CNN System Research Articles

A hybrid features fusion-based framework for classification of breast micronodules using ultrasonography

BackgroundBreast cancer is one of the leading diseases worldwide. According to estimates by the National Breast Cancer Foundation, over 42,000 women are expected to die from this disease in 2024.ObjectiveThe prognosis of breast cancer depends on the early detection of breast micronodules and the ability to distinguish benign from malignant lesions. Ultrasonography is a crucial radiological imaging technique for diagnosing the illness because it allows for biopsy and lesion characterization. The user’s level of experience and knowledge is vital since ultrasonographic diagnosis relies on the practitioner’s expertise. Furthermore, computer-aided technologies significantly contribute by potentially reducing the workload of radiologists and enhancing their expertise, especially when combined with a large patient volume in a hospital setting.MethodThis work describes the development of a hybrid CNN system for diagnosing benign and malignant breast cancer lesions. The models InceptionV3 and MobileNetV2 serve as the foundation for the hybrid framework. Features from these models are extracted and concatenated individually, resulting in a larger feature set. Finally, various classifiers are applied for the classification task.ResultsThe model achieved the best results using the softmax classifier, with an accuracy of over 95%.ConclusionComputer-aided diagnosis greatly assists radiologists and reduces their workload. Therefore, this research can serve as a foundation for other researchers to build clinical solutions.

A Research Investigation on Conventional Neural Network-Based Disease Detection Techniques for Sugarcane Plants

In the current times, the number of viruses and infections in the sugarcane plants is widespread. If we want to properly correct these infections, we need to use artificial intelligence such as CNN and RNN. Therefore, in this study, how to prevent the disease in sugarcane using the CNN and RNN system is taken as a test. This model is trained on a diverse dataset of automatic diagnosis images, with a focus on addressing the inherent class in detection. The neural network architecture is designed to capture intricate patterns indicative of sugarcane manifestations in automatic diagnosis images. Through an iterative training process, the model learns to discern subtle features associated with automatic diagnosis, achieving remarkable accuracy. The experimental results confirm the efficacy of our proposed methodology. It explores the many CNN architectures used for plant disease detection, including AlexNet, VGGNet, ResNet, InceptionNet, and DenseNet, as well as their pros and limitations. The survey also discusses the importance of RNNs in plant disease detection, specifically in time-series data analysis, where RNNs have been shown to be useful in forecasting the spread of plant diseases over time. This report also provides a successful outcome for researchers working on the creation of a recognition system for sugarcane diseases. Keywords: convolutional neural networks (cnn), recurrent neural networks (rnn), Sugarcane plants disease detection Accuracy, Genetic algorithm

Privacy-preserving AI for early diagnosis of thoracic diseases using IoTs: A federated learning approach with multi-headed self-attention for facilitating cross-institutional study

Our study recognized the crucial role of early diagnosis of pulmonary radiological abnormalities such as pneumothorax, effusion, pneumonia, cardiomegaly, and COVID-19. We proposed FedXNet, which is a collaborative deep learning model based on federated learning (FL) exploiting edge computing resources efficiently to accurately deal with them and ensure privacy. Our developed model is notable for its integration of Multi-Headed Self-Attention, a complex technique that allows the model to focus on several parts of the input data at once. This improves the model’s capacity to uncover complex patterns and correlations within the medical images. This multi-class CNN system uses a thorough four-pronged approach: (1) facilitating cross-institutional, federated training without sacrificing the integrity of individual data, (2) image preprocessing to achieve robust model accuracy, (3) efficient Feature extraction using pre-trained models and our dedicated FedXNet architecture, as well as (4) a variety of classifiers tailored to each disease, resulting in impressive diagnostic performance for a range of thoracic diseases, including COVID-19. This model paves the way for a future where timely diagnosis and better patient outcomes become a reality, empowered by the collaborative spirit of FL exploiting edge computing resources of IoT for implementing robust deep learning models.

Brain tumor detection based on a novel and high-quality prediction of the tumor pixel distributions

The work presented in this paper is in the area of brain tumor detection. We propose a fast detection system with 3D MRI scans of Flair modality. It performs 2 functions, predicting the gray level distribution and location distribution of the pixels in the tumor regions and generating tumor masks with pixel-wise precision. To facilitate 3D data analysis and processing, we introduce a 2D histogram presentation encompassing the gray-level distribution and pixel-location distribution of a 3D object. In the proposed system, specific 2D histograms highlighting tumor-related features are established by exploiting the left-right asymmetry of a brain structure. A modulation function, generated from the input data of each patient case, is applied to the 2D histograms to transform them into coarsely or finely predicted distributions of tumor pixels. The prediction result helps to identify/remove tumor-free slices. The prediction and removal operations are performed to the axial, coronal and sagittal slice series of a brain image, transforming it into a 3D minimum bounding box of its tumor region. The bounding box is utilized to finalize the prediction and generate a 3D tumor mask. The proposed system has been tested extensively with the data of more than 1200 patient cases in BraTS2018∼2021 datasets. The test results demonstrate that the predicted 2D histograms resemble closely the true ones. The system delivers also very good tumor detection results, comparable to those of state-of-the-art CNN systems with mono-modality inputs. They are reproducible and obtained at an extremely low computation cost and without need for training.

A hybrid human recognition framework using machine learning and deep neural networks.

Faces are a crucial environmental trigger. They communicate information about several key features, including identity. However, the 2019 coronavirus pandemic (COVID-19) significantly affected how we process faces. To prevent viral spread, many governments ordered citizens to wear masks in public. In this research, we focus on identifying individuals from images or videos by comparing facial features, identifying a person's biometrics, and reducing the weaknesses of person recognition technology, for example when a person does not look directly at the camera, the lighting is poor, or the person has effectively covered their face. Consequently, we propose a hybrid approach of detecting either a person with or without a mask, a person who covers large parts of their face, and a person based on their gait via deep and machine learning algorithms. The experimental results are excellent compared to the current face and gait detectors. We achieved success of between 97% and 100% in the detection of face and gait based on F1 score, precision, and recall. Compared to the baseline CNN system, our approach achieves extremely high recognition accuracy.

A Novel Additive Internet of Things (IoT) Features and Convolutional Neural Network for Classification and Source Identification of IoT Devices

The inter-class classification and source identification of IoT devices has been studied by several researchers recently due to the vast amount of available IoT devices and the huge amount of data these IoT devices generate almost every minute. As such there is every need to identify the source where the IoT data is generated and also separate an IoT device from the other using on the data they generate. This paper proposes a novel additive IoT features with the CNN system for the purpose of IoT source identification and classification. Experimental results shows that indeed the proposed method is very effective achieving an overall classification and source identification accuracy of 99.67 %. This result has a practical application to forensics purposes due to the fact that accurately identifying and classifying the source of an IoT device via the generated data can link organisations/persons to the activities they perform over the network. As such ensuring accountability and responsibility by IoT device users.

Evaluation of SEU impact on convolutional neural networks based on BRAM and CRAM in FPGAs

This study evaluates the sensitivity of FPGA-based CNN systems and the efficacy of selective hardening approaches with laser injection and proton irradiation. The LeNet-5 CNN architecture was taken as a case study. First, the entire CNN with block random access memory (BRAM) attached is evaluated by laser test scanning, and errors generated from each layer are detected to determine the most critical layer. Second, the same network is evaluated without BRAM, considering only the configuration RAM (CRAM) and errors generated from each layer are detected. Third, the impact of laser scan errors from BRAM and CRAM are compared, while tracing the layers in which these errors are generated. Then, proton testing is performed on the entire CNN, and errors from each layer are detected and compared to validate the laser scanning test. Experimental results from laser and proton demonstrate that CRAM errors have a larger impact on CNN layers than BRAM errors. Regarding layer criticality, the second convolutional layer (C2) appears to be the most critical one because it generates most CRAM errors. This was validated by serial and parallel designs, where both confirm similar trends. Accordingly, the partial triple modular redundancy (TMR) approach is only applied to the critical portions, leading to around 40 % reliability improvement with less than 20 % overhead.

An optimized CNN system to recognize handwritten characters in ancient documents in Grantha script

An optimized CNN system to recognize handwritten characters in ancient documents in Grantha script

Retracted Article: Artificial intelligence for the identification of healthy fruits and vegetables using MMDL-ABO

ABSTRACT Computer vision is a reliable and cutting-edge method of processing images, yielding desirable results and holding great promise. Many other industries, including agriculture, have begun making heavy use of computer vision. Implications for the study of agricultural processes through the application of image processing methods are quite promising. Existing works are only focused on disease categorisation of agricultural productions. They have focused only on feature descriptor, mathematical framework or identifying healthy fruits and vegetables with accuracy. While transformers have seen research for agricultural applications, deep CNNs and transformers working together have not been explored. In this paper, we examine the feasibility of using a multi-model deep learning network (MMDL) for complicated picture categorisation by combining a tweaked version of the popular deep CNN system VGG-16 with a 3D CNN network and a Swin transformer. To increase the precision of MMDL, African Buffalo Optimisation(ABO) is employed to identify healthy fruits and vegetables. The experiments are conducted using datasets that are freely accessible to the public and evaluated using a number of different criteria. Proposed method has shown more accuracy and recall than the existing methods.

A Data-Driven Cyber–Physical System Using Deep-Learning Convolutional Neural Networks: Study on False-Data Injection Attacks in an Unmanned Ground Vehicle Under Fault-Tolerant Conditions

Leveraging the benefits of deep-learning convolutional neural networks, we introduce a new data-driven cyber–physical system specifically designed to address the vulnerability of middleware software, namely, Robot Operating System (ROS), widely implemented in robotics in both civilian and military domains. As for our research platform, we employ the GVR-BOT unmanned ground vehicle, which is a replicate of the U.S. Army ground robot. We focus our study on the ability of the ground robot to operate under fault-tolerant conditions, making it challenging from the standpoint of cybersecurity to differentiate between legitimate and malicious operations. The GVR-BOT ground vehicle belongs to a class of differential drive ground robots and employs ROS in its onboard computer to interact with users. To facilitate deep learning, we develop a large database of images based on the network-traffic data of ROS, describing the dynamics of the GVR-BOT ground robot under legitimate and malicious operations. We use our image database to train and validate the performance of our deep-learning CNN system. Given a set of RGB/grayscale images describing the normalized time-series data representing the dynamics of the GVR-BOT ground robot, the objective of our proposed cybersecurity algorithm is to safeguard the legitimate operation of the ground robot under fault-tolerant conditions, such that any attempts to compromise its performance (e.g., malicious attacks) can be prevented within the minimum detection time. Our research indicates a promising result as our system is capable of detecting malicious attacks with high accuracy while recognizing its legitimate operations with reasonably small false-positive rates.

Intelligent Multi-Level Feature Fusion Using Remote Sensing and CNN Image Classification Algorithm

The collection of fetures in both multispectral and hyperspectral domains is possible with Hyperspectral Image (HSI). It comprises a vast array of multispectral bands with functional relationships. However, they become more complex when dealing with small samples. To tackle this issue, researchers employed a deep learning convolutionary neural network system (DL-CNN) and implemented a temporal abstraction strategy to grade HSI. This approach is an intelligent multi-level feature fusion that combines the temporal abstraction strategy and DL-CNN for HSI grading. Researchers have introduced the impact of seven separate classifiers in implementing the Location estimation on our broad CNN framework, which plays the shallow CNN model's main training phase. PSO, Adagrad, Plans to implement, Alexnet, Adam, Environmental benefits, and Nadam are the seven distinct remained significantly included in this analysis. A detailed study of the four multispectral remote sensing techniques sets showed the deep CNN system's supremacy followed with the HSI identification AlexNet Optimizer.

A dual tri-path CNN system for brain tumor segmentation

The research on developing CNN-based fully-automated brain-tumor-segmentation systems has been progressing rapidly. For the systems to be applicable in practice, a good processing quality and reliability are necessary. Moreover, as the parameters in a CNN are determined by training, based on statistical losses in training epochs, more parameters may cause more randomness in the process and a minimization of the number of parameters is required to achieve a good reproducibility of the results. To this end, the CNN in the proposed system has a unique structure with 2 distinguished characters. Firstly, the three paths of its feature extraction block are designed to extract, from the multi-modality input, comprehensive feature information of mono-modality, paired-modality and cross-modality data, respectively. Also, it has a particular three-branch classification block to distinguish pixels in each of the 3 intra-tumoral classes from the background. Each branch is trained separately so that the parameters are updated specifically with the corresponding ground truth data of target tumor areas. The convolutional layers of the system are custom-designed with specific purposes, resulting in a very simple config of 61,843 parameters in total. The proposed system has been tested extensively with BraTS2019 and BraTS2018 datasets. The mean Dice scores, obtained from the ten experiments on BraTS2019 validation samples, are 0.751 ± 0.007, 0.885 ± 0.002, 0.776 ± 0.004, for enhancing tumor, whole tumor and tumor core, respectively. The test results demonstrate that the proposed system is able to reproduce a high-quality segmentation result quite consistently. Furthermore, its extremely low computation complexity will facilitate its implementation/application in various environments.

Deep learning to predict Pulmonary Tuberculosis from Lung Posterior Chest Radiographs

Tuberculosis is one of the most dangerous health conditions on the globe. As it affects the human body, tuberculosis is an infectious illness. According to the World Health Organization, roughly 1.7 million individuals get TB throughout the course of their lifetimes. Pakistan ranks fifth among high-burden nations and is responsible for 61% of the TB burden within the WHO Eastern Mediterranean Region. Various methods and procedures exist for the early identification of TB. However, all methods and techniques have their limits. The bulk of currently known approaches for detecting TB rely on model-based segmentation of the lung. The primary purpose of the proposed study is to identify pulmonary TB utilising chest X-ray (Poster Anterior) lung pictures processed using image processing and machine learning methods. The recommended study introduces a unique model segmentation strategy for TB identification. For classification, CNN, Google Net, and other systems based on deep learning are used. On merged datasets, the best accuracy attained by the suggested method utilising Google Net was 89.58 percent. The recommended study will aid in the detection and accurate diagnosis of TB.

Implementation of Spatial-Level Augmentation on Pneumonia Classification with Convolutional Neural Network

Pneumonia is a disease with symptoms of difficulty breathing, fever, dry cough, and chest pain. As an indication that a person has COVID-19, it is necessary to immediately identify pneumonia by a doctor whether the lung X-ray image of the patient is classified as pneumonia or not. The importance of early diagnosis can reduce the risk of death in patients. Convolutional Neural Network is one of the fields of study in the realm of Computer Science that can perform image-based object classification. CNN can be used to identify pneumonia based on thorax images based on the features or features displayed on the image. One of the important elements in CNN is the amount of data that can be used for training, validation, and testing. Generally, the more data entered, the more learning material for the CNN system so that the system can classify more accurately. This study aims to measure the accuracy of the CNN model on thoracic-pneumonia images with spatial level augmentation changes. Image augmentation is implemented to increase image variance with initial data of 5856 images. The applied augmentations are Affine, Flip, Pixel Dropout, Random Size Crop, and Shift Scale Rotate. The stages of this research are manually grouping images, implementing augmentation on images, applying training-validation-testing on CNN, and analyzing the output results of the developed system. By using 5 types of augmentation, the dataset used as learning material can be increased up to 5x the original amount. From the research carried out, it was found that the Random-sized Crop type augmentation gave the highest accuracy value of 94.719% or an increase of 3.808% from the non-augmentation testing data. From this research, it is hoped that studies related to augmentation can be a reference regarding the type of augmentation process and its results in finding the CNN accuracy value, especially in the case of pneumonia classification

Research on Multimodal Dance Movement Recognition Based on Artificial Intelligence Image Technology.

At present, most robot dances are precompiled. Changing music requires manual adjustment of relevant parameters and metamovements, which greatly reduces the fun and intelligence. In view of the above problems, this paper designed CNN system, studied the multimodal dance movement recognition algorithm of artificial intelligence image technology, and completed the construction of a multimodal dance movement calculation system example. The results show that the CNN algorithm and the Winograd algorithm-based coprocessor-optimized CNN network in multimodal dance movement recognition with image technology reduce from a maximum of 132s to 26s in the runtime criterion, with a maximum reduction of 80%; from a maximum of 73.5% to 16.2% in the memory access criterion, with a maximum reduction of 57.3%; and from a maximum of 93.6% to 25.2% in the power consumption ratio criterion, with a maximum reduction of 68.4%. In the power consumption ratio criterion, the maximum reduction from 93.6% to 25.2% is 68.4%. The maximum accuracy of the proposed optimization method is 95.1%. The solution is proposed to address the problem of insufficient performance of traditional dance movement recognition, which will contribute to the development of artificial intelligence and dance industry.

Aortic Enlargement Detection using Chest X-rays to Identify Potential Marfan Syndrome

Marfan syndrome is diagnosed by aortic enlargement, which can lead to complications such as aortic dilation, aortic dissection, and even rupture of the aortic wall in extreme cases. As part of this work, we developed a CNN system whose function is to interpret chest X-ray images and determine whether they show an enlarged aorta. Two architectures were studied for best hyper-parameters that affect the performance of DenseNet and Inception-v3. In most tests, the system obtained high accuracy rates, low loss rates, and no overfitting, making it a valuable tool for medical practitioners to help provide an initial diagnosis to potential patients.

Convolutional Neural Networks based classification of breast ultrasonography images by hybrid method with respect to benign, malignant, and normal using mRMR

Early diagnosis of breast lesions and differentiation of malignant lesions from benign lesions are important for the prognosis of breast cancer. In the diagnosis of this disease ultrasound is an extremely important radiological imaging method because it enables biopsy as well as lesion characterization. Since ultrasonographic diagnosis depends on the expert, the knowledge level and experience of the user is very important. In addition, the contribution of computer aided systems is quite high, as these systems can reduce the workload of radiologists and reinforce their knowledge and experience when considered together with a dense patient population in hospital conditions. In this paper, a hybrid based CNN system is developed for diagnosing breast cancer lesions with respect to benign, malignant and normal. Alexnet, MobilenetV2, and Resnet50 models are used as the base for the Hybrid structure. The features of these models used are obtained and concatenated separately. Thus, the number of features used are increased. Later, the most valuable of these features are selected by the mRMR (Minimum Redundancy Maximum Relevance) feature selection method and classified with machine learning classifiers such as SVM, KNN. The highest rate is obtained in the SVM classifier with 95.6% in accuracy.

Preface

It was a great pleasure to welcome all the participants from different countries to the 2021 6th International Symposium on Advances in Electrical, Electronics and Computer Engineering (ISAEECE 2021) in Nanjing, China during March 12-14, 2021. ISAEECE 2021 is to bring together innovative academics and industrial experts in the field of electrical, electronics and computer engineering to a common forum. The primary goal of the conference is to promote research and developmental activities in electrical, electronics and computer engineering and another goal is to promote scientific information interchange between researchers, developers, engineers, students, and practitioners working all around the world. The conference will be held every year to make it an ideal platform for people to share views and experiences in electrical, electronics and computer engineering and related areas.The conference model was divided into two sessions, including oral presentations and keynote speeches. In the first part, some scholars, whose submissions were selected as the excellent papers, were given 15 minutes to perform their oral presentations one by one. Then in the second part, keynote speakers were each allocated 30-45 minutes to hold their speeches. In this conference, we were greatly honor to have invited four outstanding researchers to serve as our Conference Chairman.During the conference, we invited three industry elites to present their insightful speeches. The first keynote speaker, Dr. Sunil Kumar Jha, from Nanjing University of Information Science and Technology, China. He presented a speech: Data Fusion Approaches in Human Body Odor Data Mining. The objective of the present research talk is to introduce audience about the data fusion and human body odor and to demonstrate research results related to search for an optimal subset of VOCs in body odor, which can produce differentiation in an individual by using the combination of analytical methods and chemometric analysis. The second keynote speaker, Prof. Chunbo Xiu, from Tiangong University, China. He shared a speech on his specialty related to Memristive Cellular Neural Network & Its Dynamic Characteristic Analysis. He demonstrated that chaotic memristive CNN system can be used in the secure communication by the chaos synchronization based on sliding mode control. Assoc. Prof. Pavel Loskot, from Zhejiang University-University of Illinois at Urbana-Champaign Institute (ZJUI), our third keynote speakers. He presented a speech: Monte Carlo Simulations Revisited. Their insightful speeches made every participant praised this conference for disseminating useful knowledge.We were glad to share with you that we received lots of submissions from the conference and we selected a bunch of high-quality papers and compiled them into the proceedings after rigorously reviewed them. These papers feature following topics but are not limited to: Electronics Engineering, Electrical Engineering, Computer Engineering and other related topics. All the papers have been through rigorous review and process to meet the requirements of International publication standard.We were grateful to the International/National advisory committee, keynote speakers, session chairs, organising committee members, student volunteers and administrative assistance of the management section of University, including accounts section, digital media and publication house. Also, we were thankful to all the authors for contributing large number of papers in the conference, because of which the conference became a story of success. It was the quality of their presentations and their passion to communicate with the other participants that really make this conference series a great success.The Organizing Committee of ISAEECE 2021List of Committee member are available in this pdf.

A multi-level method noise based image denoising using convolution neural network

Gaussian noise has been the bane of any and every denoising process under the sun. Being a very corrosive noise with huge disruptive potential, this has received much attention form the image restoration community. Building on the premise, a novel framework is proposed to leverage multi-level image denoising that iteratively removes gaussian noise while recovering details lost during processing. This framework uses existing deep learning based CNN systems whilst enhancing the same by the addition of method denoising to the process. This framework is habile in competing with state-of-the-art technologies and outperforming them in some cases.

Computer-based detection of depression and dementia in spontaneous speech

IntroductionThere is a significant relation between old-age depression and subsequent dementia in patients aged 50. This supports the hypothesis of old-age depression being a predictor, and possibly a causal factor, of subsequent dementia. The number of people aged 60 years and over has tripled since 1950, reaching 16% in 2050, leading to new medical challenges. Depression is the most common mental disorder in older adults, affecting 7% of the older population. Dementia is the second most common with about 5% prevalence worldwide, but it is the first leading cause of disease burden.ObjectivesEarly detection and treatment is essential in promoting remission, preventing relapse, and reducing emotional burden. Speech is a well established early indicator of cognitive deficits. Speech processing methods offer great potential to fully automatically screen for prototypic indicators of both dementia and depressive disorders.MethodsWe present two different methods to detect pathological speech with artificial neural networks. We use both deep architectures, as well as more traditional machine learning approaches.ResultsThe models developed using a two-stage deep architecture achieved 59% classification accuracy on the test set from DementiaBank. Our CNN system achieved the best classification accuracy of 63.6% for dementia, but reaching 70% for depressive disorders on the test set from Distress Analysis Interview Corpus.ConclusionsThese methods offer a promising classification accuracy ranging from 63% to 70%, applicable in an innovative speech-based screening system.