Open-set semi-supervised learning (OSSL) provides a practical solution by filtering out-of-distribution (OOD) samples from unlabeled data to guarantee the reliance on large unlabeled data in semi-supervised setting. However, existing OSSL methods mainly focus on identifying in-distribution (ID) samples and discarding OOD samples, while ignoring to make full use of samples that could not be exactly identified as ID or OOD samples. Those samples are more likely to be hard samples, which should be carefully explored to boost the performance in OSSL task. Hence, in this paper, we propose a novel framework, named Mutual Filter Teaching (MFT), where two networks are trained simultaneously to divide the unlabeled data into three parts: ID samples, OOD samples and hard samples. The samples are regarded as ID or OOD samples only if two networks give consistent decisions according to Mahalanobis distance between the unlabeled samples and their closest class prototypes. For those samples with inconsistent decisions, we treat them as hard samples and design an efficient mutual teaching scheme where the samples detected by only one network as positive samples are fed to its peer network for training. Furthermore, we propose to employ the prediction variance of two networks to dynamically rectify the learning from hard samples. Experiments on multiple benchmark datasets demonstrate that our approach achieves the state-of-the-art performance.

Before the Stereotactic Radiosurgery (SRS) treatment, it is of great clinical significance to avoid secondary genetic damage and guide the personalized treatment plans for patients with brain metastases (BM) by predicting the response to SRS treatment of brain metastatic lesions. Thus, we developed a multi-task learning model termed SRTRP-Net to provide prior knowledge of BM ROI and predict the SRS treatment response of the lesion. In dual-encoder tumor segmentation Network (DTS-Net), two parallel encoders encode the original and mirrored multi-modal MRI images. The differences in the dual-encoder features between foreground and background are enhanced by the symmetrical visual difference block (SVDB). In the bottom layer of the encoder, a transformer is used to extract local contextual features in the spatial and depth dimensions of low-resolution images. Then, the decoder of DTS-Net provides the prior knowledge for predicting the response to SRS treatment by performing BM segmentation. SRS response prediction network (SRP-Net) directly utilizes shared multi-modal MRI features weighted by the signed distance map (SDM) of the masks. The bidirectional multi-dimensional feature fusion module (BMDF) fuses the shared features and the clinical text information features to obtain comprehensive tumor information for characterizing tumors and predicting SRS treatment response. Experiments based on internal and external clinical datasets have shown that SRTRP-Net achieves comparable or better results. We believe that SRTRP-Net can help clinicians accurately develop personalized first-time treatment regimens for BM patients and improve their survival.

Training deep neural networks (DNNs) with noisy labels often leads to poorly generalized models as DNNs tend to memorize the noisy labels in training. Various strategies have been developed for improving sample selection precision and mitigating the noisy label memorization issue. However, most existing works adopt a class-dependent softmax classifier that is vulnerable to noisy labels by entangling the classification of multi-class features. This paper presents a class-independent regularization (CIR) method that can effectively alleviate the negative impact of noisy labels in DNN training. CIR regularizes the class-dependent softmax classifier by introducing multi-binary classifiers each of which takes care of one class only. Thanks to its class-independent nature, CIR is tolerant to noisy labels as misclassification by one binary classifier does not affect others. For effective training of CIR, we design a heterogeneous adaptive co-teaching strategy that forces the class-independent and class-dependent classifiers to focus on sample selection and image classification, respectively, in a cooperative manner. Extensive experiments show that CIR achieves superior performance consistently across multiple benchmarks with both synthetic and real images. Code is available at https://github.com/RumengYi/CIR.

It is crucial to distinguish mislabeled samples for dealing with noisy labels. Previous methods such as "Co-teaching" and "JoCoR" introduce two different networks to select clean samples out of the noisy ones and only use these clean samples to train the deep models. Different from these methods which require to train two networks simultaneously, we propose a simple and effective method to identify clean samples only using one single network. We discover that the clean samples prefer to reach consistent predictions for the original images and the transformed images while noisy samples usually suffer from inconsistent predictions. Motivated by this observation, we propose a noisy label detection approach, named Transform Consistency Network (TC-Net), which constrains the transform consistency (i.e., category consistency and visual attention consistency) between the original images and the transformed images for network training. Then we can select small-loss samples to update the parameters of the network. Furthermore, in order to mitigate the negative influence of noisy labels, we design a classification loss by using the off-line hard labels and on-line soft labels to provide more reliable supervisions for training a robust model. We conduct comprehensive experiments on CIFAR-10, CIFAR-100 and Clothing1M datasets. Compared with the clean sample selection baselines, we achieve the state-of-the-art performance. Especially, in most cases, our proposed method outperforms the baselines by a large margin.

It is crucial to distinguish mislabeled samples for dealing with noisy labels. Previous methods such as Coteaching and JoCoR introduce two different networks to select clean samples out of the noisy ones and only use these clean ones to train the deep models. Different from these methods which require to train two networks simultaneously, we propose a simple and effective method to identify clean samples only using one single network. We discover that the clean samples prefer to reach consistent predictions for the original images and the transformed images while noisy samples usually suffer from inconsistent predictions. Motivated by this observation, we introduce to constrain the transform consistency between the original images and the transformed images for network training, and then select small-loss samples to update the parameters of the network. Furthermore, in order to mitigate the negative influence of noisy labels, we design a classification loss by using the off-line hard labels and on-line soft labels to provide more reliable supervisions for training a robust model. We conduct comprehensive experiments on CIFAR-10, CIFAR-100 and Clothing1M datasets. Compared with the baselines, we achieve the state-of-the-art performance. Especially, in most cases, our proposed method outperforms the baselines by a large margin.

This paper addresses semi-supervised semantic segmentation by exploiting a small set of images with pixel-level annotations (strong supervisions) and a large set of images with only image-level annotations (weak supervisions). Most existing approaches aim to generate accurate pixel-level labels from weak supervisions. However, we observe that those generated labels still inevitably contain noisy labels. Motivated by this observation, we present a novel perspective and formulate this task as a problem of learning with pixel-level label noise. Existing noisy label methods, nevertheless, mainly aim at image-level tasks, which can not capture the relationship between neighboring labels in one image. Therefore, we propose a graph-based label noise detection and correction framework to deal with pixel-level noisy labels. In particular, for the generated pixel-level noisy labels from weak supervisions by Class Activation Map (CAM), we train a clean segmentation model with strong supervisions to detect the clean labels from these noisy labels according to the cross-entropy loss. Then, we adopt a superpixel-based graph to represent the relations of spatial adjacency and semantic similarity between pixels in one image. Finally we correct the noisy labels using a Graph Attention Network (GAT) supervised by detected clean labels. We comprehensively conduct experiments on PASCAL VOC 2012, PASCAL-Context, MS-COCO and Cityscapes datasets. The experimental results show that our proposed semi-supervised method achieves the state-of-the-art performances and even outperforms the fully-supervised models on PASCAL VOC 2012 and MS-COCO datasets in some cases.

A new method of SPR signal amplification based on dynamic field enhancement at the sensor surface to detect C-reactive protein (CRP) is proposed. The gold nanoparticles-detection antibody (AuNPs-dAb) conjugates, as the dynamical agent, were applied to enhance the biosensor sensitivity. The mechanism of SPR sensitivity enhancement is due to the strong evanescent field coupling between the localized SPR (LSPR) of AuNPs and SPR of gold-silver alloy film which proved by COMSOL Multiphysics and MATLAB simulation. The biosensor sensitivity got with sandwich immunoassay (cAb/CRP/AuNPs-dAb) is 5.56-fold of the structure without AuNPs (cAb/CRP/dAb). Moreover, concentration of AuAg-SPR biosensor in CRP detection was lower for an order of magnitude than that of the conventional gold film cladding SPR (Au-SPR) biosensor, according to its direct reaction with the immobilized cAb. The dynamic field enhancement based AuAg-SPR biosensors have high sensitivity in detecting trace biomarker of CRP, and the improvement can also be applied to the detection of other biomarkers in low concentration sensitively.

In this study, we developed a gold‑silver alloy film based surface plasmon resonance (AuAg-SPR) sensor with wavelength interrogation to detect cancer antigen 125 (CA125) using a sandwich immunoassay. We first theoretically simulated the sensitivity of conventional gold film based SPR (Au-SPR) sensor and AuAg-SPR sensor, and conducted a series of experiments to investigate the sensitive characteristics of AuAg-SPR sensor, including the angle and refractive index (RI) sensitivity. We then conducted CA125 detection experiments on these two types of sensors. The results demonstrated that the limit of detection (LOD) of CA125 on the AuAg-SPR sensor was 0.1 U/mL (0.8 ng/mL) based on its direct reaction with an immobilised antibody, which was two orders of magnitude lower than that of the Au-SPR sensor (10 U/mL). The total changes in the resonance wavelength (∆λR) of the former were 1.7-fold those of the latter. The volume fractions of the adsorbates (fad) and effective RIs (nadlayer) in each adlayer were then calculated and the effect of the antibody size on the detection results was analysed. The AuAg-SPR sensors had a higher sensitivity than the conventional Au-SPR sensors for detecting CA125 due to their electric field characteristics. Therefore, these will have better application prospects.

In this work, we have developed a kind of single-layer graphene-based surface plasmon resonance (SLG-SPR) biosensor to detect C-reactive protein (CRP) and Prostate-specific antigen (PSA). In the experiment of testing CPR, the results obtained revealed that the changes in resonance wavelength of SLG-SPR biosensors are higher than that of the gold-film based SPR (Au-SPR) biosensors. Moreover, for the experiment of testing PSA, due to the dynamic evanescent field enhancement produced by a strong electric field coupling between the localized SPR (LSPR) of AuNPs and SPR of single-layer graphene-based film (SLG-film) that further amplify the evanescent field signal. We verified the SLG-SPR biosensors exhibited higher sensitivity than the Au-SPR biosensors and the SLG-SPR biosensor exceeded the traditional biosensor detection limit. Accordingly, the SLG-SPR biosensor based on dynamic optical enhancement can realize high sensitivity detection of low concentration biomarkers and can be applied to most of the trace biomarkers in theory.

In this work, we developed a kind of gold-silver alloy film based surface plasmon resonance (AuAg-SPR) biosensors with wavelength interrogation to detect C-reaction protein (CRP) by using gold nanoparticles (AuNPs)-enhanced sandwich immunoassay, and the limit of detection (LOD) of CRP was found to be 5 pg/ml . In conclusion, using the AuNPsenhanced sandwich immunoassay and the SPR chip of AuAg alloy film can detect CRP effectively and reduce the LOD significantly, and this improvement can also be applied for the detection of any biomarkers in low concentration accurately.

In this study, we have developed a kind of single-layer graphene-based surface plasmon resonance (SLG-SPR) biosensor with wavelength interrogation to detect C-reactive protein (CRP) by sandwich immunoassay. For the nature of the biosensor itself, with the decrease of the incidence angle (theta), resonance wavelength (lambda(R)) increased gradually. The refractive index sensitivity (S) and theta are inversely proportional relationship while S and lambda(R) are positively proportional relationship. Then we studied two kinds of sensitivity properties of these two SPR biosensors respectively. The refractive index (RI) sensitivity of the two biosensors is approximately the same and the high sensitivity of SLG-SPR biosensor due to the excellent adsorption properties of graphene. The experiment shows that the changes in resonance wavelength (Delta lambda(R)) of SLG-SPR biosensor are higher than that of the gold film based SPR (Au-SPR) biosensor when the final sandwich structures were formed. Meanwhile, we have applied the Fresnel reflection equations and Bruggeman approximation formula to calculate the volume fractions of adsorbates in different adlayers, and the dispersion curves of the effective RI of adsorbates with different volume fractions of adlayers and the cAb isothermal adsorption curves that conforms to Langmuir equation were obtained. In conclusion, the SLG-SPR biosensor with a higher detection sensitivity than the conventional Au-SPR biosensor in detecting CRP and it has the advantages of low-cost, stability, highly efficient for the clinical determination of CRP levels.

研究性课题的开展是培养创新创业型人才的重要途径,科研平台在研究性课题中发挥着重要的作用.为了缓解科研平台在实践教学过程中资源不足的矛盾,本文设计了一种基于教研一体化的演示平台,可支持学生开展包括综合实验、创新训练等多种模式的实践项目.该平台运行模式灵活、可扩展性好、资源利用率高、教学成效良好.