Search

Considering the poor effect of short text classification due to insufficient semantic information mining in the current short text matching methods, a new short text classification method is proposed based on explicit and implicit multiscale weighting semantic information interaction. First, the explicit and implicit representations of short text are obtained by a word vector model (word2vec), convolutional neural networks (CNNs), and long short-term memory (LSTM). Then, a multiscale convolutional neural network obtains the explicit and implicit multiscale weighting semantics information of short text. Finally, the multiscale weighting semantics is fused for more accurate short text classification. The experimental results show that this method is superior to the existing classical short text classification algorithms and two advanced short text classification models on the five short text classification datasets of MR, Subj, TREC, SST1 and SST2 with accuracies of 85.7%, 96.9%, 98.1%, 53.4% and 91.8%, respectively. [ABSTRACT FROM AUTHOR]

Effective contamination warning and control of food quality can significantly reduce the likelihood of food quality safety incidents. Existing food contamination warning models for food quality rely on supervised learning, do not model the complex feature associations between detection samples, and do not consider the unevenness of detection data categories. In this paper, To overcome these limitations, we propose a Contrastive Self-supervised learning-based Graph Neural Network framework (CSGNN) for contamination warning of food quality. Specifically, we structure the graph for detecting correlations between samples and then define the positive and negative instance pairs for contrastive learning based on attribute networks. Further, we use a self-supervised approach to capture the complex relationships between detection samples. Finally, we assessed each sample's contamination level based on the absolute value of the subtraction of the prediction scores from multiple rounds of positive and negative instances obtained by the CSGNN. Moreover, we conducted a sample study on a batch of dairy product detection data in a Chinese province. The experimental results show that CSGNN outperforms other baseline models in contamination assessment of food quality, with AUC and recall of unqualified samples reaching 0.9188 and 1.0000, respectively. Meanwhile, our framework provides interpretable contamination classification for food detection. This study provides an efficient early warning method with precise and hierarchical contamination classification for contamination warning of food quality work. [ABSTRACT FROM AUTHOR]

Simple Summary: Guanine nucleotide-binding protein-like 3-like (GNL3L) is a novel GTP-binding nucleolar protein. In this study, we analyzed the expression, prognosis, and immune roles of GNL3L in pan-cancer from multiple omics analyses. The final results showed that GNL3L is differentially expressed in a variety of cancers, plays a prognostic role, and has good immune value. Moreover, GNL3L may affect the occurrence of cancer through processes such as ribonucleoprotein, ribosomal RNA processing, and cell proliferation. At the same time, we established an esophageal cancer (ESCA) prediction model with strong predictive ability and proved that GNL3L can significantly affect the proliferation ability of esophageal cancer cells through clone formation assays. In conclusion, GNL3L is an important biomarker. Guanine nucleotide-binding protein-like 3-like protein (GNL3L) is a novel, evolutionarily conserved, GTP-binding nucleolar protein. This study aimed to investigate the expression, prognosis, and immune value of GNL3L in pan-cancer from multiple omics analyses. Firstly, the expression and prognostic value of GNL3L in pan-cancer were discussed using the TIMER2 database, the GEPIA database, the cBioportal database, COX regression analysis, and enrichment analysis. The association of GNL3L with tumor mutational burden (TMB), tumor microsatellite instability (MSI), mismatch repair (MMR) genes, and immune cells was then analyzed. Finally, an esophageal cancer (ESCA) prediction model was established, and GNL3L clone formation assays were performed. The final results showed that GNL3L is differentially expressed in the vast majority of cancers, is associated with the prognosis of various cancers, and may affect cancer occurrence through processes such as ribonucleoprotein, ribosomal RNA processing, and cell proliferation. At the same time, it was found that the correlation between GNL3L and TMB, MSI, MMR, and various immune cells is significant. The established ESCA prediction model had a strong predictive ability, and GNL3L could significantly affect the proliferation of esophageal cancer cells. In conclusion, GNL3L may serve as an important prognostic biomarker and play an immunomodulatory role in tumors. [ABSTRACT FROM AUTHOR]

To improve the detection sensitivity of a porous silicon optical biosensor in the real-time detection of biomolecules, a non-spectral porous silicon optical biosensor technology, based on dual-signal light detection, is proposed. Double-light detection is a combination of refractive index change detection and fluorescence change detection. It uses quantum dots to label probe molecules to detect target molecules. In the double-signal-light detection method, the first detection-signal light is the detection light that is reflected from the surface of the porous silicon Bragg mirror. The wavelength of the detection light is the same as the wavelength of the photonic band gap edge of the porous silicon Bragg mirror. CdSe/ZnS quantum dots are used to label the probe DNA and hybridize it with the target DNA molecules in the pores of porous silicon to improve its effective refractive index and enhance the detection-reflection light. The second detection-signal light is fluorescence, which is generated by the quantum dots in the reactant that are excited by light of a certain wavelength. The Bragg mirror structure further enhances the fluorescence signal. A digital microscope is used to simultaneously receive the digital image of two kinds of signal light superimposed on the surface of porous silicon, and the corresponding algorithm is used to calculate the change in the average grey value before and after the hybridization reaction to calculate the concentration of the DNA molecules. The detection limit of the DNA molecules was 0.42 pM. This method can not only detect target DNA by hybridization, but also detect antigen by immune reaction or parallel biochip detection for a porous silicon biosensor. [ABSTRACT FROM AUTHOR]

Food safety is a high-priority issue for all countries. Early warning analysis and risk control are essential for food safety management practices. This paper innovatively proposes an anomaly score-based risk early warning system (ASRWS) via an unsupervised auto-encoder (AE) for the effective early warning of detection products, which classifies qualified and unqualified products by reconstructing errors. The early warning analysis of qualified samples is carried out by early warning thresholds. The proposed method is applied to a batch of dairy product testing data from a Chinese province. Extensive experimental results show that the unsupervised anomaly detection model AE can effectively analyze the dairy product testing data, with a prediction accuracy and fault detection rate of 0.9954 and 0.9024, respectively, within only 0.54 s. We provided an early warning threshold-based method to conduct the risk analysis, and then a panel of food safety experts performed a risk revision on the prediction results produced by the proposed method. In this way, AI improves the panel's efficiency, whereas the panel enhances the model's reliability. This study provides a fast and cost-effective, food safety early warning method for detection data and assists market supervision departments in controlling food safety risk. [ABSTRACT FROM AUTHOR]

In this paper, carbon quantum dot-labelled β-lactoglobulin antibodies were used for refractive index magnification, and β-lactoglobulin was detected by angle spectroscopy. In this method, the detection light is provided by a He-Ne laser whose central wavelength is the same as that of the porous silicon microcavity device, and the light source was changed to a parallel beam to illuminate the porous silicon microcavity' surface by collimating beam expansion, and the reflected light was received on the porous silicon microcavity' surface by a detector. The angle corresponding to the smallest luminous intensity before and after the onset of immune response was measured by a detector for different concentrations of β-lactoglobulin antigen and carbon quantum dot-labelled β-lactoglobulin antibodies, and the relationship between the variation in angle before and after the immune response was obtained for different concentrations of the β-lactoglobulin antigen. The results of the experiment present that the angle variations changed linearly with increasing β-lactoglobulin antigen concentration before and after the immune response. The limit of detection of β-lactoglobulin by this method was 0.73 μg/L, indicating that the method can be used to detect β-lactoglobulin quickly and conveniently at low cost. [ABSTRACT FROM AUTHOR]

Cervical cancer is the fourth most common cancer in the world. Whole-slide images (WSIs) are an important standard for the diagnosis of cervical cancer. Missed diagnoses and misdiagnoses often occur due to the high similarity in pathological cervical images, the large number of readings, the long reading time, and the insufficient experience levels of pathologists. Existing models have insufficient feature extraction and representation capabilities, and they suffer from insufficient pathological classification. Therefore, this work first designs an image processing algorithm for data augmentation. Second, the deep convolutional features are extracted by fine-tuning pre-trained deep network models, including ResNet50 v2, DenseNet121, Inception v3, VGGNet19, and Inception-ResNet, and then local binary patterns and a histogram of the oriented gradient to extract traditional image features are used. Third, the features extracted by the fine-tuned models are serially fused according to the feature representation ability parameters and the accuracy of multiple experiments proposed in this paper, and spectral embedding is used for dimension reduction. Finally, the fused features are inputted into the Analysis of Variance-F value-Spectral Embedding Net (AF-SENet) for classification. There are four different pathological images of the dataset: normal, low-grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion (HSIL), and cancer. The dataset is divided into a training set (90%) and a test set (10%). The serial fusion effect of the deep features extracted by Resnet50v2 and DenseNet121 (C5) is the best, with average classification accuracy reaching 95.33%, which is 1.07% higher than ResNet50 v2 and 1.05% higher than DenseNet121. The recognition ability is significantly improved, especially in LSIL, reaching 90.89%, which is 2.88% higher than ResNet50 v2 and 2.1% higher than DenseNet121. Thus, this method significantly improves the accuracy and generalization ability of pathological cervical WSI recognition by fusing deep features. [ABSTRACT FROM AUTHOR]

Action recognition algorithms are widely used in the fields of medical health and pedestrian dead reckoning (PDR). The classification and recognition of non-normal walking actions and normal walking actions are very important for improving the accuracy of medical health indicators and PDR steps. Existing motion recognition algorithms focus on the recognition of normal walking actions, and the recognition of non-normal walking actions common to daily life is incomplete or inaccurate, resulting in a low overall recognition accuracy. This paper proposes a microelectromechanical system (MEMS) action recognition method based on Relief-F feature selection and relief-bagging-support vector machine (SVM). Feature selection using the Relief-F algorithm reduces the dimensions by 16 and reduces the optimization time by an average of 9.55 s. Experiments show that the improved algorithm for identifying non-normal walking actions has an accuracy of 96.63%; compared with Decision Tree (DT), it increased by 11.63%; compared with k-nearest neighbor (KNN), it increased by 26.62%; and compared with random forest (RF), it increased by 11.63%. The average Area Under Curve (AUC) of the improved algorithm improved by 0.1143 compared to KNN, by 0.0235 compared to DT, and by 0.04 compared to RF. [ABSTRACT FROM AUTHOR]

To improve the detection sensitivity of porous silicon microcavity biosensors, CdSe/ZnS quantum dots are used to label complementary DNA molecules for the refractive index amplification and angular spectrum method for detection. In this method, the TE mode laser is used as the detection light and the light source is changed into a parallel beam by collimating and expanding the beam, which illuminates the PSM surface and receives the reflected light from the PSM surface through the detector. The angle corresponding to the weakest reflected light intensity before and after the biological reaction between probe DNA and complementary DNA of different concentrations labeled by quantum dots was measured by the detector, and the relationship between the angle change before and after the biological reaction and the complementary DNA concentration labeled by quantum dots was obtained. The experimental results show that the angle change increases linearly with increasing complementary DNA concentration. The detection limit of the experiment, as determined by fitting, is approximately 36 pM. The detection limit of this method is approximately 1/300 of that without quantum dot labeling. Our method has a low cost because it does not require the use of a reflectance spectrometer, and it also demonstrates high sensitivity. [ABSTRACT FROM AUTHOR]

Cervical cancer is the fourth most common cancer in the world. Whole-slide images (WSIs) are an important standard for the diagnosis of cervical cancer. Missed diagnoses and misdiagnoses often occur due to the high similarity in pathological cervical images, the large number of readings, the long reading time, and the insufficient experience levels of pathologists. Existing models have insufficient feature extraction and representation capabilities, and they suffer from insufficient pathological classification. Therefore, this work first designs an image processing algorithm for data augmentation. Second, the deep convolutional features are extracted by fine-tuning pre-trained deep network models, including ResNet50 v2, DenseNet121, Inception v3, VGGNet19, and Inception-ResNet, and then local binary patterns and a histogram of the oriented gradient to extract traditional image features are used. Third, the features extracted by the fine-tuned models are serially fused according to the feature representation ability parameters and the accuracy of multiple experiments proposed in this paper, and spectral embedding is used for dimension reduction. Finally, the fused features are inputted into the Analysis of Variance-F value-Spectral Embedding Net (AF-SENet) for classification. There are four different pathological images of the dataset: normal, low-grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion (HSIL), and cancer. The dataset is divided into a training set (90%) and a test set (10%). The serial fusion effect of the deep features extracted by Resnet50v2 and DenseNet121 () is the best, with average classification accuracy reaching 95.33%, which is 1.07% higher than ResNet50 v2 and 1.05% higher than DenseNet121. The recognition ability is significantly improved, especially in LSIL, reaching 90.89%, which is 2.88% higher than ResNet50 v2 and 2.1% higher than DenseNet121. Thus, this method significantly improves the accuracy and generalization ability of pathological cervical WSI recognition by fusing deep features.