5 and SO2 exposure, respectively. This study suggests that air pollution might be an important trigger of respiratory admissions, and result in substantial burden of HAs for respiratory diseases in Sichuan Basin.

Manifold-based learning and synthesis.

This paper proposes a new approach to analyze high-dimensional data set using low-dimensional manifold. This manifold-based approach provides a unified formulation for both learning from and synthesis back to the input space. The manifold learning method desires to solve two problems in many existing algorithms. The first problem is the local manifold distortion caused by the cost averaging of the global cost optimization during the manifold learning. The second problem results from the unit variance constraint generally used in those spectral embedding methods where global metric information is lost. For the out-of-sample data points, the proposed approach gives simple solutions to transverse between the input space and the feature space. In addition, this method can be used to estimate the underlying dimension and is robust to the number of neighbors. Experiments on both low-dimensional data and real image data are performed to illustrate the theory.

Binary fingerprint image thinning using template-based PCNNs.

This correspondence presents a coarse-to-fine binary-image-thinning algorithm by proposing a template-based pulse-coupled neural-network model. Under the control of coupled templates, this algorithm iteratively skeletonizes a binary image by changing the load signals of pulse neurons. A direction-constraining scheme for avoiding fingerprint ridge spikes has been discussed. Experiments show that this algorithm is effective for fingerprint thinning, as well as other common images. Moreover, this algorithm can be coupled with a fingerprint identification system to improve the recognition performance.

[Development and clinical trial of fluorescence real time PCR to detect rubella virus].

To establish a novel rapid, convenient, sensitive and specific method applicable to quantitative analysis of the rubella virus extensively, RV total RNA was extracted with Trizol. The envelope glycoprotein E1 gene was amplified from rubella virus by PCR, and the PCR products were cloned into the pMD18-T cloning vector and transfected into DH5alpha. After Amp selection and analysis of restriction enzyme, the clones carrying the E1 gene were identified. After quantitation and serial dilution, the quantitative analysis of E1 gene was made by real-time PCR with the use of FAM as indicator. Standard curve of the real-time PCR was plotted with starting cDNA concentration versus threshold cycle. Then the new method was used to measure 50 cases with suspectable RV infection. The results were compared with those obtained by ELISA assay. TaqMan(r)MGB real-time PCR could help evaluate the level of virus reliably. The correlation coefficient of the standard curve is 0.998, and the linear range of the system is from 10(3) copies/microl to 10(9) copies/microl in clinical samples. The CV value is 0.94% in batch assay and 3.36% in day to day assay. The new method is more sensitive and specific than ELISA assay. For its simplicity, sensitivity, specificity and digitized results, the real-time PCR for quantification of RV cDNA in clinical samples is available.