An incremental learning preprocessor for feed-forward neural network

Outpost Vector model synthesizes new vectors from two classes of data at their boundary to maintain the shape of the current system in order to increase the level of accuracy of classification. This paper presents an incremental learning preprocessor for Feed-forward Neural Network (FFNN) which utilizes Outpost Vector model to improve the level of accuracy of classification of both new data and old data. The preprocessor generates outpost vectors from selected new samples, selected prior samples, both samples, or generates no outpost vector at all. After that, they are included in the final training set, as well as selected new samples and selected prior samples, based on the specified parameters. The final training set is then trained with FFNN. The whole process is repeated again when new samples are sufficiently collected in order to learn newer knowledge. The experiments are conducted with a 2-dimension partition problem. The distribution of training and test samples is created in a limited location of a 2-dimension donut ring. The context of the problem is assumed to shift 45° in counterclockwise direction. There are two classes of data which are represented as 0 and 1. Every consecutive partition is set to have different class of both new data and old data. The experimental results show that the use of outpost vectors generated from either selected new samples or selected prior or both samples helps improve the level of accuracy of classification for all data. The run-time complexity of the algorithm presents that the overhead from outpost vector generation process is insignificant and is compensated by the improved level of accuracy of classification.     

A convolutional neural network model for semantic segmentation of mitotic events in microscopy images

Neural Computing and Applications - Mitosis, which has important effects such as healing and growing for human body, has attracted considerable attention in recent years. Especially, cell division...     

An incremental online semi-supervised active learning algorithm based on self-organizing incremental neural network

An incremental online semi-supervised active learning algorithm, which is based on a self-organizing incremental neural network (SOINN), is proposed. This paper describes improvement of the two-layer SOINN to a single-layer SOINN to represent the topological structure of input data and to separate the generated nodes into different groups and subclusters. We then actively label some teacher nodes and use such teacher nodes to label all unlabeled nodes. The proposed method can learn from both labeled and unlabeled samples. It can query the labels of some important samples rather than selecting the labeled samples randomly. It requires neither prior knowledge, such as the number of nodes, nor the number of classes. It can automatically learn the number of nodes and teacher vectors required for a current task. Moreover, it can realize online incremental learning. Experiments using artificial data and real-world data show that the proposed method performs effectively and efficiently.     

Multi-atlas segmentation of optic disc in retinal images via convolutional neural network

Multi-atlas segmentation is widely accepted as an essential image segmentation approach. Through leveraging on the information from the atlases instead of utilizing the model-based segmentation techniques, the multi-atlas segmentation could significantly enhance the accuracy of segmentation. However, label fusion, which plays an important role for multi-atlas segmentation still remains the primary challenge. Bearing this in mind, a deep learning-based approach is presented through integrating feature extraction and label fusion. The proposed deep learning architecture consists of two independent channels composing of continuous convolutional layers. To evaluate the performance our approach, we conducted comparison experiments between state-of-the-art techniques and the proposed approach on publicly available datasets. Experimental results demonstrate that the accuracy of the proposed approach outperforms state-of-the-art techniques both in efficiency and effectiveness.

     

Fuzzy cluster based neural network classifier for classifying breast tumors in ultrasound images

The performance of supervised classification algorithms is highly dependent on the quality of training data. Ambiguous training patterns may misguide the classifier leading to poor classification performance. Further, the manual exploration of class labels is an expensive and time consuming process. An automatic method is needed to identify noisy samples in the training data to improve the decision making process. This article presents a new classification technique by combining an unsupervised learning technique (i.e. fuzzy c-means clustering (FCM)) and supervised learning technique (i.e. back-propagation artificial neural network (BPANN)) to categorize benign and malignant tumors in breast ultrasound images. Unsupervised learning is employed to identify ambiguous examples in the training data. Experiments were conducted on 178 B-mode breast ultrasound images containing 88 benign and 90 malignant cases on MATLAB® software platform. A total of 457 features were extracted from ultrasound images followed by feature selection to determine the most significant features. Accuracy, sensitivity, specificity, area under the receiver operating characteristic curve (AUC) and Mathew's correlation coefficient (MCC) were used to access the performance of different classifiers. The result shows that the proposed approach achieves classification accuracy of 95.862% when all the 457 features were used for classification. However, the accuracy is reduced to 94.138% when only 19 most relevant features selected by multi-criterion feature selection approach were used for classification. The results were discussed in light of some recently reported studies. The empirical results suggest that eliminating doubtful training examples can improve the decision making performance of expert systems. The proposed approach show promising results and need further evaluation in other applications of expert and intelligent systems.     

Segmentation of ultrasound images by using a hybrid neural network

A hybrid neural network is presented for the segmentation of ultrasound images.

Feature vectors are formed by the discrete cosine transform of pixel intensities in region of interest (ROI). The elements and the dimension of the feature vectors are determined by considering only two parameters: The amount of ignored coefficients, and the dimension of the ROI.

First-layer-nodes of the proposed hybrid network represent hyperspheres (HSs) in the feature space. Feature space is partitioned by intersecting these HSs to represent the distribution of classes. The locations and radii of the HSs are found by the genetic algorithms.

Restricted Coulomb energy (RCE) network, modified RCE network, multi-layer perceptron and the proposed hybrid neural network are examined comparatively for the segmentation of ultrasound images.     

A reinforcement agent for object segmentation in ultrasound images

The principal contribution of this work is to design a general framework for an intelligent system to extract one object of interest from ultrasound images. This system is based on reinforcement learning. The input image is divided into several sub-images, and the proposed system finds the appropriate local values for each of them so that it can extract the object of interest. The agent uses some images and their ground-truth (manually segmented) version to learn from. A reward function is employed to measure the similarities between the output and the manually segmented images, and to provide feedback to the agent. The information obtained can be used as valuable knowledge stored in the Q-matrix. The agent can then use this knowledge for new input images. The experimental results for prostate segmentation in trans-rectal ultrasound images show high potential of this approach in the field of ultrasound image segmentation.     

Medical image segmentation with transform and moment based features and incremental supervised neural network

In this study, a novel incremental supervised neural network (ISNN) is proposed for the segmentation of medical images. Performance of the ISNN is investigated for tissue segmentation in medical images obtained from various imaging modalities. Two feature extraction methods based on transform and moments are comparatively investigated to segment the tissues in medical images. Two-dimensional (2D) continuous wavelet transform (CWT) and the moments of the gray-level histogram (MGH) are computed in order to form the feature vectors of ultrasound (US) bladder and phantom images, X-ray computerized tomography (CT) and magnetic resonance (MR) head images. In the 2D-CWT method, feature vectors are formed by the intensity of one pixel of each wavelet-plane of different energy bands. The MGH represents the tissues within the sub-windows by using the spatial variation of image intensities. In this study, the ISNN and Grow and Learn (GAL) network are employed for the segmentation task. It is observed that the ISNN has significantly eliminated the disadvantages of the GAL network in the segmentation of the medical images.     

An incremental-learning neural network for the classification of remote-sensing images

A novel classifier for the analysis of remote-sensing images is proposed. Such a classifier is based on Radial Basis Function (RBF) neural networks and relies on an incremental-learning technique. This technique allows the periodical acquisition of new information whenever a new training set becomes available, while preserving the knowledge learnt by the network on previous training sets. In addition, in each retraining phase, the network architecture is automatically updated so that new classes may be considered. These characteristics make the proposed neural classifier a promising tool for several remote-sensing applications.     

A multilayer incremental neural network architecture for classification

A new multilayer incremental neural network (MINN) architecture and its performance in classification of biomedical images is discussed. The MINN consists of an input layer, two hidden layers and an output layer. The first stage between the input and first hidden layer consists of perceptrons. The number of perceptrons and their weights are determined by defining a fitness function which is maximized by the genetic algorithm (GA). The second stage involves feature vectors which are the codewords obtained automaticaly after learning the first stage. The last stage consists of OR gates which combine the nodes of the second hidden layer representing the same class. The comparative performance results of the MINN and the backpropagation (BP) network indicates that the MINN results in faster learning, much simpler network and equal or better classification performance.     

基于卷积神经网络的超声图像左心室分割方法

超声图像左心室的分割在临床上对医生的作用巨大。由于超声图像含有大量噪声，轮廓特征不明显，目前的卷积神经网络（CNN）方法对左心室分割容易得到不必要的区域，并且分割目标不完整。为了解决上述问题，在全卷积神经网络（FCN）基础上加入了关键点定位和求取图像凸包方法对分割结果进行优化。首先采用FCN获取初步的分割结果；然后为了去除分割结果中的错误区域，提出一种CNN定位左心室三个关键点的位置，通过关键点筛选掉分割结果中不必要的区域；最后为保证剩余区域能够组合成一个完整的心室，利用求取图像凸包算法将所有有效区域进行合并。实验结果表明，在超声图像左心室分割效果上，所提方法能够在普通FCN的基础上获得很大的提升，在交并比评价标准下，该方法获取的左心室结果能够比传统CNN方法提升近15%。     

A new convolutional neural network model for peripapillary atrophy area segmentation from retinal fundus images

Peripapillary atrophy (PPA) is a clinical finding, which reflects the atrophy of retina layer and retinal pigment epithelium. The size of PPA area is a useful medical indicator, as it is highly associated with many diseases such as glaucoma and myopia. Therefore, separating the PPA area from retinal images, which is called PPA area segmentation, is very important. It is a challenging task, because PPA areas are irregular and non-uniform, and their contours are blurry and change gradually. To solve these issues, we transform the PPA area segmentation task into a task of segmenting another two areas with relatively regular and uniform shapes, and then propose a novel multi-task fully convolutional network (MFCN) model to jointly extract them from retinal images. Meanwhile, we take edge continuity of the target area into consideration. To evaluate the performance of the proposed model, we conduct experiments on images with PPA areas labelled by experts and achieve an average precision of 0.8928, outperforming the state-of-the-art models. To demonstrate the application of PPA segmentation in medical research, we apply PPA related features based on the segmented PPA area on differentiating glaucomatous and physiologic large cup cases. Experiment conducted on real datasets confirms the effectiveness of using these features for glaucoma diagnosis.     

A robust incremental learning framework for accurate skin region segmentation in color images

In this paper, we propose a robust incremental learning framework for accurate skin region segmentation in real-life images. The proposed framework is able to automatically learn the skin color information from each test image in real-time and generate the specific skin model (SSM) for that image. Consequently, the SSM can adapt to a certain image, in which the skin colors may vary from one region to another due to illumination conditions and inherent skin colors. The proposed framework consists of multiple iterations to learn the SSM, and each iteration comprises two major steps: (1) collecting new skin samples by region growing; (2) updating the skin model incrementally with the available skin samples. After the skin model converges (i.e., becomes the SSM), a post-processing can be further performed to fill up the interstices on the skin map. We performed a set of experiments on a large-scale real-life image database and our method observably outperformed the well-known Bayesian histogram. The experimental results confirm that the SSM is more robust than static skin models.     

前列腺磁共振图像分割的反卷积神经网络方法

目的 前列腺磁共振图像存在组织边界对比度低、有效区域少等问题,手工勾勒组织轮廓边界的传统分割方法无法满足临床实时性要求,针对这些问题提出了一种基于深度反卷积神经网络的前列腺磁共振图像分割算法。方法 基于深度学习理论,将训练图像样本输入设计好的卷积神经网络,提取具有高度区分性的前列腺图像特征,反卷积策略用于拓展特征图尺寸,使网络的输入尺寸与输出预测图大小匹配。网络生成的概率预测图通过训练一个softmax分类器,对预测图像取二值化,获得最终的分割结果。为克服原始图像中有效组织较少的问题,采用dice相似性系数作为卷积网络的损失函数。结果 本文算法以Dice相似性系数和Hausdorff距离作为评价指标,在MICCAI 2012数据集中,Dice相似性系数大于89.75%,Hausdorff距离小于1.3 mm,达到了传统方法的分割精度,并且将处理时间缩短在1 min以内,明显优于其他方法。结论 定量与定性的实验表明,基于反卷积神经网络的前列腺分割方法可以准确地对磁共振图像进行分割,相比于其他分割算法大幅度减小了处理时间,能够很好地适用于临床的前列腺图像分割任务。     

An end-to-end neural network for detecting hidden people in images based on multiple attention network

Camouflaged people like soldiers on the battlefield or even camouflaged objects in the natural environments are hard to be detected because of the strong resemblances between the hidden target and the background. That’s why seeing these hidden objects is a challenging task. Due to the nature of hidden objects, identifying them require a significant level of visual perception. To overcome this problem, we present a new end-to-end framework via a multi-level attention network in this paper. We design a novel inception module to extract multi-scale receptive fields features aiming at enhancing feature representation. Furthermore, we use a dense feature pyramid taking advantage of multi-scale semantic features. At last, to locate and distinguish the camouflaged target better from the background, we develop a multi-attention module that generates more discriminative feature representation and combines semantic information with spatial information from different levels. Experiments on the camouflaged people dataset show that our approach outperformed all state-of-the-art methods.

     

Structured neural networks with learning for texture segmentation in images

The structure and operation of a neural network with Hebb cell assembly organization is described. The structure is designed for solving problems of texture segmentation and recognition. The network is implemented by a computer program.Translated from Kibernetika i Sistemnyi Analiz, No. 6, pp. 149–161, November–December, 1991.     

多尺度融合增强的纵膈淋巴结超声弹性图像分割

目的 支气管超声弹性成像具有丰富的通道语义信息,精准的分割纵膈淋巴结对诊断肺癌是否转移具有重要意义,也对癌症的分期和治疗有着重要作用。目前,超声弹性图像分割研究较少,没有充分挖掘图像通道特征之间的关系。因此,提出一种结合注意力机制的多尺度融合增强的纵膈淋巴结超声弹性图像分割U-Net(attention-based multi-scale fusion enhanced ultrasound elastic images segmentation network for mediastinal lymph node, AMFE-UNet)。方法首先,考虑到图像可以提供纵膈淋巴结的位置和通道信息,设计密集卷积网络（dense convolutional network,DenseNet）作为模型编码器;其次,结合注意力机制和空洞卷积设计多尺度融合增强解码器,从多尺度和范围对结节的边界和纹理进行建模;最后,用选择性内核网络设计跳跃连接,将编码器的中间特征与解码器的输出特征充分融合。根据解码器特征进行数值或通道融合的方式不同,将AMFE-UNet分为A和B两个子型。结果 在超声弹性图像数据集...     

面向舌体分割的两阶段卷积神经网络设计

目的 由于舌体与周围组织颜色相似,轮廓模糊,传统的分割方法难以精准分割舌体,为此提出一种基于两阶段卷积神经网络的舌体分割方法。方法 首先,在粗分割阶段,将卷积层和全连接层相结合构建网络Rsnet,采用区域建议策略得到舌体候选框,从候选框中进一步确定舌体,从而实现对舌体的定位,去除大量的干扰信息;然后,在精分割阶段,将卷积层与反卷积层相结合构建网络Fsnet,对粗分割舌象中的每一个像素点进行分类进而实现精分割;最后,采用形态学相关算法对精分割后的舌体图像进行后续处理,进一步消除噪点和边缘粗糙点。结果 本文构建了包含2 764张舌象的数据集,在该数据集上进行五折交叉实验。实验结果表明,本文算法能够取得较为理想的分割结果且具有较快的处理速度。选取了精确度、召回率及F值作为评价标准,与3种常用的传统分割方法相比,在综合指标F值上分别提高了0.58、0.34、0.12,效率上至少提高6倍,与同样基于深度学习思想的MNC（multi-task network cascades）算法相比,在F值上提高0.17,效率上提高1.9倍。结论 将基于深度学习的方法应用到舌体分割中,有利于实现舌象的准确、鲁棒、快速分割。在分割之前,先对舌体进行定位,有助于进一步减少分割中的错分与漏分。实验结果表明,本文算法有效提升了舌体分割的准确性,能够为后续的舌象自动识别和分析奠定坚实的基础。     

深度卷积神经网络图像语义分割研究进展

在计算机视觉领域中,语义分割是场景解析和行为识别的关键任务,基于深度卷积神经网络的图像语义分割方法已经取得突破性进展。语义分割的任务是对图像中的每一个像素分配所属的类别标签,属于像素级的图像理解。目标检测仅定位目标的边界框,而语义分割需要分割出图像中的目标。本文首先分析和描述了语义分割领域存在的困难和挑战,介绍了语义分割算法性能评价的常用数据集和客观评测指标。然后,归纳和总结了现阶段主流的基于深度卷积神经网络的图像语义分割方法的国内外研究现状,依据网络训练是否需要像素级的标注图像,将现有方法分为基于监督学习的语义分割和基于弱监督学习的语义分割两类,详细阐述并分析这两类方法各自的优势和不足。本文在PASCAL VOC（pattern analysis, statistical modelling and computational learning visual object classes）2012数据集上比较了部分监督学习和弱监督学习的语义分割模型,并给出了监督学习模型和弱监督学习模型中的最优方法,以及对应的MIoU（mean intersection-over-union）。最后,指出了图像语义分割领域未来可能的热点方向。     

随机权神经网络增量构造学习方法研究进展

随机权神经网络(random weight neural network,RWNN)在解决数据定性和定量分析方面具有强大的潜力,其最显著的特征是隐含层参数随机生成.这一特征使得RWNN相比于基于梯度下降优化微调节点参数的神经网络具有诸多优势,如结构简单、易于实现和低人工干预等.RWNN的隐含层和输入层之间的参数是在一个固定区间内随机生成,而隐含层和输出层之间的输出权值则通过解析法进行求解.增量构造方法从一个小的初始网络开始,逐渐添加新的隐含层节点以提升模型品质,直到满足预期性能目标.基于此,重点从基础理论、增量构造学习方法和未来开放研究方向等方面切入,全面综述增量RWNN的研究进展.首先介绍RWNN的基本结构、理论和分析;进一步重点介绍RWNN在增量构造学习方法上的各种改进及应用;最后指出RWNN增量构造学习未来开放的研究方向.