Classification of leaf images

There are tremendous content‐based retrieval systems. Most of them are applied to general image databases. Some were proposed for specified databases such as texture databases, ancient paintings, document image databases, digital mammography, face image databases, etc. However, there are fewer for plant databases. Plants are used in various fields such as in foodstuff, medicine, and industry. Recently, plant is important for environment protection. On the other hand, the problem of plant destruction becomes worse in the few years. We should train people to know about plants, in turn, to treasure and protect them. In addition to the limited number of expert botanists, the convenient content‐based retrieval system for plant is necessary and useful, since it can retrieve related information and knowledge from plant database for the query leaf so as to facilitate fast learning of plants. In this study, a leaf database is constructed and a classification method for leaves is proposed. Most approaches for leaf classification in literature used contour‐based features. The proposed method tries to use region‐based features. The reasons are that region‐based features are more robust than contour‐based features since significant curvature points are hard to find. Those features adopted include aspect ratio, compactness, centroid, and horizontal/vertical projections. The effectiveness of the proposed method has been demonstrated by various experiments. On the average, our method has the classification accuracy for 1‐NN rule as 82.33% and the recall rate for 10 returned images as 48.2%, while the contour‐based method has 37.6% and 21.7%, respectively. © 2006 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 16, 15–23, 2006     

Image‐guided navigation of single‐element focused ultrasound transducer

The spatial specificity and controllability of focused ultrasound (FUS), in addition to its ability to modify the excitability of neural tissue, allows for the selective and reversible neuromodulation of the brain function, with great potential in neurotherapeutics. Intraoperative magnetic resonance imaging (MRI) guidance has limitations due to its complicated examination logistics, such as fixation through skull screws to mount the stereotactic frame, simultaneous sonication in the MRI environment, and restrictions in choosing MR‐compatible materials. To overcome these limitations, an image‐guidance system based on optical tracking and preoperative imaging data is developed, separating the imaging acquisition for guidance and sonication procedure for treatment. Techniques to define the local coordinates of the focal point of sonication are presented. First, mechanical calibration detects the concentric rotational motion of a rigid‐body optical tracker, attached to a straight rod mimicking the sonication path, pivoted at the virtual FUS focus. The spatial error presented in the mechanical calibration was compensated further by MRI‐based calibration, which estimates the spatial offset between the navigated focal point and the ground‐truth location of the sonication focus obtained from a temperature‐sensitive MR sequence. MRI‐based calibration offered a significant decrease in spatial errors (1.9 ± 0.8 mm; 57% reduction) compared to the mechanical calibration method alone (4.4 ± 0.9 mm). Using the presented method, pulse‐mode FUS was applied to the motor area of the rat brain, and successfully stimulated the motor cortex. The presented techniques can be readily adapted for the transcranial application of FUS to intact human brain. © 2012 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 22, 177–184, 2012     

A content-based image retrieval system based on object-moment feature

The last decade has witnessed great interest in research on content-based image retrieval (CBIR). In 2009, Lin et al. proposed a smart CBIR system based on colour and texture feature. Their system has a high detection rate except the cases where image objects have similar shapes. To enhance the detection rate a shape-based image feature called object-moment is proposed in this paper. Object-moment uses the moment of force to compute the object edge feature by calculating the distance from each edge pixel to the axis, and adding them up as a feature. Besides, we integrate the colour features (NSOM, CSOM) and the texture features (CCM, DBPSP) to enhance image detection rate and simplify computation of image retrieval. A series of analyses and comparisons are performed in our experiments to demonstrate that our proposed method improves the retrieval accuracy significantly.     

Optimized Identification with Severity Factors of Gastric Cancer for Internet of Medical Things

The Internet of Medical Things (IoMT) emerges with the vision of the Wireless Body Sensor Network (WBSN) to improve the health monitoring systems and has an enormous impact on the healthcare system for recognizing the levels of risk/severity factors (premature diagnosis, treatment, and supervision of chronic disease i.e., cancer) via wearable/electronic health sensor i.e., wireless endoscopic capsule. However, AI-assisted endoscopy plays a very significant role in the detection of gastric cancer. Convolutional Neural Network (CNN) has been widely used to diagnose gastric cancer based on various feature extraction models, consequently, limiting the identification and categorization performance in terms of cancerous stages and grades associated with each type of gastric cancer. This paper proposed an optimized AI-based approach to diagnose and assess the risk factor of gastric cancer based on its type, stage, and grade in the endoscopic images for smart healthcare applications. The proposed method is categorized into five phases such as image pre-processing, Four-Dimensional (4D) image conversion, image segmentation, K-Nearest Neighbour (K-NN) classification, and multi-grading and staging of image intensities. Moreover, the performance of the proposed method has experimented on two different datasets consisting of color and black and white endoscopic images. The simulation results verified that the proposed approach is capable of perceiving gastric cancer with 88.09% sensitivity, 95.77% specificity, and 96.55% overall accuracy respectively.     

Medical image fusion based on improved sum‐modified‐Laplacian

Sum‐modified‐Laplacian (SML) plays an important role in medical image fusion. However, fused rules based on larger SML always lead to fusion image distortion in transform domain image fusion or image information loss in spatial domain image fusion. Combined with average filter and median filter, a new medical image fusion method based on improved SML (ISML) is proposed. First, a basic fused image is gained by ISML, which is used for evaluation of the selection map of medical images. Second, difference images can be obtained by subtracting average image of all sources of medical images. Finally, basic fused image can be refined by difference images. The algorithm can both preserve the information of the source images well and suppress pixel distortion. Experimental results demonstrate that the proposed method outperforms the state‐of‐the‐art medical image fusion methods. © 2015 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 25, 206–212, 2015     

Deep transfer learning based hepatitis B virus diagnosis using spectroscopic images

Accurate identification of Hepatitis B virus (HBV) disease by analyzing the Raman spectroscopic images is a challenge for pathologists. To save precious human lives, an efficient technique is required with higher diagnostic accuracy at early‐stage of HBV. We proposed a novel method of HBV diagnosis using deep neural networks with the concept of transfer learning and Raman spectroscopic images. The proposed approach developed by utilizing pretrained convolutional neural networks ResNet101 by employing transfer learning on a real dataset of HBV‐infected blood plasma samples. Dataset consists of 1000 Raman images in which 600 are HBV‐infected blood plasma samples, and 400 are healthy ones. The developed model is capable to detect minute variation between infected and healthy samples and achieved enhanced performance. The proposed approach has been assessed and attained high classification accuracy, sensitivity, specificity, and AUC of 99.7%, 100%, 99.25%, and 98.7%, respectively. The proposed TL‐ResNet101 model outperformed the conventional approaches such as PCA‐SVM and PCA‐LDA and demonstrated improved accuracy more than 7%. High performance indicates that the developed TL‐ResNet101 model has potential to use for HBV diagnosis. Moreover, the developed automated approach can be extended for other disease.     

Image indexing by Color Plane Moment

Among the features of content‐based image retrieval, color features provide important clues to search similar image contents. In addition to color, there exists more information in the image. In this article, we propose a new method for content‐based image retrieval: Color Plane Moment (CPM). This method combines colors' content and their spatial distribution to improve image query results. It integrates the ideas of color histogram, backprojection, and moments. The CPM uses backprojection as one of the image preprocessing methods and computes the invariant moments with those dominant color plane images after preprocessing. There are several dominant colors chosen by color histogram in an image, and each can be expressed by seven invariant moment values that represent spatial distribution of those dominant color planes, respectively. Simulation results demonstrate that the proposed technique outperforms the other techniques. © 2002 Wiley Periodicals, Inc. Int J Imaging Syst Technol 12, 139–148, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.10022     

Color‐edge detection based on discrimination of noticeable color contrasts

Color‐edge detection is an important research task in the field of image processing. Efficient and accurate edge detection will lead to higher performance of subsequent image processing techniques, including image segmentation, object‐based image coding, and image retrieval. To improve the performance of color‐edge detection while considering that human eyes are ultimate receiver of color images, the perceptually insignificant edges should avoid being over‐detected. In this article, a color‐edge detection scheme based on the perceptual color contrast is proposed. The perceptual color contrast is defined as the visible color difference across an edge in the CIE‐Lab color space. A perceptual metric for measuring the visible color difference of a target color pixel is defined by utilizing the associated perceptually indistinguishable region. The perceptually indistinguishable region for each color pixel in the CIE‐Lab color space is estimated by the design of an experiment that considers the local property due to local changes in luminance. Simulation results show that the perceptual color contrast is effectively defined and the color edges in color images are detected while most of the perceptually insignificant edges are successfully suppressed through the proposed color‐edge detection scheme. © 2009 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 19, 332–339, 2009     

MI-STEG: A Medical Image Steganalysis Framework Based on Ensemble Deep Learning

Medical image steganography aims to increase data security by concealing patient-personal information as well as diagnostic and therapeutic data in the spatial or frequency domain of radiological images. On the other hand, the discipline of image steganalysis generally provides a classification based on whether an image has hidden data or not. Inspired by previous studies on image steganalysis, this study proposes a deep ensemble learning model for medical image steganalysis to detect malicious hidden data in medical images and develop medical image steganography methods aimed at securing personal information. With this purpose in mind, a dataset containing brain Magnetic Resonance (MR) images of healthy individuals and epileptic patients was built. Spatial Version of the Universal Wavelet Relative Distortion (S-UNIWARD), Highly Undetectable Stego (HUGO), and Minimizing the Power of Optimal Detector (MIPOD) techniques used in spatial image steganalysis were adapted to the problem, and various payloads of confidential data were hidden in medical images. The architectures of medical image steganalysis networks were transferred separately from eleven Dense Convolutional Network (DenseNet), Residual Neural Network (ResNet), and Inception-based models. The steganalysis outputs of these networks were determined by assembling models separately for each spatial embedding method with different payload ratios. The study demonstrated the success of pre-trained ResNet, DenseNet, and Inception models in the cover-stego mismatch scenario for each hiding technique with different payloads. Due to the high detection accuracy achieved, the proposed model has the potential to lead to the development of novel medical image steganography algorithms that existing deep learning-based steganalysis methods cannot detect. The experiments and the evaluations clearly proved this attempt.     

Multi‐Modal,Surface‐Focused Anticoagulation Using Poly‐2‐methoxyethylacrylate Polymer Grafts and Surface Nitric Oxide Release

This study examines platelet adhesion on surfaces that combine coatings to limit protein adsorption along with “anti‐platelet” nitric oxide (NO) release. Uncoated and poly‐2‐methoxyethylacrylate (PMEA) coated, gas permeable polypropylene (PP) membranes were placed in a bioreactor to separate plasma and gas flows. Nitrogen with 100/500/1000 ppm of NO was supplied to the gas side as a proof of concept. On the plasma side, platelet rich plasma (PRP, 1 × 108 cell/mL) was recirculated at low (60)/high (300) flows (mL/min). After 8 hours, adsorbed platelets on PP was quantified via a lactate dehydrogenase assay. Compared to plain PP, the PMEA coating alone reduced adsorption by 17.4 ± 9.2% and 29.6 ± 16.6% at low and high flow (p < 0.05), respectively. NO was more effective at low plasma flow. At 100 and 500 ppm of NO, adsorption fell by 37.9 ± 6.1% and 100 ± 4.7%, (p < 0.001), on plain PP. At high flow with 100, 500, and 1000 ppm of NO, adsorption reduced by 17.9 ± 17.8%, 46.4 ± 23.2%, and 100 ± 4.8%, (p < 0.001), respectively. On PMEA‐coated PP with only 100 ppm, adsorption fell by 69.7 ± 6.8 and 65.6% ± 16.9%, (p < 0.001), at low and high flows respectively. Therefore, the combination of an anti‐adsorptive coating with NO has great potential to reduce platelet adhesion and coagulation at biomaterial surfaces.     

基于深度学习耦合稀疏语义度量的商标检索算法

目的针对当前商标图像检索中的语义鸿沟问题,提出一种深度学习耦合稀疏语义度量的商标图像检索方案,有效抑制噪声干扰,降低冗余特征维数。方法首先,根据由卷积与池化组成的无监督学习机制,对输入商标图像进行多层特征提取,输出一维特征向量。随后,通过L2-支持向量机(L2-SVM)进行分类,利用特征向量进行训练,获得多级联特征。然后,根据商标图像的多级联特征和用户标签信息的异构数据结构,设计一种稀疏语义度量方法进行相似检索,减少语义鸿沟。此外,引入一种混合范数作为相似度量的稀疏约束,以抑制原始输入空间中的冗余特征维数和噪声,优化检索结果。结果实验表明,与当前流行的商标检索方案相比,所提算法具有更高的检索精度,其输出的结果中仅有1幅无关图像。结论该方案具有较高的检索精度和较强的鲁棒性,在商标检测、商标保护等方面中具有良好的应用价值。     

面向印刷质量控制的元学习盲图像质量评价方法

目的 为了更好地检测印刷图文复制效果,提高生产效率,提出一种针对图像复杂失真和内容变化的元学习盲图像质量评价模型.方法 首先在元训练部分,通过ResNet50网络获取多个失真数据集的共有失真先验知识;然后在元测试部分,融合ResNet50的多层次特征,实现对图像局部失真和全局失真的完整描述;最后通过特征降维、融合获得多层次特征的权值,建立图像质量评价网络模型.结果 模型在真实失真数据集LIVEC上SRCC达到0.87以及在合成失真数据集LIVE上SRCC达到0.97,且模型的预测性能和泛化性能都要优于其他算法.结论 所提出的元学习盲图像评价方法能够准确预测不同类型图像质量分数,可为印刷图像质量评价和印刷生产控制提供一定指导.     

基于方块编码的图像特征提取及检索算法

提出了一种基于方块编码的图像检索算法.首先将图像分成互不重叠的子图像块,根据图像块中各像素间的色差,利用方块编码的思想对这些子图像进行编码,然后根据人眼的视觉特性来定义图像的关键块,最后借助于基于关键字的文本检索技术进行图像检索.同时,考虑到不同类型的关键块在表征图像内容时重要程度的不同而赋予其不同的权值.实验结果表明本文算法在图像的相似性检索时是有效的,并具有较高的检索效率.     

Fuzzy entropy‐based MR brain image segmentation using modified particle swarm optimization

This article presents an image segmentation technique based on fuzzy entropy, which is applied to magnetic resonance (MR) brain images in order to detect brain tumors. The proposed method performs image segmentation based on adaptive thresholding of the input MR images. The image is classified into two membership functions (MFs) of the fuzzy region: Z‐function and S‐function. The optimal parameters of these fuzzy MFs are obtained using modified particle swarm optimization (MPSO) algorithm. The objective function for obtaining the optimal fuzzy MF parameters is considered to be the maximum fuzzy entropy. Through a number of examples, The performance is compared with existing entropy based object segmentation approaches and the superiority of the proposed method is demonstrated. The experimental results are compared with the exhaustive search method and Otsu's segmentation technique. The result shows the proposed fuzzy entropy‐based segmentation method optimized using MPSO achieves maximum entropy with proper segmentation of infected areas and with minimum computational time. © 2013 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 23, 281–288, 2013     

ANFIS‐EM approach for PET brain image reconstruction

In this article, for the reconstruction of the positron emission tomography (PET) images, an iterative MAP algorithm was instigated with its adaptive neurofuzzy inference system based image segmentation techniques which we call adaptive neurofuzzy inference system based expectation maximization algorithm (ANFIS‐EM). This expectation maximization (EM) algorithm provides better image quality when compared with other traditional methodologies. The efficient result can be obtained using ANFIS‐EM algorithm. Unlike any usual EM algorithm, the predicted method that we call ANFIS‐EM minimizes the EM objective function using maximum a posteriori (MAP) method. In proposed method, the ANFIS‐EM algorithm was instigated by neural network based segmentation process in the image reconstruction. By the image quality parameter of PSNR value, the adaptive neurofuzzy based MAP algorithm and de‐noising algorithm compared and the PET input image is reconstructed and simulated in MATLAB/simulink package. Thus ANFIS‐EM algorithm provides 40% better peak signal to noise ratio (PSNR) when compared with MAP algorithm. © 2015 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 25, 1–6, 2015     

Automatic Detection of Aortic Dissection Based on Morphology and Deep Learning

Aortic dissection (AD) is a kind of acute and rapidly progressing cardiovascular disease. In this work, we build a CTA image library with 88 CT cases, 43 cases of aortic dissection and 45 cases of health. An aortic dissection detection method based on CTA images is proposed. ROI is extracted based on binarization and morphology opening operation. The deep learning networks (InceptionV3, ResNet50, and DenseNet) are applied after the preprocessing of the datasets. Recall, F1-score, Matthews correlation coefficient (MCC) and other performance indexes are investigated. It is shown that the deep learning methods have much better performance than the traditional method. And among those deep learning methods, DenseNet121 can exceed other networks such as ResNet50 and InceptionV3.     

Image reconstruction with a shift‐variant filtration in circular cone‐beam CT

It is well known that cone‐beam data acquired with a circular orbit are insufficient for exact image reconstruction. Despite this, because a cone‐beam scanning configuration with a circular orbit is easy to implement in practice, it has been widely employed for data acquisition in, e.g., micro‐CT and CT imaging in radiation therapy. The algorithm developed by Feldkamp, Davis, and Kress (FDK) and its modifications, such as the Tent–FDK (T‐FDK) algorithm, have been used for image reconstruction from circular cone‐beam data. In this work, we present an algorithm with spatially shift‐variant filtration for image reconstruction in circular cone‐beam CT. We performed computer‐simulation studies to compare the proposed and existing algorithms. Numerical results in these studies demonstrated that the proposed algorithm has resolution properties comparable to, and noise properties better than, the FDK algorithm. As compared to the T‐FDK algorithm, our proposed algorithm reconstructs images with an improved in‐plane spatial resolution. © 2005 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 14, 213–221, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.20026