Use Of Convolutional Neural Networks Research Articles

Применение сверточных нейронных сетей для определения эндотелиальной дистрофии Фукса

Purpose. To evaluate the application of convolutional neural networks for the automatic detection of Fuchs' dystrophy. Material and methods. The study included 700 biomicroscopic images of the corneal endothelium (Tomey EM-3000) randomly selected from the database of the Saint-Petersburg brunch of the S. Fyodorov Eye Microsurgery Federal State Institution. At the first stage, the images were divided into 2 groups. The first group included images with the presence of Fuchs' dystrophy, the second – another pathology or a healthy cornea. The corneal endothelial cell density images were divided into three categories: training, validation, and test datasets. In our study we tested various architectures of convolutional neural networks: ResNet18, ResNet50, VGG16, VGG19 and GoogleNet. Results. The approbation of the neural network on the test sample has given the following values of the F-measure: ResNet18: 0.985; ResNet50: 1,000; VGG16: 0.940; VGG19: 0.990; GoogleNet: 0.987. Pre-trained network ResNet50 performed best with frozen layers, Adam optimizer, cross-entropy as a loss function, and a training step of 0.000005. Conclusion. The use of convolutional neural networks for the automatic detection of Fuchs' dystrophy can be successfully implemented as part of a doctor's decision support system. ResNet50 showed the best results among all types of models and did not give a single error on the test sample, which indicates the high efficiency of using this network in the classification algorithm for corneal endothelial images. Keywords: artificial intelligence, Fuchs corneal dystrophy, convolutional neural networks

Quantitative analysis of artificial intelligence on liver cancer: A bibliometric analysis.

To provide the current research progress, hotspots, and emerging trends for AI in liver cancer, we have compiled a relative comprehensive and quantitative report on the research of liver disease using artificial intelligence by employing bibliometrics in this study. In this study, the Web of Science Core Collection (WoSCC) database was used to perform systematic searches using keywords and a manual screening strategy, VOSviewer was used to analyze the degree of cooperation between countries/regions and institutions, as well as the co-occurrence of cooperation between authors and cited authors. Citespace was applied to generate a dual map to analyze the relationship of citing journals and citied journals and conduct a strong citation bursts ranking analysis of references. Online SRplot was used for in-depth keyword analysis and Microsoft Excel 2019 was used to collect the targeted variables from retrieved articles. 1724 papers were collected in this study, including 1547 original articles and 177 reviews. The study of AI in liver cancer mostly began from 2003 and has developed rapidly from 2017. China has the largest number of publications, and the United States has the highest H-index and total citation counts. The top three most productive institutions are the League of European Research Universities, Sun Yat Sen University, and Zhejiang University. Jasjit S. Suri and Frontiers in Oncology are the most published author and journal, respectively. Keyword analysis showed that in addition to the research on liver cancer, research on liver cirrhosis, fatty liver disease, and liver fibrosis were also common. Computed tomography was the most used diagnostic tool, followed by ultrasound and magnetic resonance imaging. The diagnosis and differential diagnosis of liver cancer are currently the most widely adopted research goals, and comprehensive analyses of multi-type data and postoperative analysis of patients with advanced liver cancer are rare. The use of convolutional neural networks is the main technical method used in studies of AI on liver cancer. AI has undergone rapid development and has a wide application in the diagnosis and treatment of liver diseases, especially in China. Imaging is an indispensable tool in this filed. Mmulti-type data fusion analysis and development of multimodal treatment plans for liver cancer could become the major trend of future research in AI in liver cancer.

Simulating a full-sky high resolution Galactic synchrotron spectral index map using neural networks

ABSTRACT We present a model for the full-sky diffuse Galactic synchrotron spectral index with an appropriate level of spatial structure for a resolution of 56 arcmin (to match the resolution of the Haslam 408 MHz data). Observational data at 408 MHz and 23 GHz have been used to provide spectral indices at a resolution of 5 degrees. In this work, we make use of convolutional neural networks to provide a realistic proxy for the higher resolution information, in place of the genuine structure. Our deep learning algorithm has been trained using 14.4 arcmin observational data from the 1.4 GHz Parkes radio continuum survey. We compare synchrotron emission maps constructed by extrapolating the Haslam data using various spectral index maps, of different angular resolution, with the Global Sky Model. We add these foreground maps to a total emission model for a 21-cm intensity mapping experiment, then attempt to remove the foregrounds. The different models all display different spectral or spatial behaviour and so each provide a useful and different tool to the community for testing component separation techniques. We find that for an experiment operating using a cosine aperture taper beam with a primary full width at half maximum between 1.1 and 1.6 degrees, and the principal component analysis technique of foreground removal, there is a discernible difference between synchrotron spectral index models with a resolution larger than 5 degrees but that no greater resolution than 5 degrees is required.

Automatic Detection of Indian Currency Denominations using Deep Learning

Identification of the denomination of the currency note to pay physically without UPI is the first step of paying to the seller by the consumer. In this project, we have proposed an approach to detect denominations of Indian currency using Convolutional Neural Networks. Computer Vision and object detection is an area of great interest for research in today’s world. It has several applications like detection of defects in machinery, intruder detection, computer vision for code and character recognition among many others. Through the work we have done, we explored something that could be of great help to people in day-to-day life. In this project we have tried to investigate the approaches to detect currency denominations using Convolutional Neural Networks. The objective is to build a model that would be able to detect Indian currency denominations efficiently. Typically the model will be useful for people with vision impairment. The experimental results show that the use of Convolutional Neural Networks is a good way and the model can further be improved if it is trained in such a way that it could also identify the regions of interest.

DeepInsight-3D architecture for anti-cancer drug response prediction with deep-learning on multi-omics

Modern oncology offers a wide range of treatments and therefore choosing the best option for particular patient is very important for optimal outcome. Multi-omics profiling in combination with AI-based predictive models have great potential for streamlining these treatment decisions. However, these encouraging developments continue to be hampered by very high dimensionality of the datasets in combination with insufficiently large numbers of annotated samples. Here we proposed a novel deep learning-based method to predict patient-specific anticancer drug response from three types of multi-omics data. The proposed DeepInsight-3D approach relies on structured data-to-image conversion that then allows use of convolutional neural networks, which are particularly robust to high dimensionality of the inputs while retaining capabilities to model highly complex relationships between variables. Of particular note, we demonstrate that in this formalism additional channels of an image can be effectively used to accommodate data from different omics layers while implicitly encoding the connection between them. DeepInsight-3D was able to outperform other state-of-the-art methods applied to this task. The proposed improvements can facilitate the development of better personalized treatment strategies for different cancers in the future.

Приложения компьютерного зрения в горнодобывающей промышленности

n the last decade, there has been an active digitalization of industrial production based on rapidly developing information technologies, including artificial intelligence technologies. This is largely due to the development of deep learning methods and their applications in computer vision. Since the mid 2010s convolutional neural networks demonstrate exceptional efficiency in solving problems such as the detection, classification and segmentation of various objects. As a result, computer vision methods are beginning to be actively used in the problems of quality control of raw materials and finished products. All this applies to the mining industry. However, in the Russian scientific literature there are practically no systematic reviews of computer vision applications in this area. The present study aims to fill this gap. The paper provides a systematic review of the history of development and the current state of the methods and technologies of machine vision used in the mining industry for the analysis of solid materials, demonstrates the latest achievements in this area and examples of their application in the mining industry. The authors have analyzed 29 research papers in the field of application of computer vision in the mining industry and classified the stages of technology development from the mid-1980s, when computer vision was used without the use of machine learning, and ending with modern research based on the use of deep convolutional neural networks for solving problems of classification and segmentation. The effectiveness of the methods used is compared, their advantages and disadvantages are discussed, and forecasts are made for the development of computer vision methods in the mining industry in the near future. Examples are given showing that the use of convolutional neural networks made it possible to move to a qualitatively higher level of quality in solving problems of classification and segmentation as applied to the analysis of output volume, particle size distribution, including flakiness, angularity and roughness, dust and clay content, bulk density and emptiness, etc.

The Use of Correlation Features in the Problem of Speech Recognition

The problem solved in the article is connected with the increase in the efficiency of phraseological radio exchange message recognition, which sometimes takes place in conditions of increased tension for the pilot. For high-quality recognition, signal preprocessing methods are needed. The article considers new data preprocessing algorithms used to extract features from a speech message. In this case, two approaches were proposed. The first approach is building autocorrelation functions of messages based on the Fourier transform, the second one uses the idea of building autocorrelation portraits of speech signals. The proposed approaches are quite simple to implement, although they require cyclic operators, since they work with pairs of samples from the original signal. Approbation of the developed method was carried out with the problem of recognizing phraseological radio exchange messages in Russian. The algorithm with preliminary feature extraction provides a gain of 1.7% in recognition accuracy. The use of convolutional neural networks also provides an increase in recognition efficiency. The gain for autocorrelation portraits processing is about 3–4%. Quantization is used to optimize the proposed models. The algorithm’s performance increased by 2.8 times after the quantization. It was also possible to increase accuracy of recognition by 1–2% using digital signal processing algorithms. An important feature of the proposed algorithms is the possibility of generalizing them to arbitrary data with time correlation. The speech message preprocessing algorithms discussed in this article are based on classical digital signal processing algorithms. The idea of constructing autocorrelation portraits based on the time series of a signal has a novelty. At the same time, this approach ensures high recognition accuracy. However, the study also showed that all the algorithms under consideration perform quite poorly under the influence of strong noise.

The use of convolutional neural network in the analysis of electrocardiograms

The use of convolutional neural network in the analysis of electrocardiograms

Plant Disease Diagnosing Based on Deep Learning Techniques

Agriculture crops are highly significant for the sustenance of human life and act as an essential source for national income development worldwide. Plant diseases and pests are considered one of the most imperative factors influencing food production, quality, and minimize losses in production. Farmers are currently facing difficulty in identifying various plant diseases and pests, which are important to prevent plant diseases effectively in a complicated environment. The recent development of deep learning techniques has found use in the diagnosis of plant diseases and pests, providing a robust tool with highly accurate results. In this context, this paper presents a comprehensive review of the literature that aims to identify the state of the art of the use of convolutional neural networks (CNNs) in the process of diagnosing and identification of plant pest and diseases. In addition, it presents some issues that are facing the models performance, and also indicates gaps that should be addressed in the future. In this regard, we review studies with various methods that addressed plant disease detection, dataset characteristics, the crops, and pathogens. Moreover, it discusses the commonly employed five-step methodology for plant disease recognition, involving data acquisition, preprocessing, segmentation, feature extraction, and classification. It discusses various deep learning architecture-based solutions that have a faster convergence rate of plant disease recognition. From this review, it is possible to understand the innovative trends regarding the use of CNN’s algorithms in the plant diseases diagnosis and to recognize the gaps that need the attention of the research community.

Single 24-GHz FMCW Radar-Based Indoor Device-Free Human Localization and Posture Sensing With CNN

The position information and posture information of a device-free human in a confined indoor environment have multiple uses in health monitoring and other areas. In this study, we aim to use a single-input-single-output (SISO) frequency-modulated continuous-wave (FMCW) radar at 24-GHz band for simultaneous localization and posture estimation of one device-free human target. This device is easy to deploy in new setups. We use image formation to convert temporal measurements of the radar signal into image-like data and convert the problem of simultaneous position estimation and posture perception into an image classification problem. Leveraging the use of convolutional neural networks (CNNs) for image classification, we design a variety of tests to explore the most favorable parameters of the CNN model and then use the best practice model to accomplish the classification task involving a fusion of localization and pose recognition. To explain the primary causes of inaccuracies, we examine not only cases based on a fused position and posture dataset but also cases based on position-only and posture-only datasets. On both real-life datasets, our proposed scheme can achieve 98% classification accuracy and less than 1-m localization accuracy within 0.95 cumulative error probability within the area of interest, outperforming traditional classification methods.

Identifying pregnancy in cows using ovarian ultrasound images and convolutional neural networks - a proof-of-concept study

Visual analysis of ovarian structures by ultrasound in cows is a relevant support tool for improving reproductive performance in livestock. However, with human visual analysis, subjectiveness is a limiting factor, and computer vision technologies are a way to overcome this. We thus aimed to evaluate the use of convolutional neural networks (CNNs) for identifying pregnancy in cows using ovarian ultrasound images obtained 30 days after artificial insemination. The dataset consisted of 510 images from individual independent functional ovaries of 238 pregnant and 272 non-pregnant Nellore cows. All the images were collected from the same commercial farm. To evaluate the dependency of CNN performance on image quality, the images were classified by two independent veterinarians with significant experience in ultrasound evaluation as good, regular, or bad for visual identification of ovarian structures. Four CNN architectures were evaluated, namely ResNet50, ResNeXt50, InceptionResNetV2, and DenseNet121. All the CNNs were evaluated using the complete dataset (ALL; n = 510) and a subset of good and regular images (GR; n = 462). Ten evaluations were conducted for each architecture and dataset combination. In each run, the datasets were randomly divided into 70 % for training and 30 % for testing according to the holdout method. A regularization-based method was used during the training phase using dropout, data augmentation, and a dynamic learning rate. The model’s performance was evaluated in terms of accuracy, precision, sensitivity, and specificity, which ranged from 46.5 to 78.8, 34.4–80.0, 28.3–75.4, and 11.2–82.2 %, respectively. The DenseNet121 architecture performed better based on higher accuracy, sensitivity, and specificity, and both lower SD and variation due to dataset. ResNet50 and DenseNet121 models performed better when using the GR dataset. InceptionResNetV2 had a more precise architecture when using the ALL dataset but had lower performance for the GR dataset. We concluded that CNNs are a promising tool to identify the pregnancy status of cows in ovarian images collected 30 days after artificial insemination. The effect of image quality on the performance depends on the CNN architecture.

Deep learning applications for stellar parameter determination: II-application to the observed spectra of AFGK stars

Abstract In this follow-up article, we investigate the use of convolutional neural network for deriving stellar parameters from observed spectra. Using hyperparameters determined previously, we have constructed a Neural Network architecture suitable for the derivation of T eff {T}_{{\rm{eff}}} , log g \log g , [ M / H ] \left[M\hspace{0.1em}\text{/}\hspace{0.1em}H] , and v e sin i {v}_{e}\sin i . The network was constrained by applying it to databases of AFGK synthetic spectra at different resolutions. Then, parameters of A stars from Polarbase, SOPHIE, and ELODIE databases are derived, as well as those of FGK stars from the spectroscopic survey of stars in the solar neighbourhood. The network model’s average accuracy on the stellar parameters is found to be as low as 80 K for T eff {T}_{{\rm{eff}}} , 0.06 dex for log g \log g , 0.08 dex for [ M / H ] \left[M\hspace{0.1em}\text{/}\hspace{0.1em}H] , and 3 km/s for v e sin i {v}_{e}\sin i for AFGK stars.

Pre-computation of image features for the classification of dynamic properties in breaking waves

ABSTRACT The use of convolutional neural networks (CNNs) in image classification has become the standard method of approaching many computer vision problems. Here we apply pre-trained networks to classify images of non-breaking, plunging and spilling breaking waves. The CNNs are used as basic feature extractors and a classifier is then trained on top of these networks. The dynamic nature of breaking waves is exploited by using image sequences extracted from videos to gain extra information and improve the classification results. We also see improved classification performance by using pre-computed image features such as the Optical Flow (OF) between image pairs to create new models in combination with infra-red images with reduction in errors of up to 60%. The inclusion of this dynamic information improves the classification between breaking wave classes. We also provide corrections to a methodology in the literature from which the data originates to achieve a more accurate assessment of model performance.

Convolutional Neural Network Predictions for Unsteady Reynolds-Averaged Navier–Stokes-Based Numerical Simulations

The application of computational fluid dynamics (CFD) to turbulent flow has been a considerable topic of research for many years. Nonetheless, using CFD tools results in a large computational cost, which implies that, for some applications, CFD may be unviable. To date, several authors have carried out research applying deep learning (DL) techniques to CFD-based simulations. One of the main applications of DL with CFD is in the use of convolutional neural networks (CNNs) to predict which samples will have the desired magnitude. In this study, a CNN which predicts the streamwise and vertical velocities and the pressure fields downstream of a circular cylinder for a series of time instants is presented. The CNN was trained using a signed distance function (SDF), a flow region channel (FRC) and the t-1 sample as inputs, and the ground-truth CFD data as the output. The results showed that the CNN was able to predict multiple time instants with low error rates for turbulent flows with variable input velocities to the domain.

Auditory stimulation and deep learning predict awakening from coma after cardiac arrest.

Assessing the integrity of neural functions in coma after cardiac arrest remains an open challenge. Prognostication of coma outcome relies mainly on visual expert scoring of physiological signals, which is prone to subjectivity and leaves a considerable number of patients in a 'grey zone', with uncertain prognosis. Quantitative analysis of EEG responses to auditory stimuli can provide a window into neural functions in coma and information about patients' chances of awakening. However, responses to standardized auditory stimulation are far from being used in a clinical routine due to heterogeneous and cumbersome protocols. Here, we hypothesize that convolutional neural networks can assist in extracting interpretable patterns of EEG responses to auditory stimuli during the first day of coma that are predictive of patients' chances of awakening and survival at 3 months. We used convolutional neural networks (CNNs) to model single-trial EEG responses to auditory stimuli in the first day of coma, under standardized sedation and targeted temperature management, in a multicentre and multiprotocol patient cohort and predict outcome at 3 months. The use of CNNs resulted in a positive predictive power for predicting awakening of 0.83 ± 0.04 and 0.81 ± 0.06 and an area under the curve in predicting outcome of 0.69 ± 0.05 and 0.70 ± 0.05, for patients undergoing therapeutic hypothermia and normothermia, respectively. These results also persisted in a subset of patients that were in a clinical 'grey zone'. The network's confidence in predicting outcome was based on interpretable features: it strongly correlated to the neural synchrony and complexity of EEG responses and was modulated by independent clinical evaluations, such as the EEG reactivity, background burst-suppression or motor responses. Our results highlight the strong potential of interpretable deep learning algorithms in combination with auditory stimulation to improve prognostication of coma outcome.

Survey on Detection of Manipulated Multimedia in Digital Forensics Using Machine Learning

The manipulation of multimedia has increased all over the world. Different tools are used to alter the multimedia and it is difficult to detect genuine and fake media. People are facing problems to detect if the media is real or fake. Due to manipulated media, cybercrime has becomeincreasingly widespread. We believe that personal security and privacy should be carried out easily and intelligently in this digital environment where all fundamental tasks are completed without issue. When we looked into the numbers, we discovered that a sizable proportion of people experience harassment or other forms of abuseregularly. Based on a review of the existing system, we presented an application that would use the CNN (Convolutional Neural Networks) method to distinguish between real and fraudulent media in a single application. CNN performs better with picture and voice or audio inputs than earlier networks and other techniques. CNN hidden extract feature from the input using pixels value and computation based on edges and outline of the inputs using pixels value and computation based on edges and outline of the inputs. The growing use of convolutional neural networks (CNNs) has had a substantial effect on defenders.

Portable motorized telescope system for wind turbine blades damage detection

AbstractWind turbines are among the fastest‐growing sources of energy production and the maintenance operations include regular inspection of their blades, causing considerable downtime and cost. In addition, the manual inspection process involves a great risk. To address this challenge, in this article a preventive maintenance system for wind turbines based on deep computational learning techniques is presented. This open‐source project aims to detect and classify possible surface damages on wind turbine blades to facilitate and improve the inspection of such infrastructures. The system consists of a stand‐alone Android application that makes use of convolutional neural networks for image processing, a portable telescope to take precise photographs of the turbine blades, and a motorized mount that allows the movement of the telescope. The application tries to carry out a complete sweep of the surface of the wind turbine blades in an autonomous way based on the predictions of neural network models and finally presents the defects found to the user. Thanks to this, maintenance time would be reduced and the risk of manual intervention would be avoided. Accuracies of around 97% for label predictions and 90% for bounding box coordinate predictions have been achieved on the validation dataset. The proposed low‐cost inspection system for detecting surface damages on blades has been experimentally validated in a real wind farm.

Pavement Crack Monitoring: Literature Review

Crack monitoring of pavements is an ever-evolving technology with new crack identification technologies being introduced frequently. Although older technologies consisted of physical removing the pavement section using coring, however new methods are available that are non-destructive and yield a higher performance than conventional technologies. This paper compiles various crack monitoring technologies such as wireless sensor networks, photo imaging, laser imaging, 3D road surface profile scans, acoustics wave propagation technology, embedded strain sensors and onboard vehicle sensors that majorly use an artificial intelligence algorithm to identify and categorize the cracks. The research also includes the use of convolutional neural network that can be used to analyze pavement images and such neural network can localize and classify the cracks for crack initiation and propagation stage. The research concludes with the favor of using the optical imaging technology called Syncrack which serves better performance in terms of time of prediction by 25% and accuracy by 30% when compared to other sensing technologies.

Applications of CNN in leaf diseases: A critical survey

Crop diseases can significantly impact crop yield and overall productivity, posing challenges for farmers in increasing output and market prices. Early detection of these diseases is crucial for preventing further spread and reducing their impact. To overcome this, researchers have utilized image processing technology, including deep learning techniques such as convolutional neural networks (CNNs), to detect crop diseases. In this critical survey, we provide a comprehensive review of recent studies and developments in the use of CNNs for identifying leaf diseases in agricultural plants. We discuss the benefits and drawbacks of different deep learning techniques and image processing methods for disease diagnosis and management in agriculture. Our research highlights the potential of CNNs and deep learning to significantly advance the field of agricultural research and development. We also analyze the factors affecting the outcomes of each technique, including the accuracy, precision. Our study emphasizes the need for further research and development to optimize the use of CNNs in agricultural applications, particularly for improving disease management and crop productivity.

Age assessment using CT of knee joint and neural network technologies

Age assessment of living persons plays an important role in clinical and sports medicine, as well as in law practice. Traditional methods have a number of problems: age restrictions, technical difficulties of visualization, low reproducibility and subjectivity of estimation. The proposed approach, which implies the use of multispiral computed tomography (MSCT) and database mining, will eliminate these drawbacks and improve the estimation of age. The aim of the study was to investigate the use of deep learning algorithms to classify the age groups (with a threshold level of 18 years) for CT images of knee joint. The study included 455 MSCT images of the knee joint of male and female subjects aged from 13 to 24. The method included score assessment of the distal femur's epiphyseal synostosis stages, tibia and fibula proximal epiphyses and a preliminary statistical analysis of correlations between age and stages of synostosis. The challenge of binary classification of target age groups with the use of convolutional neural networks was implemented at the second phase of the trial. Various architectures of convolutional neural networks and their ensembles were tested. The result of the study showed that the total score of epiphyseal synostosis has the highest correlation with the age (r=0.88). The proposed method of chronological age assessment on the basis of the knee area CT images research using deep learning algorithms demonstrated a good result. The classification accuracy (threshold level of 18 years) was 0.86.