Empirical Comparisons of Deep Learning Networks on Liver Segmentation

Accurate segmentation of CT images of liver tumors is an important adjunct for the liver diagnosis and treatment of liver diseases. In recent years, due to the great improvement of hard device, many deep learning based methods have been proposed for automatic liver segmentation. Among them, there are the plain neural network headed by FCN and the residual neural network headed by Resnet, both of which have many variations. They have achieved certain achievements in medical image segmentation. In this paper, we firstly select five representative structures, i.e., FCN, U-Net, Segnet, Resnet and Densenet, to investigate their performance on liver segmentation. Since original Resnet and Densenet could not perform image segmentation directly, we make some adjustments for them to perform live segmentation. Our experimental results show that Densenet performs the best on liver segmentation, followed by Resnet. Both perform much better than Segnet, U-Net, and FCN. Among Segnet, U-Net, and FCN, U-Net performs the best, followed by Segnet. FCN performs the worst.     

Neural network classifier and multiple hypothesis image segmentation for dietary assessment using calorie calculator

Nowadays, dietary assessment becomes the emerging system for evaluating the person’s food intake. In this paper, the multiple hypothesis image segmentation and feed-forward neural network classifier are proposed for dietary assessment to enhance the performance. Initially, the segmentation is applied to input image which is used to determine the regions where a particular food item is located using salient region detection, multi-scale segmentation, and fast rejection. Then, the significant feature of food items is extracted by the global feature and local feature extraction method. After the features are obtained, the classification is performed for each segmented region using feed-forward neural network model. Finally, the calorie value is computed with the aid of (i) food area volume and (ii) calorie and nutrition measure based on mass value. The outcome of the proposed method attains 96% of accuracy value which provides the better classification performance.     

Food recognition using neural network classifier and multiple hypotheses image segmentation

ABSTRACT

This paper proposes the multiple-hypotheses image segmentation and feed-forward neural network classifier for food recognition to improve the performance. Initially, the food or meal image is given as input. Then, the segmentation is applied to identify the regions, where a particular food item is located using salient region detection, multi-scale segmentation, and fast rejection. Then, the features of every food item are extracted by the global feature and local feature extraction. After the features are obtained, the classification is performed for each segmented region using a feed-forward neural network model. Finally, the calorie value is computed with the aid of (i) food volume and (ii) calorie and nutrition measure based on mass value. The experimental results and performance evaluation are validated. The outcome of the proposed method attains 0.947 for Macro Average Accuracy (MAA) and 0.959 for Standard Accuracy (SA), which provides better classification performance.     

Fractional gravitational search-radial basis neural network for bone marrow white blood cell classification

Automotive image segmentation systems are becoming an important tool in the medical field for disease diagnosis. The white blood cell (WBC) segmentation is crucial, because it plays an important role in the determination of the diseases and helps experts to diagnose the blood disease disorders. The precise segmentation of the WBCs is quite challenging because of the complex contents in the bone marrow smears. In this paper, a novel neural network (NN) classifier is proposed for the classification of the bone marrow WBCs. The proposed NN classifier integrates the fractional gravitation search (FGS) algorithm for updating the weight in the radial basis function mapping for the classification of the WBC based on the cell nucleus feature. The experimentation of the proposed FGS-RBNN classifier is carried on the images collected from the publically available dataset. The performance of the proposed methodology is evaluated over the existing classifier approaches using the measures accuracy, sensitivity, and specificity. The results show that the classification using the nucleus features alone can be utilized to achieve the classification with the better accuracy. Moreover, the classification performance of the proposed FGS-RBNN is better than the existing classifiers, and it is proved to be the efficacious classifier with a classification accuracy of 95%.     

Sailfish Optimizer with EfficientNet Model for Apple Leaf Disease Detection

Recent developments in digital cameras and electronic gadgets coupled with Machine Learning (ML) and Deep Learning (DL)-based automated apple leaf disease detection models are commonly employed as reasonable alternatives to traditional visual inspection models. In this background, the current paper devises an Effective Sailfish Optimizer with EfficientNet-based Apple Leaf disease detection (ESFO-EALD) model. The goal of the proposed ESFO-EALD technique is to identify the occurrence of plant leaf diseases automatically. In this scenario, Median Filtering (MF) approach is utilized to boost the quality of apple plant leaf images. Moreover, SFO with Kapur's entropy-based segmentation technique is also utilized for the identification of the affected plant region from test image. Furthermore, Adam optimizer with EfficientNet-based feature extraction and Spiking Neural Network (SNN)-based classification are employed to detect and classify the apple plant leaf images. A wide range of simulations was conducted to ensure the effective outcomes of ESFO-EALD technique on benchmark dataset. The results reported the supremacy of the proposed ESFO-EALD approach than the existing approaches.     

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.     

Fruits and Vegetable Diseases Recognition Using Convolutional Neural Networks

As they have nutritional, therapeutic, so values, plants were regarded as important and they’re the main source of humankind’s energy supply. Plant pathogens will affect its leaves at a certain time during crop cultivation, leading to substantial harm to crop productivity & economic selling price. In the agriculture industry, the identification of fungal diseases plays a vital role. However, it requires immense labor, greater planning time, and extensive knowledge of plant pathogens. Computerized approaches are developed and tested by different researchers to classify plant disease identification, and that in many cases they have also had important results several times. Therefore, the proposed study presents a new framework for the recognition of fruits and vegetable diseases. This work comprises of the two phases wherein the phase-I improved localization model is presented that comprises of the two different types of the deep learning models such as You Only Look Once (YOLO)v2 and Open Exchange Neural (ONNX) model. The localization model is constructed by the combination of the deep features that are extracted from the ONNX model and features learning has been done through the convolutional-05 layer and transferred as input to the YOLOv2 model. The localized images passed as input to classify the different types of plant diseases. The classification model is constructed by ensembling the deep features learning, where features are extracted dimension of from pre-trained Efficientnetb0 model and supplied to next 07 layers of the convolutional neural network such as 01 features input, 01 ReLU, 01 Batch-normalization, 02 fully-connected. The proposed model classifies the plant input images into associated labels with approximately 95% prediction scores that are far better as compared to current published work in this domain.     

Fused and Modified Evolutionary Optimization of Multiple Intelligent Systems Using ANN,SVM Approaches

The Fused Modified Grasshopper Optimization Algorithm has been proposed, which selects the most specific feature sets from images of the disease of plant leaves. The Proposed algorithm ensures the detection of diseases during the early stages of the diagnosis of leaf disease by farmers and, finally, the crop needed to be controlled by farmers to ensure the survival and protection of plants. In this study, a novel approach has been suggested based on the standard optimization algorithm for grasshopper and the selection of features. Leaf conditions in plants are a major factor in reducing crop yield and quality. Any delay or errors in the diagnosis of the disease can lead to delays in the management of plant disease spreading and damage and related material losses. Comparative new heuristic optimization of swarm intelligence, Grasshopper Optimization Algorithm was inspired by grasshopper movements for their feeding strategy. It simulates the attitude and social interaction of grasshopper swarm in terms of gravity and wind advection. In the decision on features extracted by an accelerated feature selection algorithm, popular approaches such as ANN and SVM classifiers had been used. For the evaluation of the proposed model, different data sets of plant leaves were used. The proposed model was successful in the diagnosis of the diseases of leaves the plant with an accuracy of 99.41 percent (average). The proposed biologically inspired model was sufficiently satisfied, and the best or most desirable characteristics were established. Finally, the results of the research for these data sets were estimated by the proposed Fused Modified Grasshopper Optimization Algorithm (FMGOA). The results of that experiment were demonstrated to allow classification models to reduce input features and thus to increase the precision with the presented Modified Grasshopper Optimization Algorithm. Measurement and analysis were performed to prove the model validity through model parameters such as precision, recall, f-measure, and precision.     

From 3-D Sonar Images to Augmented Reality Models for Objects Buried on the Seafloor

The investigation of man-made objects lying on or embedded in the sea floor can be carried out with acoustic imaging techniques and subsequent data processing. In this paper, we describe a processing chain that starts with a 3-D acoustic image of the object to be examined and ends with an augmented reality model, which requires minimal user involvement. Essentially, the chain includes blocks devoted to statistical 3-D segmentation, semi-automatic surface fitting, extraction of measurements, and augmented reality modeling. In particular, the 3-D segmentation method presented here is based on a volume-growing approach, which is essentially a 3-D extension of the traditional 2-D region growing. The volume-growing operation is guided by a statistical approach based on the optimal decision theory. The surface-fitting block is based on predefined geometric models, i.e., one of them is tentatively selected by the user after a preliminary study of the segmented object and is automatically or partially manually adapted to the segmented data by exploiting an inertial tensor. The proposed chain was successfully applied to the analysis of some 3-D acoustic images obtained from both simulated and real signals acquired by different sonar systems and containing objects that were completely or partially buried. The segmentation results provided an effective help in the identification of the object's shape, i.e., facilitating the subsequent surface-fitting step and the extraction of related measurements.     

Parametric modeling in food package defect imaging

A novel approach in food package defect detection is proposed based on system identification theory, in which the channel defect detection problem can be regarded as the conventional system identification problem, i.e., estimation of the system impulse response based on the input-output sequence using parametric and nonparametric models. The well-known parametric model ARX has been investigated in this paper. The data are collected with a focused ultrasound transducer (17.3 MHz, 6.35-mm diameter, f/2, 173 mum -6 dB pulse-echo lateral beam width at the focus) scanned over a rectangular grid, keeping the packages in the focus. Performance is measured in terms of detection rate, image contrast, and contrast-to-noise ratio. The results using the ARX model are compared with previous image formation techniques and also compared with the non-parametric method, i.e., spectral analysis. The results show that the ARX model has the comparable detection rate as RFCS and higher detection rate than BAI and RFS (except 6-mum air-filled channel in plastic trilaminate film) for channel in plastic trilaminate film. The ARX model has achieved the moderate contrast enhancement and ranks second in contrast-to-noise ratio enhancement among the compared techniques. The ARX model has a low detection rate for channel defects in aluminum trilaminate film, which shows that its performance is material-dependent. Finally, the parametric method, ARX model demonstrates better performance than the non-parametric method, spectral analysis for food package defect detection.     

Recognition and Classification of Pomegranate Leaves Diseases by Image Processing and Machine Learning Techniques

Disease recognition in plants is one of the essential problems in agricultural image processing. This article focuses on designing a framework that can recognize and classify diseases on pomegranate plants exactly. The framework utilizes image processing techniques such as image acquisition, image resizing, image enhancement, image segmentation, ROI extraction (region of interest), and feature extraction. An image dataset related to pomegranate leaf disease is utilized to implement the framework, divided into a training set and a test set. In the implementation process, techniques such as image enhancement and image segmentation are primarily used for identifying ROI and features. An image classification will then be implemented by combining a supervised learning model with a support vector machine. The proposed framework is developed based on MATLAB with a graphical user interface. According to the experimental results, the proposed framework can achieve 98.39% accuracy for classifying diseased and healthy leaves. Moreover, the framework can achieve an accuracy of 98.07% for classifying diseases on pomegranate leaves.     

MRI brain segmentation in combination of clustering methods with Markov random field

Medical image segmentation is a preliminary stage of inclusion in identification tools. The correct segmentation of brain Magnetic Resonance Imaging (MRI) images is crucial for an accurate detection of the disease diagnosis. Due to in‐homogeneity, low distinction and noise the segmentation of the brain MRI images is treated as the most challenging task. In this article, we proposed hybrid segmentation, by combining the clustering methods with Hidden Markov Random Field (HMRF) technique. This aims to decrease the computational load and improves the runtime of segmentation method, as MRF methodology is used in post‐processing the images. Its evaluation has performed on real imaging data, resulting in the classification of brain tissues with dice similarity metric. These results indicate the improvement in performance of the proposed method with various noise levels, compared with existing algorithms. In implementation, selection of clustering method provides better results in the segmentation of MRI brain images.     

基于超像素快速模糊聚类的印刷品图像分割方法

针对当前彩色印刷品色差检测过程中效率低、复杂性高等问题,提出了一种基于超像素快速模糊聚类的印刷品图像分割方法（SFFCM）。先用简单线性迭代聚类（SLIC）算法将图像分割为紧密相邻的超像素区域。每个超像素区域被视为一个独立的聚类单元。随后,将模糊C均值聚类（FCM）算法应用于超像素的归属关系计算中,即引入隶属度值,允许超像素归属于多个聚类中心,并通过权衡归属度值来实现模糊聚类。实验结果表明,相对于其他算法,本文方法在保持良好实时性的同时,实现了较好的分割效果,有效平衡了算法复杂度与分割效果之间的关系。     

Epileptic Seizures Diagnosis Using Amalgamated Extremely Focused EEG Signals and Brain MRI

There exists various neurological disorder based diseases like tumor, sleep disorder, headache, dementia and Epilepsy. Among these, epilepsy is the most common neurological illness in humans, comparable to stroke. Epilepsy is a severe chronic neurological illness that can be discovered through analysis of the signals generated by brain neurons and brain Magnetic resonance imaging (MRI). Neurons are intricately coupled in order to communicate and generate signals from human organs. Due to the complex nature of electroencephalogram (EEG) signals and MRI’s the epileptic seizures detection and brain related problems diagnosis becomes a challenging task. Computer based techniques and machine learning models are continuously giving their contributions to diagnose all such diseases in a better way than the normal process of diagnosis. Their performance may sometime degrade due to missing information, selection of poor classification model and unavailability of quality data that are used to train the models for better prediction. This research work is an attempt to epileptic seizures detection by using a multi focus dataset based on EEG signals and brain MRI. The key steps of this work are: feature extraction having two different streams i.e., EEG using wavelet transformation along with SVD-Entropy, and MRI using convolutional neural network (CNN), after extracting features from both streams, feature fusion is applied to generate feature vector used by support vector machine (SVM) to diagnose the epileptic seizures. From the experimental evaluation and results comparison with the current state-of-the-art techniques, it has been concluded that the performance of the proposed scheme is better than the existing models.

     

A lightweight deep learning model for retinal optical coherence tomography image classification

Automated retinal disease detection and grading is one of the most researched areas in medical image analysis. In recent years, Deep Learning models have attracted much attention in this field. Hence, in this paper, we present a Deep Learning-based, lightweight, fully automated end-to-end diagnostic system for the detection of the two major retinal diseases, namely diabetic macular oedema (DME) and drusen macular degeneration (DMD). Early detection of these diseases is important to prevent vision impairment. Optical coherence tomography (OCT) is the main imaging technique for detecting these diseases. The model proposed in this work is based on residual blocks and channel attention modules. The performance of the model is evaluated using the publicly available Mendeley OCT dataset and the Duke dataset. We were able to achieve a classification accuracy of 99.5% in the Mendeley test dataset and 94.9% in the Duke dataset with the proposed model. For the application, we performed an extensive evaluation of pre-trained models (LeNet, AlexNet, VGG-16, ResNet50 and SE-ResNet). The proposed model has a much smaller number of trainable parameters and shows superior performance compared to existing methods.     

Research on automatic identification system of tobacco diseases

In order to improve recognition accuracy of tobacco diseases, an identification method based on multi-feature and genetic algorithms optimizing BP neural network was proposed. First, Otsu method was used to obtain disease location information and GrabCut function was initialized for extracting diseased area effectively. Second, colour moments, disease contour and GLCM were used to get colour, multi-contour and texture features. Once again, BP neural network was optimized by genetic algorithm, and the optimal initial weights and thresholds were obtained, which shortened the training time and improved the accuracy of disease identification. Finally, BP neural network model for tobacco diseases diagnosis was established with the mobile client as input and the user services as output. The field experiment showed that the method could diagnose eight types of tobacco diseases effectively and automatically. The average recognition accuracy rate of selected tobacco diseases was about 92.5%.     

Factorization-based active contour segmentation and pelican optimization-based modified bidirectional long short-term memory for ovarian tumor detection

This paper proposes a novel three-dimensional convolution neural network-based modified bidirectional long short-term memory with pelican optimization (3D CNN based MBiLSTM with PO) algorithm for multiclass ovarian tumor detection. Initially, the International Collaboration on Cancer Reporting endometrial cancer dataset images are provided in pre-processing phase, which uses a pre-emphasis filter to process the input image. In the segmentation phase, pre-processed data is then partitioned into diverse subgroups (i.e., pixels), which minimizes the complexity of images. In this paper, a factorization-based active contour technique is employed in the effective segmentation of images. The segmented features are then extracted and classified using the 3D CNN-MBiLSTM with PO algorithm. Finally, the experimental results are conducted and compared with various other approaches for various performance metrics. Each metric is evaluated with respect to the different number of iterations. The accuracy, sensitivity, and specificity have reached a higher value of 98.5%, 96%, and 98.25%, respectively.     

Early detection of brain cancer using association allotment hierarchical clustering

Medical applications uses the computer assisted diagnosis system for recognizing the enormous amount of diseases, but cancer is one of the major challenge to the doctors because it is difficult to recognize in earlier stage with accurate manner. To overcome these issues, in this article, an early detection of brain cancer using Association Allotment Hierarchical Clustering technique is proposed. The proposed automatic detection process includes various processing steps such as preprocessing, image segmentation, feature extraction, selection, and cancer identification process. Initially the microscopic images are captured which contaminated several noises that have been removed by applying the mutual piece-wise linear transformation filtering approach. The method successfully eliminates irrelevant information and enhances the quality of the captured biopsy image. After that affected cancer cell region has been segmented with the help of Association Allotment Hierarchical Clustering method which examines the cell, tissues, and relevant border while segmenting the cancer cell. From the segmented region, different textures, statistical features are extracted depending on the tissue level, cell level, region, and contour level. Then to improve the performance of classification, the optimum features are selected using gray wolf optimization. The novelty of the proposed method is to give a better performance and the accuracy is obtained nearly 100%. Finally, the selected features are classified using neural network. Experimental result demonstrates that the performance of proposed method in terms of segmentation and computation time is better when compared with state-of-art approaches.     

Macro-level pedestrian and bicycle crash analysis: Incorporating spatial spillover effects in dual state count models

This study attempts to explore the viability of dual-state models (i.e., zero-inflated and hurdle models) for traffic analysis zones (TAZs) based pedestrian and bicycle crash frequency analysis. Additionally, spatial spillover effects are explored in the models by employing exogenous variables from neighboring zones. The dual-state models such as zero-inflated negative binomial and hurdle negative binomial models (with and without spatial effects) are compared with the conventional single-state model (i.e., negative binomial). The model comparison for pedestrian and bicycle crashes revealed that the models that considered observed spatial effects perform better than the models that did not consider the observed spatial effects. Across the models with spatial spillover effects, the dual-state models especially zero-inflated negative binomial model offered better performance compared to single-state models. Moreover, the model results clearly highlighted the importance of various traffic, roadway, and sociodemographic characteristics of the TAZ as well as neighboring TAZs on pedestrian and bicycle crash frequency.     

采用递归门限分析的红外目标分割

提出了一种有效的基于递归门限分析的红外目标分割方法。针对传统方法在目标的相对面积较小时背景信息容易误分的问题,将传统分割方法和递归处理结合起来,用于分割红外目标。在分割时,将每次分割得到的背景部分(即暗部分)淘汰掉,而保留分割得到的目标部分(即亮部分)。对得到的目标部分进行再分割,又得到新的目标和背景部分,如此重复下去,直至得到目标为止。对传统的Otsu方法、一维熵方法、二维熵方法的递归分割特性进行了分析比较,并根据目标的先验知识提出一种合理的递归终止准则。试验结果证明,基于递归门限分析的方法是一种行之有效的目标分割方法,分割性能优于传统方法。