Glaucoma is a leading cause of irreversible blindness globally. When glaucoma is diagnosed, Anterior Chamber Angle (ACA) evaluation is the necessary step for the prognosis and treatment of glaucoma. However, current clinical evaluation methods are labor intensive and rely on expert judgment, which makes them inefficient. Automating ACA classification based on images using machine learning, especially deep neural networks, holds promise. Yet, image samples alone can't provide sufficient high-level semantic information on ACA, resulting in suboptimal classification performance. This paper proposes a novel end-to-end knowledge-guided bi-modal network (KGNet) for ACA evaluation. Specifically, we consider two modalities of ACA data: textual domain knowledge and images. We first design a new strategy to refine class-based knowledge into textual descriptions, thereby increasing the diversity of features learned by the model. We then extract two types of representations using two distinct components: 1) a supervised loss is applied to learn modality-specific representations by incorporating domain knowledge; 2) a fusion module that uses knowledge-guided learning to highlight key clinical structures in ACA images leveraging bimodal correlations. Experimental results on an ACA dataset and three public datasets show that our method outperforms several state-of-the-art deep learning models in eye image evaluation, indicating the potential medical interest of our method. Furthermore, our approach improves interpretability by explicitly aligning visual representations with structured clinical knowledge, enabling more structured and clinically grounded explanations than conventional models.

Retinopathy of prematurity (ROP) is a potentially blinding retinal disease that often affects low birth weight premature infants. Lesion detection and recognition are crucial for ROP diagnosis and clinical treatment. However, this task poses challenges for both ophthalmologists and computer-based systems due to the small size and subtle nature of many ROP lesions. To address these challenges, we present a Sequence encoding and Block balancing-based Segmentation Network (SeBSNet), which incorporates domain knowledge coding, sequence coding learning (SCL), and block-weighted balancing (BWB) techniques into the segmentation of ROP lesions. The experimental results demonstrate that SeBSNet outperforms existing state-of-the-art methods in the segmentation of ROP lesions, with average ROC_AUC, PR_AUC, and Dice scores of 98.84%, 71.90%, and 66.88%, respectively. Furthermore, the integration of the proposed techniques into ROP classification networks as an enhancing module leads to considerable improvements in classification performance.

As machine learning and deep learning gain increasing prominence across diverse domains, the realms of intelligent education and computer-assisted educational technologies are constantly converging, facilitating the development of innovative pedagogical approaches. Simultaneously, with the escalating emphasis on personalized education, traditional test scores alone are no longer sufficient to comprehensively reflect students’ learning status and progress. Cognitive diagnosis aims to uncover the latent cognitive information and skills underlying students’ performance scores. By tapping into the educational data hidden beyond scores, it is possible to gain a deeper and more nuanced understanding of students’ learning profiles and abilities. Consequently, this paper proposes a novel cognitive diagnosis system based on Broad Learning System (BLS). Within this system, an attention residual block is designed to efficiently extract information from interrelated modules, taking into full consideration their interdependencies and correlations. To validate and assess the accuracy of the extracted diagnostic information as well as the rationality and effectiveness of the proposed model, student performance is predicted utilizing the Broad Learning System. Experimental results demonstrate that the proposed system exhibits excellent performance in both the extraction of latent cognitive information and the prediction of student performance.

Alzheimer's disease is a progressive neurological disorder. The disease is not reversible, but mild cognitive impairment is a transitional state between Alzheimer's disease and cognitively normal, it is a stage in which the brain is so minimally diseased that it can be treated or slowed or prevented from developing further lesions. At present, there are many methods for the aided diagnosis of mild cognitive impairment, among which the method based on fMRI medical images has emerged in recent years. However, the current aided diagnosis methods in this direction have some problems such as low accuracy and cumbersome feature extraction. A novel feature extraction method is put forward, namely the mean time series difference method in this paper. The feature extraction method effectively improves the accuracy in the auxiliary diagnosis task of mild cognitive impairment. An enhanced broad learning system with mean time series difference also was proposed. The results show that the enhanced broad learning system with mean time series can not only effectively optimize the feature extraction process, but also the accuracy of mild cognitive impairment classification tasks., which has significance for the clinical auxiliary diagnosis of mild cognitive impairment.

目的:利用人工智能神经网络方式构建ICF康复组合(ICF-RS)评定量化标准总体及三个维度(身体功能、活动、参与)功能分级的算法模型,为应用ICF-RS评定量化标准进行数据分析及功能分级提供解决方案.方法:本研究利用中文版ICF-RS评定量化标准,通过多中心合作,采用分层比例抽样的方法收集了 6家已开展ICF-RS评定量化标准临床应用的康复医学科住院患者ICF-RS数据,以多个专家对同一患者的方式获取ICF-RS评定量化标准三个维度及整体功能状况的等级评价结果.借助于神经网络算法构建ICF-RS评定量化标准的各维度及整体功能分级模型,采用计算机k折交叉验证法选择最优模型参数,并计算模型预测受试者工作特征曲线[receiver operating characteristic(ROC)curve]下面积(area under ROC curve,AUC)、准确率(accuracy,ACC)和F1 分数(F1-score).模型构建后,再通过专家和患者一对一的方式收集100例ICF-RS数据用于对已建立的模型进行临床再测试,通过计算ROC、AUC、ACC和F1对模型性能进行评价.结果:共收集584例住院患者的ICF-RS数据,其中484例数据用于构建及验证模型,100例数据用于测试模型的预测性能.根据k折交叉验证法结果显示,身体功能维度、活动维度、参与维度及整体ICF-RS功能分级模型的AUC分别是 89.00％、92.00％、87.00％和87.00％,ACC分别达到75.19％、78.10％、72.91％和73.53％,F1分别是73.68％、77.04％％、69.28％、58.95％.在模型建立后将重新收集到的100例ICF-RS数据输入模型计算,发现各模型ROC曲线良好,AUC 分别是 89.04％、91.81％、86.85％、86.89％,ACC 分别是 64.00％、72.00％、61.00％、65.00％,F1 分别是 48.30％、59.95％、64.06％、49.35％.结论:基于神经网络建立的ICF-RS整体及各维度功能分级算法模型对ICF-RS数据的功能等级预测准确率良好,预测价值较高,具有良好的临床应用价值.

Histopathological tissue classification is a simpler way to achieve semantic segmentation for the whole slide images, which can alleviate the requirement of pixel-level dense annotations. Existing works mostly leverage the popular CNN classification backbones in computer vision to achieve histopathological tissue classification. In this paper, we propose a super lightweight plug-and-play module, named Pyramidal Deep-Broad Learning (PDBL), for any well-trained classification backbone to improve the classification performance without a re-training burden. For each patch, we construct a multi-resolution image pyramid to obtain the pyramidal contextual information. For each level in the pyramid, we extract the multi-scale deep-broad features by our proposed Deep-Broad block (DB-block). We equip PDBL in three popular classification backbones, ShuffLeNetV2, EfficientNetb0, and ResNet50 to evaluate the effectiveness and efficiency of our proposed module on two datasets (Kather Multiclass Dataset and the LC25000 Dataset). Experimental results demonstrate the proposed PDBL can steadily improve the tissue-level classification performance for any CNN backbones, especially for the lightweight models when given a small among of training samples (less than 10%). It greatly saves the computational resources and annotation efforts. The source code is available at: https://github.com/linjiatai/PDBL.

As the basis for developing glaucoma treatment strategies, Anterior Chamber Angle (ACA) evaluation is usually dependent on experts' Judgements. However, experienced ophthalmologists needed for these Judgements are not widely available. Thus, computer-aided ACA evaluations become a pressing and efficient solution for this issue. In this paper, we propose a novel end-to-end frame-work GCNet for automated Glaucoma Classification based on ACA images or other Glaucoma-related medical images. We first collect and label an ACA image dataset with some pixel-level annotations. Next, we introduce a segmentation module and an embedding module to enhance the performance of classifying ACA images. Within GCNet, we design a Cross-Module Aggregation Net (CMANet) which is a weakly-supervised metric learning network to capture contextual information exchanging across these modules. We conduct experiments on the ACA dataset and two public datasets REFUGE and SIGF. Our experimental results demonstrate that GCNet outperforms several state-of-the-art deep models in the tasks of glaucoma medical image classifications. The source code of GCNet can be found at https://github.com/Jingqi-H/GCNet.

Small data means that the number of samples in data is small. In such case, their features are often hand-crafted so that they may be insufficient and inconsistent. Currently some feature augmentation methods are proposed to solve the problem. They are mainly for the larger data composed of images and videos, instead of small data. In the case of small data, the human cognition such as cognitive relativity can be more beneficial to machine learning, as humans routinely classify objects according to both their individual features and their environments. The relative transformation is an efficient way to formalize the cognitive relativity and has been validated in improving the performance of the machine learning. Although relationships among categorical central vectors of small data can greatly improve the performance, they have not been considered as features in current methods. This paper uses the relative transformation to model these relationships and then proposes a new feature augmentation method. These relationships as features are learned automatically for each sample through the neural network, instead of determined by the predefined rules. As it only uses the information of small data itself, without requiring any additional knowledge, it can be applied to any data with features. Lots of experimental results on small data validate the proposed method.

Image attack can disable a deep neural network in image classification by adding a small amount of interference to the input image.However,most of the existing image attack algorithms are relatively fragile against Variational Auto-Encoder(VAE) defense.Based on the AdvGAN algorithm,an algorithm named AntiVAEGAN is proposed,which employs the adversary mechanism to penetrate VAE defenses persistent,but AntiVAEGAN suffers from a loss of attack performance when improving the defense performance.To address the problem,this paper proposes an improved image attack algorithm,VAEAdvGAN,by combining both the generator-discriminator defense mechanism and the generator-VAE defense mechanism.Experimental results on the MNIST dataset and GTSRB dataset show that without defense,AntiVAEGAN and VAEAdvGAN can achieve almost the same classification accuracy and attack success rate as AdvGAN.In the case of VAE defense,VAEAdvGAN exhibits better overall attack effect than AdvGAN and AntiVAEGAN.

在眼底图像分析中,视盘分割对于糖尿病视网膜病变、青光眼等眼部疾病的计算机辅助诊断具有重要意义.然而,由于眼底图像数据集存在样本容量有限、数据正负样本不均衡及视盘边缘受血管遮挡严重等问题,这给视盘分割带来了极大挑战.针对以上问题,本文提出一种融合迁移学习和特征金字塔网络的视盘分割模型TriNet,通过迁移学习缓解数据集不足带来的影响;通过结合特征金字塔网络,利用多尺度特征图,提高对视盘的识别率;通过使用Focal Loss损失函数,动态改变正负样本权重,在正负样本失衡的情况下提高网络对稀少样本的识别.在Baidu Research Open-Access Dataset中的iChallenge-AMD项目的 眼底图像数据集上进行的仿真结果表明,视盘分割的IoU和DICE精度分别达到了93.48％和96.59％.

针对基于孪生网络的目标跟踪算法易受物体形变等因素影响而导致跟踪任务失败的问题,提出一种多层特征融合方法.通过联合多层网络的特征图组成特征空间来丰富模型提取的特征信息;引入无锚点方法,在预测物体位置的分支上,通过预测中心点与边框四周的距离,此部分的网络参数量是基于锚点目标跟踪算法中RPN部分参数量的20％,有效缩短目标跟踪所需要的计算时间.在VOT和GOT数据集上的实验结果显示,此方法在准确度上取得最好成绩.

基于文本的描述编辑图像是图像生成领域的一个重要研究方向,目的是识别出文本中描述的语义关系,改变源图像中与其相关的部分,通过图像生成技术获得新的图像.使用现有的方法得到的人物图像在图像清晰度、人物个性特征一致性以及姿态结构保持等方面有待进一步改进.本文设计了一个编辑人物图像的方法,根据文本描述的语义,通过编辑图像中人物服装的颜色、款式等合成新的人物图像.该模型基于条件式生成对抗网络进行设计,以文本描述、初始人物图像以及高斯噪音作为输入,以多阶段的方式输出由低到高分辨率多输出的人物图像.实验表明,该模型生成的图像在纹理细节、姿态保持等方面优于现有的同类任务模型生成的图像.

针对图像检索中多区域加权聚合算法局部特征提取过程较繁琐、耗时及内存占用高等问题,提出多区域深度特征加权聚合算法(MR-CroW).该算法通过增大深度卷积神经网络(DCNN)最后一个卷积层池化窗口,调整预训练DCNN模型VGG16,得到全局特征提取器;选择子区域作为检索对象响应区域,抑制非主要目标噪声;引入跨维度加权聚合算法(CroW),将多个区域特征进行加权聚合,得到最终的特征向量.在Oxford5K和Paris6K数据集上的实验结果表明,MR-CroW平均准确率(mAP)高于其它几种算法,特征处理阶段时间低于其它几种算法.

医学图像分割是计算机辅助诊断的关键技术.青光眼作为全球第二大致盲眼病,其早期筛查和临床诊断依赖于眼底图的视盘和视杯的准确分割.但传统的视盘和视杯分割方法采用人工构建特征,模型泛化能力差.近年来,基于卷积神经网络的端对端学习模型可通过自动发现特征来分割视盘和视杯,但由于标注样本有限,模型难以训练.提出一个基于半监督条件生成对抗网络的视盘和视杯两阶段分割模型——CDR-GANs.该模型的每个分割阶段均由语义分割网络、生成器和判别器构成,通过对抗学习,判别器引导语义分割网络和生成器学习眼底图及其分割图的联合概率分布.在真实数据集ORIGA上的实验结果表明,CDR-GANs在均交并比(mean intersection over union,简称MIoU)、CDR绝对误差(absolute CDR error)和实际分割效果这些指标上明显优于现有模型.

Water level measure is the key task in hydrological monitoring, wherein water level ruler (WLR) is the necessary enabled measure instrument. Due to its low cost and easy deployment, WLR has been widely used for flood monitoring in many hydrological stations. For saving labor costs, it is very desirable to achieve automatic reading of WLR, which has recently become a focus of the research in hydrological field. However, in automatic reading of WLR, there exist several issues or challenges, which mainly include low recognition rate of characters on WLR and low precision in calculating water level. To address these issues, in this paper we propose a practical and efficient method which is based on image processing. The proposed method consists of three important components, i.e., (1) a multi-template matching algorithm to recognize the characters on WLR, (2) a sequence verification algorithm to check and refine the recognized characters, and (3) a projection height comparison method to achieve accurate reading even under the circumstance of incomplete characters. We conduct experiments on real-world data to verify the efficiency of our proposed method. The experiment results show that the proposed method achieves $$63\%$$ recognition rate of characters on WLR, as well as average measure error of $$\pm {0.90\hbox { cm}}$$ which is much smaller than the national error standards on water-level monitoring in China ( $$\pm {1.0\hbox { cm}}$$ ). Therefore, we believe our proposed method could be useful for facilitating effective water level measure in practice.

Glaucoma is a chronic and widespread eye disease threatening humans' irreversible vision loss. The cup-to-disc ratio (CDR), one of the most important measurements used for glaucoma screening and diagnosis, requires accurate segmentation of optic disc and cup from fundus images. However, most existing techniques fail to obtain satisfactory segmentation performance because a significant number of pixel-level annotated data are often unavailable during training. To cope with this limitation, in this paper, we propose an effective joint optic disc and cup segmentation method based on semi-supervised conditional Generative Adversarial Nets (GANs). Our architecture consists of a segmentation net, a generator and a discriminator, to learn a mapping between the fundus images and the corresponding segmentation maps. Additionally, we employ both labeled and unlabeled data to improve the segmentation performance. The extensive experiments show that our method achieves state-of-the-art optic disc and cup segmentation results on both ORIGA and REFUGE datasets.

移动学习平台是网络信息时代高速发展的产物,其丰富的学习资源和教学功能,是线下传统授课模式的有益补充.提出基于超星学习通线上线下结合的教学模式,并应用到计算机科学技术导论的实践中,激发学生参与教学活动,教学相长.

Sparse coding-based single image super-resolution has attracted much interest. In this paper, a super-resolution reconstruction algorithm based on sparse coding with multi-class dictionaries is put forward. We propose a novel method for image patch classification, using the phase congruency information. A sub-dictionary is learned from patches in each category. For a given image patch, the sub-dictionary that belongs to the same category is selected adaptively. Since the given patch has similar pattern with the selected sub-dictionary, it can be better represented. Finally, iterative back-projection is used to enforce global reconstruction constraint. Experiments demonstrate that our approach can produce comparable or even better super-resolution reconstruction results with some existing algorithms, in both subjective visual quality and numerical measures.

为了快速、有效地分类出敏感视频,提出一种基于人体肤色识别的敏感视频分类方法.首先通过帧差图像提取前景运动区域,然后构建人体肤色模型,并结合前景运动区域识别出视频中的人体,最后用多因素的级联分类器进行敏感视频分类.实验结果分析表明,文中方法有较好的分类准确率和查全率.

镜头边界检测是基于内容的视频检索中的关键技术,提出一种利用TextTiling方法来识别视频镜头边界的算法。通过滑动窗口对视频进行初步切割,利用主成分分析将视频帧投影到特征子空间,并在投影空间上计算相邻帧间距离,再根据相邻窗口之间的深度值确定视频镜头边界。针对TREC-2001视频测试数据集的实验结果显示,该算法检测镜头边界的平均查全率和平均查准率分别为89%和96.5%。