Search

Learning styles cover various attributes related to the attitude and the learning behavior of individuals. Research and educational theories confirm that considering learning styles in distance learning environments can improve academic performance and learner satisfaction. The traditional approach to identify learning styles is based on asking students to fill out a questionnaire. This approach is considerably less accurate due to the learners’ lack of awareness of their own preferences. Furthermore, learners’ learning styles are defined only once. In this study, we propose an automatic approach to identify learners’ learning styles based on patterns of learning behavior with respect to Felder and Silverman Learning Style Model (FSLSM), in an online learning environment. Patterns of behavior were analysed based on a data-driven approach. This approach exploits different Machine Learning (ML) techniques to detect the learning styles of learners. To validate our proposals, experiments were carried out in a higher education institution with 73 students enrolled in online courses on the ADLS (Automatic Detection of Learning Styles) system that we implemented. A 9 runs cross-validation was used to evaluate the selected ML techniques. Detection accuracy, recall, precision, and F measure were observed. The findings show the possibility of detecting learning styles automatically based on learning behavior with high performances. Different levels of accuracy were found for the different dimensions of FSLSM. However, Support Vector Machines (SVM) have exhibited great ability in predicting learning styles for all dimensions of FSLSM with an accuracy average of 88%.

Automatic fault detection (AFD) is a key technology to optimize the Operation and Maintenance of photovoltaic (PV) systems portfolios. A very common approach to detect faults in PV systems is based on the comparison between measured and simulated performance. Although this approach has been explored by many authors, due to the lack a common basis for evaluating their performance, it is still unclear what are the influencing aspects in the design of AFD algorithms. In this study, a series of AFD algorithms have been tested under real operating conditions, using monitoring data collected over 58 months on 80 rooftop-type PV systems installed in Germany. The results shown that this type of AFD algorithm have the potential to detect up to 82.8% of the energy losses with specificity above 90%. In general, the higher the simulation accuracy, the higher the specificity. The use of less accurate simulations can increase sensitivity at the cost of decreasing specificity. Analyzing the measurements individually makes the algorithm less sensitive to the simulation accuracy. The use of machine learning clustering algorithm for the statistical analysis showed exceptional ability to prevent false alerts, even in cases where the modeling accuracy is not high. If a slightly higher level of false alerts can be tolerated, the analysis of daily PR using a Shewhart chart provides the high sensitivity with an exceptionally simple solution with no need for more complex algorithms for modeling or clustering., 33 pages, 30 figures, preprint submitted to Elsevier Solar Energy

In the present report, we evaluated adrenergic mechanisms of generalized spike-wave epileptic discharges (SWDs), which are the encephalographic hallmarks of idiopathic generalized epilepsies. SWDs link to a hyper-synchronization in the thalamocortical neuronal activity. We unclosed some alpha2-adrenergic mechanisms of sedation and provocation of SWDs in rats with spontaneous spike-wave epilepsy (WAG/Rij and Wistar) and in control non-epileptic rats (NEW) of both sexes. Dexmedetomidine (Dex) was a highly selective alpha-2 agonist (0.003–0.049 mg/kg, i.p.). Injections of Dex did not elicit de novo SWDs in non-epileptic rats. Dex can be used to disclose the latent form of spike-wave epilepsy. Subjects with long-lasting SWDs at baseline were at high risk of absence status after activation of alpha2- adrenergic receptors. We create the concept of alpha1- and alpha2-ARs regulation of SWDs via modulation of thalamocortical network activity. Dex induced the specific abnormal state favorable for SWDs—“alpha2 wakefulness”. Dex is regularly used in clinical practice. EEG examination in patients using low doses of Dex might help to diagnose the latent forms of absence epilepsy (or pathology of cortico-thalamo-cortical circuitry).

In this paper, we present the implementation of a system that identifies the modulation of complex radio signals. This is realized using an artificial intelligence model developed, trained, and integrated with Microsoft Azure cloud. We consider that cloud-based platforms offer enough flexibility and processing power to use them instead of conventional computers for signal processing based on artificial intelligence. We tested the implementation using a software-defined radio platform developed in GNU Radio that generates and receives real modulated signals. This process ensures that the solution proposed is viable to be used in real signal processing systems. The results obtained show that for certain modulation types, the identification is performed with a high degree of success. The use of a cloud-based platform allows quick access to the system. The user is able to identify the signal modulation using only a laptop that has access to the internet.

This study aims to evaluate the diagnostic accuracy of artificial intelligence in detecting apical pathosis on periapical radiographs. A total of twenty anonymized periapical radiographs were retrieved from the database of Poznan University of Medical Sciences. These radiographs displayed a sequence of 60 visible teeth. The evaluation of the radiographs was conducted using two methods (manual and automatic), and the results obtained from each technique were afterward compared. For the ground-truth method, one oral and maxillofacial radiology expert with more than ten years of experience and one trainee in oral and maxillofacial radiology evaluated the radiographs by classifying teeth as healthy and unhealthy. A tooth was considered unhealthy when periapical periodontitis related to this tooth had been detected on the radiograph. At the same time, a tooth was classified as healthy when no periapical radiolucency was detected on the periapical radiographs. Then, the same radiographs were evaluated by artificial intelligence, Diagnocat (Diagnocat Ltd., San Francisco, CA, USA). Diagnocat (Diagnocat Ltd., San Francisco, CA, USA) correctly identified periapical lesions on periapical radiographs with a sensitivity of 92.30% and identified healthy teeth with a specificity of 97.87%. The recorded accuracy and F1 score were 96.66% and 0.92, respectively. The artificial intelligence algorithm misdiagnosed one unhealthy tooth (false negative) and over-diagnosed one healthy tooth (false positive) compared to the ground-truth results. Diagnocat (Diagnocat Ltd., San Francisco, CA, USA) showed an optimum accuracy for detecting periapical periodontitis on periapical radiographs. However, more research is needed to assess the diagnostic accuracy of artificial intelligence-based algorithms in dentistry.

A meteoroid impact against the lunar surface at speeds exceeding several kilometers per second generates a light flash less than 0.1 seconds in duration. Observers record the Moon’s nightside videos for a few hours, then search for the flashes in the video files by a software tool. Recently, some observers record the videos in SER format that allows 16-bit grayscale. However, there has been no tool for automatic search in the SER videos. Therefore, we developed a software tool for detecting lunar impact flashes in monochromatic SER videos. We describe the method for detection and how to use the tool. The tool, its source code, and sample videos are available on the Internet.

月面のクレータの単位面積当たりの多寡により，月面がいつ出来たか，あるいはいつ熔岩によって被覆されたかを調べることが可能である．したがって，クレータの検知（数え上げ）は，惑星科学にとって重要である．日本の月周回探査セレーネ（SELENE；愛称「かぐや」）に搭載された地形カメラによって得られたデータから，月全球の10m/画素の月面の可視画像データ（オルソ画像）と標高データ（数値地形/標高モデル；DTM/DEM）が得られている．本稿では，これらのデータを使って月面のクレータを自動検知する手法を紹介する．この手法では，まずオルソ画像についてクレータの影・日向を強調し，そこから，クレータ候補を抽出，その後，クレータ候補の傾斜パターンから，クレータ候補を絞ることで，クレータ検知の精度を上げるというものである．, It is possible to investigate when the lunar surface was formed or when it was covered with lava by the amount of lunar craters per unit area. Theus, crater detection (counting) is important for planetary science. Based on the data obtained by the Terrain Camera onboard the Japanese lunar orbiter SELENE (nicknamed ''Kaguya''), visible ortho-rectified image data and elevation data (Digital Terrain Model or Digital Elevation Model; DTM or DEM) have been produced. In this paper, we will introduce a method for automatically detecting lunar craters using these data. In this method, shadowed and illuminated areas of possible craters are emphasized in the ortho-rectified images, the crater candidates are extracted from them, and then they are narrowed down from the patterns of their inner walls, improving the accuracy of crater detection., 形態: カラー図版あり, Physical characteristics: Original contains color illustrations, 資料番号: AA2130018000, レポート番号: JAXA-RM-21-007

International audience; In this study, we develop a novel method to automatically detect areas of snow avalanche debris using a color space segmentation technique applied to SAR image time series through January 2018 in the Swiss Alps. Debris avalanche zones are detected assuming that these areas are characterised by a significant and localised increase in SAR signal relative to the surrounding environment. We undertake a sensitivity study by calculating debris products by varying the D-M reference images (a stable reference image taken several weeks before the event). We examine the results according to the direction of the orbit, the characteristics of the terrain (slope, altitude, orientation) and also by evaluating the relevance of the detection with the help of an independent SPOT database ([1]) including 18,737 avalanche events. Small avalanches are not detected by SAR images and depending on the orientation of the terrain some avalanches are not detected by either the ascending or the descending orbit. The detection results vary with the reference image; best detection results are obtained with some selected individual dates with almost 70 % of verified avalanche events using the ascending orbit.

International audience; The use of CNN and segmentation to extract image features for the prediction of four postures for sows kept in crates was examined. The extracted features were used as input variables in an SVM classification method to estimate posture. The possibility of using a posture prediction model with images not necessarily obtained under the same conditions as those used for the training set was explored. As a reference case, the efficacy of the posture prediction model was explored when training and testing datasets were built using the same pool of images. In this case, all the models produced satisfactory results, with a maximum f1-score of 97.7% with CNNs and 93.3% with segmentation. To evaluate the impact of environmental variations, the models were trained and tested on different monitoring days. In this case, the best f1-score dropped to 86.7%. The impact of using the posture prediction model on animals that were not present in the training dataset was then explored. The best f1-score reduced to 63.4% when the posture prediction models were trained on one animal and tested on 11 other different animals. Conversely, when the models were tested on one animal and trained on the 11 others, the f1-score only decreased to 86% with the best model. On average, a decrease of around 17% caused by environmental and individual variations between training and testing was observed.; Nous avons comparé l’utilisation des CNN et de la segmentation pour extraire des caractéristiques d’intérêts dans les images (features), afin de prédire la posture de truies allaitantes. Une fois extraite, les features peuvent être utilisées comme variable d’entrée d’une méthode de classification, de type SVM, pour prédire la posture de la truie. Nous avons exploré l’impact de l’ensemble d’apprentissage sur la qualité de la prédiction, notamment lorsque l’ensemble d’apprentissage et de test ne sont pas obtenues dans les mêmes conditions. Nous avons d’abord considéré un cas de référence, où la méthode d’estimation de la posture était entrainée et testée sur des images prises dans les mêmes conditions. Dans ce cas, tous les modèles testés ont fournis des résultats satisfaisants, avec un f1-score de 97.7% pour le meilleur CNN et de 93.3% pour la méthode basée sur la segmentation d’images. Pour évaluer l’impact des changements environnementaux lors de la prise de vue des images, nous avons ensuite entrainé et testé les modèles sur des images provenants de deux jours différents. Dans ce cas, le meilleur f1-score est réduit à 86.7%. Enfin, nous avons considéré le cas où des animaux présent dans l’ensemble test, ne le sont pas dans l’ensemble d’apprentissage. Lorsque le modèle de prediction de posture est entrainé sur une truie, et testé sur les 11 truies, le meilleur f1-score obtenu est de 63.4%. Lorsque le modèle est entrainé sur 11 truies et testé sur 1 truie, le f1-score est de 86%. En moyenne, nous avons observé que les variations environnementales et individuelles faisaient baisser le f1-score de 17%.

Obstructive Sleep Apnea (OSA) is a common sleep-breathing disorder that highly reduces the quality of human life. The most powerful method for the detection and classification of sleep apnea is the Polysomnogram. However, this method is time-consuming and cost-inefficient. Therefore, several methods focus on using electrocardiogram (ECG) signals to detect sleep apnea. This paper proposed a novel automated approach to detect and classify apneic events from single-lead ECG signals. Wavelet Scattering Transformation (WST) was applied to the ECG signals to decompose the signal into smaller segments. Then, a set of features, including higher-order statistics and entropy-based features, was extracted from the WST coefficients to formulate a search space. The obtained features were fed to a random forest classifier to classify the ECG segments. The experiment was validated using the 10-fold and hold-out cross-validation methods, which resulted in an accuracy of 91.65% and 90.35%, respectively. The findings were compared with different classifiers to show the significance of the proposed approach. The proposed approach achieved better performance measures than most of the existing methodologies.

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies Wildfires are one of the world's most costly and deadly natural disasters, damaging millions of hectares of vegetation and threatening the lives of people and animals. The risks to civilian agents and task forces are particularly high, which emphasizes the value of leveraging technology to minimize their impacts on nature and people. The use of drone imagery coupled with deep learning for automated fire detection can provide new solutions to this problem, limiting the damage that result. In this context, our work aims to implement a solution for the automatic detection of forest fires in real time by exploiting convolutional neural networks (CNN) on drone images based on classification and segmentation models. The methodological approach followed in this study can be broken down into three main steps: First, the comparison of two models, namely Xception Network and EfficientNetB2, for the classification of images captured during a forest burn into 'Fire' or 'No_Fire' classes. Then we will proceed to the segmentation of the images belonging to the 'Fire' class by comparing the U-Net architecture with Attention U-Net and Trans U-Net in order to choose the best performing model. The EfficientNetB2 architecture for classification gave satisfactory results with an accuracy of 71.72%. Concerning segmentation, we adopted the U-Net model which offers a segmentation accuracy that reaches 98%. As for the deployment, a fire detection application was designed using Android Studio software by assimilating the drone's camera.

Foodborne illnesses are one of the most severe and prevalent health problems in the world. Thus, achieving the rapid and accurate identification of foodborne pathogens is important. This study presents an automatic device to perform a multistep immunoassay on a lateral flow immunoassay strip to detect foodborne pathogens from fresh lettuce. The device is automatically operated using a smartphone application that we developed, which allows users to quantify the detection results. In this study, we characterize the device’s limit of detection and demonstrate the detection and quantification of Escherichia coli O157:H7 from contaminated lettuce. We then compare the quantified result to that calculated by counting colonies from agar plates. The device is capable of detecting contamination in lettuces that have as low as 5 × 104 Escherichia coli O157:H7 per 10 g.

Möjligheten för individer att uttrycka sig på internet har underlättat för det fria ordet, som är en demokratisk grundsten i vårt samhälle. Baksidan av detta mynt är att alltmer våldsam radikalisering sker runt om på digitala plattformar. I digitala miljöer sprids idag våldsbejakande propaganda där våldsverkare hyllas som hjältar, våld mot fienden rättfärdigas, och instruktioner om hur attentat kan genomföras delas. För socialt utsatta individer som får sin verklighetsförankring i dessa våldsbejakande digitala miljöer kan världsuppfattningen ändras till den grad att de till slut själva väljer att utföra grova våldsdåd. Att kunna identifiera digitala plattformar där diskussioner om våld är mer vanligt förekommande kan därför ge en första indikation på vilka sidor som riskerar att potentiellt främja våldsutövning. Tidigare forskning om hotbedömning genom textanalys har främst fokuserat på att identifiera individer som utgör ett hot. Mindre utrymme har ägnats åt att utveckla metoder som istället kan identifiera gemenskaper som utgör ett hot för individen. Det saknas idag applicerbara och validerade metoder som genom automatiserad textanalys kan mäta diskussioner om våld på digitala forum. Arbetets forskningsmål var att skapa en metod som ska kunna mäta förekomsten av våld i diskussioner på digitala plattformar, och tillämpar metodramverket designforskning. Genom både kvalitativa och kvantitativa metoder skapades en ordlista över en mängd våldsuttryck som används på flera samtida sociala medier. Programkod utvecklades för att automatiskt kunna räkna antalet förekommande våldsuttryck i en given textsamling. Genom tidigare tillgänglig data från flera olika forum utfördes en testning och utvärdering av metoden. Resultaten visar att våldsuttryck var upp till 100 gånger mer vanligt förekommande på vissa av de mer kända högerextrema forumen jämfört med mer generella diskussionsforum. Spektrumet är i linje med vad som initialt skulle kunna förväntas utifrån karaktären på dessa olika forum och indikerar därmed att metoden levererar realistiska resultat. En djupare kvalitativ analys av inläggen skulle vara nödvändig för att identifiera hur stor del av de identifierade våldsuttrycken tas upp i diskussioner med en positiv inställning till våld. Individuals being able to express themselves on the internet has been a boon to free speech, a democratic pillar to our community. The backside of this is that an increasing amount of violent radicalization is happening all over social media. At this moment, propaganda that praises violence is being spread on digital platforms, where violent perpetrators are praised as heroes, violence against "the enemy" is justified, and instructions on how to perform violent attacks are being spread. Socially vulnerable individuals repeatedly exposed to violent social communities can have their worldview change so drastically that they end up committing violent crimes. That is why identifying digital platforms where discussions about violence are more commonly occurring could give a first indication of which sites pose a higher risk of promoting violent attacks. Previous research on threat assessment through text analysis has mainly focused on detecting warning behaviours in radicalised individuals. Less room within this area of research has been given to developing methods that instead identify communities that can pose a risk to the individual. Therefore, there is a lack of applicable and validated methods that can measure discussions of violence on social media through automated text analysis. The research goal of this thesis was to create a method that can measure the occurrence of violent discussions on digital platforms using the method framework Design Science. A list of violent words and expressions commonly used on social media was created with qualitative and quantitative methods. Code was then developed to automatically count the number of occurrences of these expressions in a given text. Testing and evaluation of the method were carried out with data previously made available from several forums. The results show that violent expressions occurred up to 100 times more often on some of the more known right-wing extremist forums than on more general discussion forums. The resulting spectrum aligns with what one initially could expect to find by judging the character of these forums and therefore indicates that the method delivers accurate results. A deeper qualitative analysis of the posts would be necessary to identify how many of the identified violent expressions appeared in discussions with a positive attitude toward violence.

Many studies have proven that urban greenness is an important factor when cyclists choose a route. Thus, detecting trees along a cycling route is a major key to assessing the quality of cycling routes and providing further arguments to improve ridership and the better design of cycling routes. The rise in the use of video recordings in data collection provides access to a new point of view of a city, with data recorded at eye level. This method may be superior to the commonly used normalized difference vegetation index (NDVI) from satellite imagery because satellite images are costly to obtain and cloud cover sometimes obscures the view. This study has two objectives: (1) to assess the number of trees along a cycling route using software object detection on videos, particularly the Detectron2 library, and (2) to compare the detected canopy on the videos to other canopy data to determine if they are comparable. Using bicycles installed with cameras and GPS, four participants cycled on 141 predefined routes in Montréal over 87 h for a total of 1199 km. More than 300,000 images were extracted and analyzed using Detectron2. The results show that the detection of trees using the software is accurate. Moreover, the comparison reveals a strong correlation (>0.75) between the two datasets. This means that the canopy data could be replaced by video-detected trees, which is particularly relevant in cities where open GIS data on street vegetation are not available.

According to research, classifiers and detectors are less accurate when images are blurry, have low contrast, or have other flaws which raise questions about the machine learning model’s ability to recognize items effectively. The chest X-ray image has proven to be the preferred image modality for medical imaging as it contains more information about a patient. Its interpretation is quite difficult, nevertheless. The goal of this research is to construct a reliable deep-learning model capable of producing high classification accuracy on chest x-ray images for lung diseases. To enable a thorough study of the chest X-ray image, the suggested framework first derived richer features using an ensemble technique, then a global second-order pooling is applied to further derive higher global features of the images. Furthermore, the images are then separated into patches and position embedding before analyzing the patches individually via a vision transformer approach. The proposed model yielded 96.01% sensitivity, 96.20% precision, and 98.00% accuracy for the COVID-19 Radiography Dataset while achieving 97.84% accuracy, 96.76% sensitivity and 96.80% precision, for the Covid-ChestX-ray-15k dataset. The experimental findings reveal that the presented models outperform traditional deep learning models and other state-of-the-art approaches provided in the literature.

International audience; Weaning a patient from mechanical ventilation is a critical task in Intensive Care Units, but it can be made safer by using Lung Ultrasound scoring. This scoring is currently done visually by specialists based on Lung Ultrasound artifacts among which are A-lines. Automating this scoring may help standardizing results and saving time for clinicians. In this paper, we propose a method to automatically detect A-lines from a manual delineation of the pleural line, and by using both the intensity profile of the LUS image and morphological operations. A score is then assigned to significant lines and represents the possibility of them being A-lines. The proposed method shows promising results in differentiating A-lines from other elements with an Area Under the Curve of 0.95; furthermore, using a threshold at 0.5 to detect A-lines leads to very good performances with an accuracy of 95% and a F0.5-score of 0.84.

The Loess Plateau is an ecologically fragile area in China; furthermore, loess landslides are typical forms of geological disasters, which severely limit the sustainable development of the local societies and the economy. Studying the automatic detection of landslides can facilitate disaster prevention and mitigation in the Loess Plateau, and help realize the climate action goal (SDG 13) of the United Nations Sustainable Development Goals (SDGs). This paper takes typical loess areas in China as the research object, and establishes a historical loess landslide sample database based on Google Earth (GEE) image data, with a total of 1451 loess landslides. The automatic detection of loess landslides is implemented by improving the You Only Look Once X (YOLOX) algorithm. The results show that the average accuracy of landslide detection in this method is 95.43%, and the accuracy rate is 96.32%, which effectively combines the earth’s big data to realize the automatic detection of loess landslides. The research results provide technical support for the promotion of disaster prevention and mitigation in China’s loess regions, the realization of sustainable development goals, and the improvement of natural disaster prevention–resistance–reduction systems.

The novel coronavirus (COVID-19) could be described as the greatest human challenge of the 21st century. The development and transmission of the disease have increased mortality in all countries. Therefore, a rapid diagnosis of COVID-19 is necessary to treat and control the disease. In this paper, a new method for the automatic identification of pneumonia (including COVID-19) is presented using a proposed deep neural network. In the proposed method, the chest X-ray images are used to separate 2–4 classes in 7 different and functional scenarios according to healthy, viral, bacterial, and COVID-19 classes. In the proposed architecture, Generative Adversarial Networks (GANs) are used together with a fusion of the deep transfer learning and LSTM networks, without involving feature extraction/selection for classification of pneumonia. We have achieved more than 90% accuracy for all scenarios except one and also achieved 99% accuracy for separating COVID-19 from healthy group. We also compared our deep proposed network with other deep transfer learning networks (including Inception-ResNet V2, Inception V4, VGG16 and MobileNet) that have been recently widely used in pneumonia detection studies. The results based on the proposed network were very promising in terms of accuracy, precision, sensitivity, and specificity compared to the other deep transfer learning approaches. Depending on the high performance of the proposed method, it can be used during the treatment of patients. This work was supported by Research and Innovation Management Center (PPPI) and Faculty of Engineering, Universiti Malaysia Sabah (UMS) under VOT (TBP0002). Publisher's Version

BACKGROUND Captures of codling moth, Cydia pomonella (L.), in traps are used to establish action thresholds and time insecticide sprays. The need for frequent trap inspections in often remote orchards has created a niche for remote sensing smart traps. A smart trap baited with a five‐component pheromone‐kairomone blend was evaluated for codling moth monitoring among an assemblage of other nontargets in apple and pear orchards. RESULTS Codling moth captures did not differ between the smart trap and a standard trap when both were checked manually. However, the correlation between automatic and manual counts of codling moth in the smart traps was low, R 2 = 0.66 ÷ 0.87. False‐negative identifications by the smart trap were infrequent, Cydia pomonella monitoring with smart traps baited with pheromone‐kairomone lures enabled remote detection of the target species, but also of other pests, including Grapholita molesta, Synanthedon myopaeformis, and Cacoecimorpha pronubana. © 2021 The Authors. Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

Male genitalia are usually extremely divergent between closely related species, but relatively constant within one species. Here we examine the effect of temperature on the shape of the ventral branches, a male genital structure involved in reproductive isolation, in the sister species Drosophila santomea and Drosophila yakuba. We designed a semi‐automatic measurement machine learning pipeline that can reliably identify curvatures and landmarks based on manually digitized contours of the ventral branches. With this method, we observed that temperature does not affect ventral branches in D. yakuba but that in D. santomea ventral branches tend to morph into a D. yakuba‐like shape at lower temperature. We found that male genitalia structures involved in reproductive isolation can be relatively variable within one species and can resemble the shape of closely related species’ genitalia through plasticity to temperature. Our results suggest that reproductive isolation mechanisms can be dependent on the environmental context., The closely related species Drosophila yakuba and Drosophila santomea have distinctive ventral branches, a male genital structure involved in reproductive isolation. We designed a semi‐automatic method to measure the shape of the ventral branches and found that temperature does not affect ventral branches in D. yakuba but that in D. santomea ventral branches tend to morph into a D. yakuba‐like shape at lower temperature. ​

The paper considers the implementation of algorithms for automatic search for signals scattered by meteor trails according to EKB ISTP SB RAS radar data. In general, the algorithm is similar to the algorithms adopted in specialized meteor systems. The algorithm is divided into two stages: detecting a meteor echo and determining its parameters. We show that on the day of the maximum Geminid shower, December 13, 2016, the scattered signals detected by the algorithm are foreshortening and correspond to scattering by irregularities extended in the direction of the meteor shower radiant. This confirms that the source of the signals detected by the algorithm is meteor trails. We implement an additional program for indirect trail height determination. It uses a decay time of echo and the NRLMSIS-00 atmosphere model to estimate the trail height. The dataset from 2017 to 2019 is used for further testing of the algorithm. We demonstrate a correlation in calculated Doppler velocity between the new algorithm and FitACF. We present a solution of the inverse problem of reconstructing the neutral wind velocity vector from the data obtained by the weighted least squares method. We compare calculated speeds and directions of horizontal neutral winds, obtained in the three-dimensional wind model, and the HWM-14 horizontal wind model. The algorithm allows real-time scattered signal processing and has been put into continuous operation at the EKB ISTP SB RAS radar.

Nowadays, during the diagnosis process, the doctor is able to obtain access to much information describing the patient’s condition using appropriate tools. However, there are always two sides to the coin. The doctor has certain limitations regarding the amount of data they can process at once. Information technology comes to the rescue, which with the help of computers is able to quickly and effectively separate important information from redundant information and support the doctor in making a diagnosis. In this work, a decision-making system was created to diagnose common lung pathologies in digital radiography images. Here, we consider four basic pulmonary diseases: pneumothorax, pneumonia, pulmonary consolidation, and lung lesions. Our objective is to develop a new automatic detection method of lung pathologies on chest X-ray radiographs using python programming language and its libraries. The approach uses solutions in the field of artificial intelligence, such as deep learning, convolutional neural network and segmentation to make a diagnosis that aims to help the radiologist at work. In the first sections, this work describes the fundamentals of the present form of diagnosis, a proposal to improve this process, the method of operation of the algorithms used, data acquisition, segmentation and processing methods. Then, the results of the operation of four different models and their implementation in a practical window program were presented. The best model, which detects pulmonary consolidation, achieves accuracy higher than 91%, which is a satisfactory result because they are not intended to replace radiologists but to improve their work. In the future, this type of program can be further developed by adding models that recognize other conditions.

The digital optical fiber repeater (DOFR) is an important infrastructure in the LTE networks, which solve the problem of poor regional signal quality. Various types of conventional measurement data from the LTE network cannot indicate whether a working DOFR is present in the cell. Currently, the detection of DOFRs relies solely on maintenance engineers for field detection. Manual detection methods are not timely or efficient, because of the large number and wide geographical distribution of DOFRs. Implementing automatic detection of DOFR can reduce the maintenance cost for mobile network operators. We treat the DOFR detection problem as a classification problem and employ a deep convolutional neural network (DCNN) to tackle it. The measurement report (MR) we used in this paper are tabular data, which is not an ideal input for DCNN. We propose a novel MR representation method that takes the overall MR data of a cell as a sample rather than a single record in the table, and represents the MR data as a pseudo-image matrix (PIM). The PIM will be used as the input for training DCNN, and the trained DCNN will be used to perform DOFR detection tasks. We conducted a series of experiments on real MR data, and the classification accuracy can achieve 93%. The proposed AI-based method can effectively detect the DOFR in a cell.

[ES] La endoscopia capsular (CE) es una ampliamente utilizada alternativa mínimamente invasiva a la endoscopia tradicional, que permite la visualización de todo el intestino delgado, mientras no es posible hacerlo fácilmente con los procedimientos más invasivos. Sin embargo, esos métodos tradicionales aún suelen ser la primera opción de tratamiento, ya que todavía existen desafíos importantes en el campo de la CE, incluyendo el tiempo necesario para el diagnóstico por vídeo después del procedimiento, el hecho de que la cápsula no se puede controlar activamente, la falta de consenso sobre una buena preparación del paciente y el coste alto. En esta tesis doctoral, nuestro objetivo es extraer más información de los procedimientos de endoscopía por cápsula para ayudar a aliviar estos problemas desde una perspectiva que parece estar subrepresentada en la investigación actual. Primero, como el objetivo principal en esta tesis, pretendemos desarrollar un método de evaluación de la limpieza en procedimientos de CE automático y objetivo para asistir la investigación médica en métodos de preparación de los pacientes. Específicamente, a pesar de que una preparación adecuada del paciente pueda ayudar a obtener una mejor visibilidad, los estudios sobre el método más efectivo son contradictorios debido a la ausencia de tal método. Por lo tanto, pretendemos proporcionar un método de ese tipo, capaz de presentar la limpieza en una escala intuitiva, con una novedosa arquitectura relativamente ligera de una red neuronal convolucional en su núcleo. Entrenamos este modelo en un conjunto de datos extensivo de más de 50,000 parches de imágenes, obtenidos de 35 procedimientos CE diferentes, y lo comparamos con métodos de clasificación del estado del arte. A partir de la clasificación, desarrollamos un método para automáticamente estimar las probabilidades a nivel de píxel y deducir los puntos en la escala de la evaluación de la limpieza a través de umbrales aprendidos. Después, validamos nuestro método en un entorno clínico en 30 videos de CE obtenidos nuevamente, comparando las puntuaciones resultantes con las asignadas de forma independiente por especialistas humanos. Obtuvimos la mayor precisión de clasificación para el método propuesto (95,23%), con tiempos de predicción promedios significativamente más bajos que para el segundo mejor método. En la validación, encontramos un acuerdo aceptable con dos especialistas humanos en comparación con el acuerdo interhumano, mostrando su validez como método de evaluación objetivo. Adicionalmente, otro objetivo de este trabajo es detectar automáticamente el túnel y ubicar el túnel en cada fotograma. Para este objetivo, entrenamos un modelo basado en R-CNN, concretamente el detector ligero YOLOv3, en un total de 1385 fotogramas, extraídos de procedimientos de CE de 10 pacientes diferentes. De tal manera, alcanzamos una precisión del 86,55% y una recuperación del 88,79% en nuestro conjunto de datos de test. Ampliando este objetivo, también pretendemos visualizar la motilidad intestinal de una manera análoga a una manometría intestinal tradicional, basada únicamente en la técnica mínimamente invasiva de CE. Para esto, alineamos los fotogramas con similar orientación y derivamos los parámetros adecuados para nuestro método de segmentación de las propiedades del rectángulo delimitador del túnel. Finalmente, calculamos el tamaño relativo del túnel para construir un equivalente de una manometría intestinal a partir de información visual. Desde que concluimos nuestro trabajo, nuestro método para la evaluación automática de la limpieza se ha utilizado en un estudio a gran escala aún en curso, en el que participamos activamente. Mientras gran parte de la investigación se centra en la detección automática de patologías, como tumores, pólipos y hemorragias, esperamos que nuestro trabajo pueda hacer una contribución significativa para extraer más información de la CE también en otras áreas frecuentemente subestimadas., [CA] L'endoscòpia capsular (CE) és una àmpliament utilitzada alternativa mínimament invasiva a l'endoscòpia tradicional, que permet la visualització de tot l'intestí prim, mentre no és possible fer-lo fàcilment amb els procediments més invasius. No obstant això, aqueixos mètodes tradicionals encara solen ser la primera opció de tractament, ja que encara existeixen desafiaments importants en el camp de la CE, incloent el temps necessari per al diagnòstic per vídeo després del procediment, el fet que la càpsula no es pot controlar activament, la falta de consens sobre una bona preparació del pacient i el cost alt. En aquesta tesi doctoral, el nostre objectiu és extraure més informació dels procediments de endoscopía per càpsula per a ajudar a alleujar aquests problemes des d'una perspectiva que sembla estar subrepresentada en la investigació actual. Primer, com l'objectiu principal en aquesta tesi, pretenem desenvolupar un mètode d'avaluació de la neteja en procediments de CE automàtic i objectiu per a assistir la investigació mèdica en mètodes de preparació dels pacients. Específicament, a pesar que una preparació adequada del pacient puga ajudar a obtindre una millor visibilitat, els estudis sobre el mètode més efectiu són contradictoris a causa de l'absència de tal mètode. Per tant, pretenem proporcionar un mètode d'aqueix tipus, capaç de presentar la neteja en una escala intuïtiva, amb una nova arquitectura relativament lleugera d'una xarxa neuronal convolucional en el seu nucli. Entrenem aquest model en un conjunt de dades extensiu de més de 50,000 pegats d'imatges, obtinguts de 35 procediments CE diferents, i el comparem amb mètodes de classificació de l'estat de l'art. A partir de la classificació, desenvolupem un mètode per a automàticament estimar les probabilitats a nivell de píxel i deduir els punts en l'escala de l'avaluació de la neteja a través de llindars apresos. Després, validem el nostre mètode en un entorn clínic en 30 vídeos de CE obtinguts novament, comparant les puntuacions resultants amb les assignades de manera independent per especialistes humans. Vam obtindre la major precisió de classificació per al mètode proposat (95,23%), amb temps de predicció mitjanes significativament més baixos que per al segon millor mètode. En la validació, trobem un acord acceptable amb dos especialistes humans en comparació amb l'acord interhumà, mostrant la seua validesa com a mètode d'avaluació objectiu. Addicionalment, un altre objectiu d'aquest treball és detectar automàticament el túnel i situar el túnel en cada fotograma. Per a aquest objectiu, entrenem un model basat en R-CNN, concretament el detector lleuger YOLOv3, en un total de 1385 fotogrames, extrets de procediments de CE de 10 pacients diferents. De tal manera, aconseguim una precisió del 86,55% i una recuperació del 88,79% en el nostre conjunt de dades de test. Ampliant aquest objectiu, també pretenem visualitzar la motilitat intestinal d'una manera anàloga a una manometría intestinal tradicional, basada únicament en la tècnica mínimament invasiva de CE. Per a això, alineem els fotogrames amb similar orientació i derivem els paràmetres adequats per al nostre mètode de segmentació de les propietats del rectangle delimitador del túnel. Finalment, calculem la grandària relativa del túnel per a construir un equivalent d'una manometría intestinal a partir d'informació visual. Des que concloem el nostre treball, el nostre mètode per a l'avaluació automàtica de la neteja s'ha utilitzat en un estudi a gran escala encara en curs, en el qual participem activament. Mentre gran part de la investigació se centra en la detecció automàtica de patologies, com a tumors, pòlips i hemorràgies, esperem que el nostre treball puga fer una contribució significativa per a extraure més informació de la CE també en altres àrees sovint subestimades., [EN] Capsule endoscopy (CE) is a widely used, minimally invasive alternative to traditional endoscopy that allows visualisation of the entire small intestine, whereas more invasive procedures cannot easily do this. However, those traditional methods are still commonly the first choice of treatment for gastroenterologists as there are still important challenges surrounding the field of CE. Among others, these include the time consuming video diagnosis following the procedure, the fact that the capsule cannot be actively controlled, lack of consensus on good patient preparation and the high cost. In this doctoral thesis, we aim to extract more information from capsule endoscopy procedures to aid in alleviating these issues from a perspective that appears to be under-represented in current research. First, and as the main objective in this thesis, we aim to develop an objective, automatic cleanliness evaluation method in CE procedures to aid medical research in patient preparation methods. Namely, even though adequate patient preparation can help to obtain a cleaner intestine and thus better visibility in the resulting videos, studies on the most effective preparation method are conflicting due to the absence of such a method. Therefore, we aim to provide such a method, capable of presenting results on an intuitive scale, with a relatively light-weight novel convolutional neural network architecture at its core. We trained this model on an extensive data set of over 50,000 image patches, collected from 35 different CE procedures, and compared it with state-of-the-art classification methods. From the patch classification results, we developed a method to automatically estimate pixel-level probabilities and deduce cleanliness evaluation scores through automatically learnt thresholds. We then validated our method in a clinical setting on 30 newly collected CE videos, comparing the resulting scores to those independently assigned by human specialists. We obtained the highest classification accuracy for the proposed method (95.23%), with significantly lower average prediction times than for the second-best method. In the validation of our method, we found acceptable agreement with two human specialists compared to interhuman agreement, showing its validity as an objective evaluation method. Additionally, we aim to automatically detect and localise the tunnel in each frame, in order to help determine the capsule orientation at any given time. For this purpose, we trained an R-CNN based model, namely the light-weight YOLOv3 detector, on a total of 1385 frames, extracted from CE procedures of 10 different patients, achieving a precision of 86.55% combined with a recall of 88.79% on our test set. Extending on this, we additionally aim to visualise intestinal motility in a manner analogous to a traditional intestinal manometry, solely based on the minimally invasive technique of CE, through aligning the frames with similar orientation and using the bounding box parameters to derive adequate parameters for our tunnel segmentation method. Finally, we calculate the relative tunnel size to construct an equivalent of an intestinal manometry from visual information. Since we concluded our work, our method for automatic cleanliness evaluation has been used in a still on-going, large-scale study, with in which we actively participate. While much research focuses on automatic detection of pathologies, such as tumors, polyps and bleedings, we hope our work can make a significant contribution to extract more information from CE also in other areas that are often overlooked.

Acoustic monitoring allows the evaluation of spatio-temporal changes in animal populations. However, analyzing large volumes of information is challenging. We evaluate the performance of a detection technique (autodetec function of the R warbleR package) to identify vocalizations of Megascops centralis, using 6877 one-minute recordings from 21 sites in the vicinity of the Jaguas dam, Andes of Antioquia Colombia., All vocalizations were manually annotated and two sites (597 recordings) with the highest number of records (49 and 34) were selected to evaluate the algorithm. The function was implemented with audios at two sampling rates (44 100 Hz and 22 050 Hz) and three amplitude thresholds (5, 10, and 20). We assessed the performance of this function in terms of its sensitivity and specificity, and we estimate the probability of detection of a signal according to its quality. Sensitivity and specificity showed great variation (0-0.48 and 0.5-0.98 respectively) and the probability of detection of a signal increased with its quality (poor: 0.12, medium: 0.27 andhigh: 0.64). Acoustic monitoring has an enormous potential, and its success depends, in part, on the availability of automatic recognition tools, that are open access and can be easily implemented. This development can be achieved by strengthening acoustic collections.

Efficient leadership plays an important role in organizations, with the military being one of the more obvious examples of this statement. In this context, it is not surprising that ensuring a culture of excellence is at the heart of Navy leadership. However, it is not easy to maintain or increase the quality of leadership among staff, as such efforts require constant training and practice. To address this need for continuous monitoring and improvement in human leadership expressed in everyday communication, we demonstrate the feasibility of automatically detecting and classifying military leadership messages. We achieve this goal by 1) curating a data set of short text messages that are written in the military-specific language, have some characteristics of spoken language, and are human-annotated with labels referring to selected leadership roles and 2) demonstrating the performance of selected automation methods that allow classes to be predicted for each analyzed message. This study shows that recent deep learning methods provide reasonable performance, even when limited data is provided. Future efforts should focus on creating an automated self-assessment tool that would enable continuous monitoring and training of leadership skills required in the Navy domain.

Invasive ductal carcinoma (IDC) is the most common type of breast cancer which is the leading cause of cancer-related deaths in middle-aged women. Pathological analysis of biopsy is the gold standard for diagnosis of breast cancer, and early detection, diagnosis, and treatment can significantly increase the survival rate. This paper proposes a method for the automatic detection of IDC based on the fusion of multi-scale residual convolutional neural network (MSRCNN) and SVM. First, the patches from whole slide images (WSI) were preprocessed by data enhancement and normalization and then input into the MSRCNN for features extraction. Second, the extracted features were input to the SVM and are classified into two categories: healthy and diseased patches. Finally, it is restored to the WSI according to the coordinate information of the patches, therefore the IDC and healthy tissue regions were built. Experimental results show that after 5-fold cross-validation, our method obtained an average accuracy of 87.45±0.81%, an average balance accuracy of 85.7±0.95%, and an average F1 score of 79.89±1.11%. Consequently, it has important practical value and scientific research significance.

Automatic weeding is a hot research field of smart agriculture, which has many benefits such as achieving precise weed control, saving human cost, and avoiding damage on crops, etc. Recently, many researchers have focused on the research using the deep learning method, such as the convolutional neural network (CNN) and recurrent neural network (RNN) and have achieved decent outcomes related to the automatic weed detection. However, there are still generally problems of the projects such as weak robustness and excessive reliance on a large number of samples. To solve these problems, a recognition algorithm for automatic identification and weed removal was designed, and a soybean field weed detection and localization method based on the fusion of multiple image processing methods was proposed in this study. The images and video stream were obtained through the camera mounted on a mobile robot platform. Firstly, the soil background inside the image was segmented from the foreground (including the weeds and crops) by setting the threshold for a specific color space (hue). Then, three different methods including the area threshold method, template matching and saturation threshold method were used to classify the crops and weeds. Finally, based on a proposed innovative voting method, the three recognition methods were comprehensively weighed and fused to achieve more accurate recognition and localization results of the crops and weeds inside the image. Experimental validations were carried out using the samples obtained through the moving platform, and the experimental results showed that the average accuracy of the proposed weed detection algorithm was as high as 98.21%, while the recognition error was only 1.79%. Meanwhile, compared with each single method as the scale threshold, template matching and saturation threshold, the fused method based on the weighted voting has been able to raise the average accuracy by 5.71%. Even though the samples used in the validations were limited in covering different scenarios, the high recognition accuracy has proved the practicability of the proposed method. In addition, the robustness test that images with raindrop and shadow interference in the complex and unstructured agricultural scene was carried out, and satisfied results showed that above 90% of the plant were successfully detected, which verified the fine adaptability and robustness of the proposed method.

In order to weaken the limitation between the adjacent nodes of shooting action teaching image and fully expose the necessary image features, an automatic detection method of shooting action teaching image features based on artificial intelligence is proposed. Starting from the decomposition of shooting teaching action, the practical attribute of shooting teaching image is defined, and then the standardized image detection environment based on artificial intelligence is built according to the action essentials of consistent shooting. On this basis, through the way of extracting the characteristics of teaching image, calculating the characteristic scale value of shooting action, combining with the feature description principle of automatic detection, the design of automatic detection method of shooting action teaching image characteristics is completed. The experimental results show that, compared with the traditional feature method, the UTI traction coefficient between the adjacent nodes of the image increases to 0.33, while the matching time required for detection decreases to 17.5 mm, so as to fully expose the necessary features of the shooting action teaching image.

In the currentworld the best modefor transport is two wheeler transports. But huge risk is involved because of having protection at lower level. This paper proposed an approach for automatically detecting whether the rider wear helmet or not with any manual intervention. If the rider of the bike identified not wearing helmet, then corresponding two wheeler number plate was read and noted. Data records of every offender who were not wore helmet will be automatically saved into the database A database will be generated with records to identify every offender accurately. This paper aims to explain automatic detection of motor cyclists without helmet and sending the messages to detected persons. Our proposed system is to extract the vehicle number from the RTO website using number plate of a vehicle by using the captured images. After getting the information use mobile number as a input to our model using this mobile number sending the fine details to the detected person. Our system maintains a database for number of detections of that particular vehicle numbers and also sending the detecting location details.

Despite the recent recognition of jellyfish as an important component of marine ecosystems and existing concerns on their potential population increase, they are rarely monitored at the appropriate spatial and temporal scales. Traditional jellyfish monitoring techniques are costly and generally restrict the spatial–temporal resolution limiting the quantity and quality of monitoring data. We introduce Jellytoring 2.0, an automatic recognition tool for jellyfish species based on convolutional neural networks (CNN). We trained Jellytoring 2.0 to identify 15 jellyfish species with a global distribution. Our aim is to offer Jellytoring 2.0 as an open-access tool to serve as the backbone for a system that promotes the creation of large-scale and long-term jellyfish monitoring data. Results reveal that Jellytoring 2.0 performed well in the identification of the 15 species with average precision values ranging between 90% and 99% for most of the species. Four of the species presented slightly lower values (75%–80%). Our system was trained on a relatively small dataset, implying that additional integration of image data will further improve the performance of the CNN. We show how the application of CNNs to image data can deliver a tool that will enable the cost-effective collection of jellyfish data on larger spatial and temporal scales. For Jellytoring 2.0 to become a truly global automatic identification system, we ask scientists and nonscientists to actively contribute with jellyfish image data to extend the number of species it can identify., During the conception, implementation and writing of the study A.R.F were supported by a Marie-Sklodowska-Curie Individual Fellowship (JellyPacts project number 655475) and a Juan de la Cierva-Incorporación (IJC2019-040836-I/AEI/10.13039/501100011033) from the Spanish Government. M.M.A. was supported by Ministry of Economy and Competitiveness (AEI, FEDER, UE), under contract DPI2017-86372-C3-3-R. C.L.J. was supported by the EU funded Erasmus+ program. Y.G.C. was supported by Ministry of Economy and Competitiveness (AEI, FEDER, UE) under contract DPI2017-86372-C3-3-R; Ministry of Science and Innovation (AEI, FEDER, UE) under contracts PID2020-115332RB-C33 and PLEC2021-007525 and by the Comunitat Autonoma de les Illes Balears through the Direcció General de Política Universitaria i Recerca with funds from the Tourist Stay Tax Law (PRD2018/34). H.H. was partially supported through a Ramón y Cajal Fellowship financed by the Ministerio de Economía y Competitividad de España and the Conselleria d'Educació, Cultura i Universitats Comunidad Autónoma de las Islas Baleares (RyC 2013 14729).

Purpose: We assessed three commercial automated spike detection software packages (Persyst, Encevis and BESA) to see which had the best performance. Methods: Thirty prolonged EEG records from people aged at least 16 years were collected and 30-minute representative epochs were selected. Interictal epileptiform discharges (IEDs) were marked by three human experts and by all three software packages. For each 30-minutes selection and for each 10-second epoch we measured whether or not IEDs had occurred. We defined the gold standard as the combined detections of the experts. Kappa scores, sensitivity and specificity were estimated for each software package. Results: Sensitivity for Persyst in the default setting was 95% for 30-minute selections and 82% for 10-second epochs. Sensitivity for Encevis was 86% (30-minute selections) and 61% (10-second epochs). The specificity for both packages was 88% for 30-minute selections and 96%-99% for the 10-second epochs. Interrater agreement between Persyst and Encevis and the experts was similar than between experts (0.67-0.83 versus 0.63-0.67). Sensitivity for BESA was 40% and specificity 100%. Interrater agreement (0.25) was low. Conclusions: IED detection by the Persyst automated software is better than the Encevis and BESA packages, and similar to human review, when reviewing 30-minute selections and 10-second epochs. This findings may help prospective users choose a software package.

(IF 10.75; Q1); International audience; Hyperspectral imaging (HSI) is a non-destructive, high-resolution imaging technique that is currently under significant development for analyzing geological areas with remote devices or natural samples in a laboratory. In both cases, the hyperspectral image provides several sedimentary structures that must be separated to temporally and spatially describe the sample. Sediment sequences are composed of successive deposits (strata, homogenite, flood) that are visible depending on sample properties. The classical methods to identify them are time-consuming, have a low spatial resolution (millimeters) and are generally based on naked-eye counting. In this study, we compare several supervised classification algorithms to discriminate sedimentological structures in lake sediments. Instantaneous events in lake sediments are generally linked to extreme geodynamical events (e.g., floods, earthquakes), so their identification and counting are essential to understand long-term fluctuations and improve hazard assessments. Identification and counting are done by reconstructing a chronicle of event layer occurrence, including estimation of deposit thicknesses. Here, we applied two hyperspectral imaging sensors (Visible Near-Infrared, VNIR, 60 μm, 400–1000 nm; Short Wave Infrared, SWIR, 200 μm, 1000–2500 nm) on three sediment cores from different lake systems. We highlight that the SWIR sensor is the optimal one for creating robust classification models with discriminant analyses (prediction accuracies of 0.87–0.98). Indeed, the VNIR sensor is impacted by the surface reliefs and structures that are not in the learning set, which causes mis-classification. These observations are also valid for the combined sensor (VNIR-SWIR) and the RGB images. Several spatial and spectral pre-processing were also compared and enabled one to highlight discriminant information specific to a sample and a sensor. These works show that the combined use of hyperspectral imaging and machine learning improves the characterization of sedimentary structures compared to conventional methods.

Objective: To evaluate the performance of a deep learning-based computer-aided detection (DL-CAD) system in a Chinese low-dose CT (LDCT) lung cancer screening program. Materials and methods: One-hundred-and-eighty individuals with a lung nodule on their baseline LDCT lung cancer screening scan were randomly mixed with screenees without nodules in a 1:1 ratio (total: 360 individuals). All scans were assessed by double reading and subsequently processed by an academic DL-CAD system. The findings of double reading and the DL-CAD system were then evaluated by two senior radiologists to derive the reference standard. The detection performance was evaluated by the Free Response Operating Characteristic curve, sensitivity and false-positive (FP) rate. The senior radiologists categorized nodules according to nodule diameter, type (solid, part-solid, non-solid) and Lung-RADS. Results: The reference standard consisted of 262 nodules ≥ 4 mm in 196 individuals; 359 findings were considered false positives. The DL-CAD system achieved a sensitivity of 90.1% with 1.0 FP/scan for detection of lung nodules regardless of size or type, whereas double reading had a sensitivity of 76.0% with 0.04 FP/scan (P = 0.001). The sensitivity for detection of nodules ≥ 4 - ≤ 6 mm was significantly higher with DL-CAD than with double reading (86.3% vs. 58.9% respectively; P = 0.001). Sixty-three nodules were only identified by the DL-CAD system, and 27 nodules only found by double reading. The DL-CAD system reached similar performance compared to double reading in Lung-RADS 3 (94.3% vs. 90.0%, P = 0.549) and Lung-RADS 4 nodules (100.0% vs. 97.0%, P = 1.000), but showed a higher sensitivity in Lung-RADS 2 (86.2% vs. 65.4%, P < 0.001). Conclusions: The DL-CAD system can accurately detect pulmonary nodules on LDCT, with an acceptable false-positive rate of 1 nodule per scan and has higher detection performance than double reading. This DL-CAD system may assist radiologists in nodule detection in LDCT lung cancer screening.

This paper presents a novel method for partial discharge detection. Generally, the presence of partial discharge in the power grid is unfavorable as it reduces the electricity transmission efficiency and also adversely affects the durability of the line. A partial discharge manifests itself as a transient voltage oscillation with a very small amplitude and a short duration. The resolution and sampling frequency of the measurement hardware must therefore be sufficient to capture these short transient signals. Furthermore, some algorithm must be able to automatically detect these transients and distinguish a true partial discharge from other components that resemble partial discharges with certain properties, but in reality, are out of interest. The first part of the article describes the properties of partial discharge signals in the frequency domain, which are further used for the discharge detection. Based on the time-frequency properties of partial discharge signals, a recursive filtering method was designed. The second part of the article shows the suitability of the proposed solution on examples of discharge detection in both test and field conditions. As will be shown, the proposed method with a high degree of reliability detects both test discharges generated in laboratory conditions and discharges that occurred during operation of the power grid.

Human values such as integrity, privacy, curiosity, security, and honesty are guiding principles for what people consider important in life. Such human values may be violated by mobile software applications (apps), and the negative effects of such human value violations can be seen in various ways in society. In this work, we focus on the human value of honesty. We present a model to support the automatic identification of violations of the value of honesty from app reviews from an end-user perspective. Beyond the automatic detection of honesty violations by apps, we also aim to better understand different categories of honesty violations expressed by users in their app reviews. The result of our manual analysis of our honesty violations dataset shows that honesty violations can be characterised into ten categories: unfair cancellation and refund policies; false advertisements; delusive subscriptions; cheating systems; inaccurate information; unfair fees; no service; deletion of reviews; impersonation; and fraudulent-looking apps. Based on these results, we argue for a conscious effort in developing more honest software artefacts including mobile apps, and the promotion of honesty as a key value in software development practices. Furthermore, we discuss the role of app distribution platforms as enforcers of ethical systems supporting human values, and highlight some proposed next steps for human values in software engineering (SE) research.

Industrial X-ray computed tomography (CT) has proven its value as a non-destructive method for inspecting light metal castings. The CT volume generated enables the internal and external geometry of the specimen to be measured, casting defects to be localized and their statistical properties to be investigated. On the other hand, CT volumes are very prone to artifacts that can be mistaken for defects by conventional segmentation algorithms. Based on CT data of aluminium alloy castings provided by industrial partners, we have developed an automatic approach to analyze discontinuities inside CT volumes based on a three-step pipeline: (1) 2D segmentation of CT slices with automatic deep segmentation to detect suspicious greyscale discontinuities; (2) classification of these discontinuities into true alarms (defects) or false alarms (artifacts and noise), using a trained convolutional neural network classifier; (3) localization of the validated defects in 3D to investigate their statistical properties such as sphericity, elongation and compactness. Based on this, the validated 3D defects are then classified into porosities or shrinkage cavities using an SVM classifier and a siamese neural network. The choice of each model and the training results are presented and discussed, as well as the efficiency of the approach as an automatic defect detection algorithm for industrial CT volumes., La tomographie industrielle à rayons X (CT) a prouvé sa valeur en tant que méthode non destructive pour l'inspection des pièces de fonderie en métal léger. Le volume CT généré permet de mesurer la géométrie interne et externe de la pièce, de localiser les défauts de fonderie et d'étudier leurs propriétés statistiques. D'autre part, les volumes CT sont très sujets aux artefacts qui peuvent être confondus avec des défauts par les algorithmes de segmentation conventionnels. Sur la base des données CT des pièces de fonderie en alliage d'aluminium fournies par des partenaires industriels, une approche automatique a été développée pour analyser les discontinuités à l'intérieur des volumes CT sur la base d'un pipeline en trois étapes : (1) segmentation 2D des tranches CT avec segmentation profonde automatique pour détecter les discontinuités suspectes en niveaux de gris ; (2) classification de ces discontinuités en véritables alarmes (défauts) ou fausses alarmes (artefacts et bruit), à l'aide d'un classificateur de réseau neuronal convolutif entrainé; (3) localisation des défauts validés en 3D pour étudier leurs propriétés géométriques telles que la sphéricité, l'allongement et la compacité. Les défauts 3D validés sont ensuite classés en porosités ou retassures à l'aide d'un classificateur SVM et d'un réseau de neurones siamois. Le choix de chaque modèle et les résultats d'entrainement sont présentés et discutés, ainsi que l'efficacité de l'approche en tant qu'algorithme de détection automatique de défauts pour les volumes CT industriels.

Laser beam welding is used in many areas of industry and research. There are many strategies and approaches to further improve the weld seam properties in laser beam welding. Metallography is often needed to evaluate welded seams. Typically, the images are examined and evaluated by experts. The evaluation process qualitatively provides the properties of the welds. Particularly in times when artificial intelligence is being used more and more in processes, the quantization of properties that could previously only be determined qualitatively is gaining importance. In this contribution, we propose to use deep learning to perform semantic segmentation of micrographs of complex weld areas to achieve the automatic detection and quantization of weld seam properties. A semantic segmentation dataset is created containing 282 labeled images. The training process is performed with DeepLabv3+. The trained model achieves a value of around 95% for weld contour detection and 76.88% of mean intersection over union (mIoU). © 2022 by the authors. Licensee MDPI, Basel, Switzerland.

Ulcerative colitis (or "UC") is a chronic inflammatory bowel disease. It is in curable and can lead to the removal of part of the large intestine and/or col orectal cancer. It affects the intestinal wall with lesions such as bleeding, ulcers and blurring at vascular pattern visibility. By inserting a camera into the colon (colonoscopy), the gastroenterologist can visualize the intestinal wall and thus assess the severity of the disease. Current severity scores are based on the gas troenterologist’s interpretation and only consider the most severe lesions. They ignore the spatial extent and local distribution of the lesions along with the colon segments. The goal of this thesis is to provide an objective evaluation of the ulcerative colitis severity. Our works are based on the analysis of endoscopic videos and mathematical modelling techniques. First of all, we propose an algorithm to automatically detect bleeding and ulcers. We use endoscopic videos containing rectangular annotations and subsequently, many mislabelled pixels are present. We, therefore, compute the performance of the detectors at the annotations level instead of the pixels level. We also pro pose a models sampling method that speed-up the algorithm. We show good agreement with doctors’ annotations. Then, a diagram of visualization of the lesions throughout the colon was pro posed. The obtained results show the lesions’ spatial distribution in the whole colon. In addition, they allow a finer distinction between patients with the same severity score and thus better evaluation of the disease severity. The next work is the modelling of the severity taking into account the spatial distribution of the lesions on the segments of the colon. The results show a correlation with current severity scores. Next, we study the bias in the distribution of lesions close to the rectum. We also bring proof to some medical statements related to the disease severity. Finally, a reaction-diffusion equation was introduced to model the distribution of lesions in the colon. The objective is to study the spatial dynamics of the le sions as they propagate in the colon. In application to UC patients, we compute the speed of propagation of lesions in the colon as well as the time for which the 4 whole colon will be inflamed. The results suggest dissimilar behaviour between patients with the same severity score. All these works are carried out in close collaboration with biologists from Labex- Inflamex as well as gastroenterologists from Bichat and Beaujon hospital, La rectocolite hémorragique ("RCH") est une maladie inflammatoire chronique de l’intestin. Elle est incurable et peut conduire à l’ablation d’une partie du gros intestin et/ou à un cancer colorectal. Elle affecte la paroi intestinale avec des lésions telles que des saignements, des ulcères et un brouillage au niveau de la visibilité de la trame vasculaire. En insérant une caméra dans le côlon (coloscopie), le gastro-entérologue peut visualiser la paroi intestinale et ainsi évaluer la gravité de la maladie. Les scores de gravité actuels sont basés sur l’interprétation du gastro-entérologue et ne prennent en compte que les lésions les plus graves. Ils ignorent l’étendue spatiale et la distribution locale des lésions sur les segments du côlon. L’objectif de cette thèse est de fournir une évaluation objective de la sévérité de la rectocolite hémorragique. Nos travaux sont basés sur l’analyse de vidéos endoscopiques et sur des techniques de modélisation mathématique. Tout d’abord, nous proposons un algorithme pour détecter automatiquement les saignements et les ulcères. Nous utilisons des vidéos endoscopiques contenant des annotations rectangulaires et par conséquent, de nombreux pixels mal étiquetés sont présents. Nous calculons donc la performance des détecteurs au niveau des annotations plutôt qu’au niveau des pixels. Nous proposons également une méthode d’échantillonnage de modèles qui accélère l’algorithme. Nous montrons un bon accord avec les annotations des médecins. Ensuite, un schéma de visualisation des lésions à travers le côlon a été proposé. Les résultats obtenus montrent la distribution spatiale des lésions dans l’ensemble du côlon. De plus, ils permettent une distinction plus fine entre les patients ayant le même score de gravité et donc une meilleure évaluation de la sévérité de la maladie. Le travail suivant est la modélisation de la sévérité en tenant compte de la distribution spatiale des lésions sur les segments du côlon. Les résultats montrent une corrélation avec les scores de sévérité actuels. Ensuite, nous étudions le biais de la distribution des lésions à proximité du rectum. Nous apportons également des preuves à certaines hypothèses médicales liées à la gravité de la maladie. Enfin, une équation de réaction-diffusion est introduite pour modéliser la distribution des lésions dans le colon. L’objectif est d’étudier la dynamique spatiale des lésions lorsqu’elles se propagent dans le côlon. En application aux patients atteints de la RCH, nous calculons la vitesse de propagation des lésions dans le côlon ainsi que le temps pendant lequel le côlon entier sera enflammé. Les résultats suggèrent un comportement différent entre les patients ayant le même score de sévérité. Tous ces travaux sont menés en collaboration étroite avec les biologistes du Labex-Inflamex ainsi que les gastro-entérologues des hôpitaux Bichat et Beaujon.

International audience; Historical charcoal production is one of the significant factors affecting today’s forest dynamics. A key challenge is to develop tools to investigate historical charcoal production over large areas, allowing a more comprehensive understanding of past impacts and history of charcoal production over a given landscape. In this study, high-resolution remote-sensing airborne LiDAR images over a large woodland area were used to compare manual on-screen versus algorithm-based automatic methods to inventory charcoal kilns with inputs of field-validated data. The results revealed that (1) the on-screen detection method provided less false-positives, (2) the automatic method detects a higher number of kilns and (3) kiln distribution seemed to be connected mostly to land ownership rather than to environmental variables. This study validates a new method of charcoal kilns’ inventory and spatial analysis that can be applied to other areas to better understand the effect of past biomass harvesting for charcoal production on forest dynamics.

In this article a fully automated machine-vision technique for the detection and segmentation of mesenteric cysts in computed tomography (CT) images of the abdominal space is presented. The proposed technique involves clustering, filtering, morphological operations and evaluation processes to detect and segment mesenteric cysts in the abdomen regardless of their texture variation and location with respect to other surrounding abdominal organs. The technique is comprised of various processing phases, which include K-means clustering, iterative Gaussian filtering, and an evaluation of the segmented regions using area-normalized histograms and Euclidean distances. The technique was tested using 65 different abdominal CT scan images. The results showed that the technique was able to detect and segment mesenteric cysts and achieved 99.31%, 98.44%, 99.84%, 98.86% and 99.63% for precision, recall, specificity, dice score coefficient and accuracy respectively as quantitative performance measures which indicate very high segmentation accuracy.

Myocardial infarction (MI) occurs when an artery supplying blood to the heart is abruptly occluded. The “gold standard” method for imaging MI is cardiovascular magnetic resonance imaging (MRI) with intravenously administered gadolinium-based contrast (with damaged areas apparent as late gadolinium enhancement [LGE]). However, no “gold standard” fully automated method for the quantification of MI exists. In this work, we propose an end-to-end fully automatic system (MyI-Net) for the detection and quantification of MI in MRI images. It has the potential to reduce uncertainty due to technical variability across labs and the inherent problems of data and labels. Our system consists of four processing stages designed to maintain the flow of information across scales. First, features from raw MRI images are generated using feature extractors built on ResNet and MoblieNet architectures. This is followed by atrous spatial pyramid pooling (ASPP) to produce spatial information at different scales to preserve more image context. High-level features from ASPP and initial low-level features are concatenated at the third stage and then passed to the fourth stage where spatial information is recovered via up-sampling to produce final image segmentation output into: (i) background, (ii) heart muscle, (iii) blood and (iv) LGE areas. Our experiments show that the model named MI-ResNet50-AC provides the best global accuracy (97.38%), mean accuracy (86.01%), weighted intersection over union (IoU) of 96.47%, and bfscore of 64.46% for the global segmentation. However, in detecting only LGE tissue, a smaller model, MI-ResNet18-AC, exhibited higher accuracy (74.41%) than MI-ResNet50-AC (64.29%). New models were compared with state-of-the-art models and manual quantification. Our models demonstrated favorable performance in global segmentation and LGE detection relative to the state-of-the-art, including a four-fold better performance in matching LGE pixels to contours produced by clinicians.

Despite significant advancements in object detection technology, most existing detection networks fail to investigate global aspects while extracting features from the inputs and cannot automatically adjust based on the characteristics of the inputs. The present study addresses this problem by proposing a detection network consisting of three stages: preattention, attention, and prediction. In the preattention stage, the network framework is automatically selected based on the features of the images’ objects. In the attention stage, the transformer structure is introduced. Taking into account the global features of the target, this study combines a self-attention module in the transformer model and convolution operation to integrate image features from global to local and for detection, thus improving the ship target accuracy. This model uses mathematical methods to obtain results of predictive testing in the prediction stage. The above improvements are based on the You Only Look Once version 4 (YOLOv4) framework, named “Auto-T-YOLO”. The model achieves the highest accuracy of 96.3% on the SAR Ship Detection dataset (SSDD) compared to the other state-of-the-art (SOTA) model. It achieves 98.33% and 91.78% accuracy in the offshore and inshore scenes, respectively. The experimental results verify the practicality, validity, and robustness of the proposed model.

Among the non-destructive testing (NDT) techniques, infrared thermography (IRT) is an attractive and highly reliable technology that can measure the thermal response of a wide area in real-time. In this study, thinning defects in S275 specimens were detected using lock-in thermography (LIT). After acquiring phase and amplitude images using four-point signal processing, the optimal excitation frequency was calculated. After segmentation was performed on each defect area, binarization was performed using the Otsu algorithm. For automated detection, the boundary tracking algorithm was used. The number of pixels was calculated and the detectability using RMSE was evaluated. Clarification of defective objects using image segmentation detectability evaluation technique using RMSE was presented.

International audience; Benthic diatoms are unicellular microalgae that are routinely used as bioindicators for monitoring the ecological status of freshwater. Their identification using light microscopy is a time-consuming and laborintensive task that could be automated using deep-learning. However, training such networks relies on the availability of labeled datasets, which are difficult to obtain for these organisms. Herein, we propose a method to generate synthetic microscopy images for training. We gathered individual objects, i.e. 9230 diatoms from publicly available taxonomic guides and 600 items of debris from available real images. We collated a comprehensive dataset of synthetic microscopy images including both diatoms and debris using seamless blending and a combination of parameters such as image scaling, rotation, overlap and diatom-debris ratio. We then performed sensitivity analysis of the impact of the synthetic data parameters for training state-of-the art networks for horizontal and rotated bounding box detection (YOLOv5). We first trained the networks using the synthetic dataset and fine-tuned it to several real image datasets. Using this approach, the performance of the detection network was improved by up to 25% for precision and 23% for recall at an Intersectionover-Union(IoU) threshold of 0.5. This method will be extended in the future for training segmentation and classification networks.

Diabetic retinopathy (DR) is a human eye disease that affects people who are suffering from diabetes. It causes damage to their eyes, including vision loss. It is treatable; however, it takes a long time to diagnose and may require many eye exams. Early detection of DR may prevent or delay the vision loss. Therefore, a robust, automatic and computer-based diagnosis of DR is essential. Currently, deep neural networks are being utilized in numerous medical areas to diagnose various diseases. Consequently, deep transfer learning is utilized in this article. We employ five convolutional-neural-network-based designs (AlexNet, GoogleNet, Inception V4, Inception ResNet V2 and ResNeXt-50). A collection of DR pictures is created. Subsequently, the created collections are labeled with an appropriate treatment approach. This automates the diagnosis and assists patients through subsequent therapies. Furthermore, in order to identify the severity of DR retina pictures, we use our own dataset to train deep convolutional neural networks (CNNs). Experimental results reveal that the pre-trained model Se-ResNeXt-50 obtains the best classification accuracy of 97.53% for our dataset out of all pre-trained models. Moreover, we perform five different experiments on each CNN architecture. As a result, a minimum accuracy of 84.01% is achieved for a five-degree classification.

In this paper, an automatic algorithm for the detection of subsidence areas in SAR interferograms is proposed. It is based on the analysis of spatial distribution of the interferogram phase, and its coherence and entropy. The developed method was tested for differential interferograms generated on the basis of Sentinel-1 SAR images covering mining areas in South Poland. The obtained results were compared with those achieved using a method based on circular Gabor filters. Performed analysis revealed that the detection rate for the proposed method varied from 34% to 83%. It is an improved method based on Gabor filters that achieved a detection rate from 30% to 53%.

Purpose: We assessed whether automated detection software, combined with live observation, enabled reliable seizure detection using three commercial software packages: Persyst, Encevis and BESA. Methods: Two hundred and eighty-six prolonged EEG records of individuals aged 16-86 years, collected between August 2019 and January 2020, were retrospectively processed using all three packages. The reference standard included all seizures mentioned in the clinical report supplemented with true detections made by the software and not previously detected by clinical physiologists. Sensitivity was measured for offline review by clinical physiologists and software seizure detection, both in combination with live monitoring in an EMU setting, for all three software packages at record and seizure level. Results: The database contained 249 seizures in 64 records. The sensitivity of seizure detection was 98% for Encevis and Persyst, and 95% for BESA, when a positive results was defined as detection at least one of the seizures occurring within an individual record. When positivity was defined as recognition of all seizures, sensitivity was 93% for Persyst, 88% for Encevis and 84% for BESA. Clinical physiologists' review had a sensitivity of 100% at record level and 98% at seizure level. The median false positive rate per record was 1.7 for Persyst, 2.4 for BESA and 5.5 for Encevis per 24 h. Conclusion: Automated seizure detection software does not perform as well as technicians do. However, it can be used in an EMU setting when the user is aware of its weaknesses. This assessment gives future users helpful insight into these strengths and weaknesses. The Persyst software performs best.

[EN] This Master’s Degree final work consists of the development of an Android application to detect a vertical jump through a video and show its data and measurements. The system is intended to be used by people who want to record or improve their vertical jumping ability or otherwise want to evaluate it from a health point of view. This purpose arises from the idea of improving current applications to analyze vertical jump. The idea for improvement comes from the fact that none of the existing applications, even the market leaders, present the possibility of detecting the jump automatically. Thanks to this, the idea arose of creating an application that allows the user to obtain the measurements of their vertical jump in a more automatic way, without having to view a complete video to inform the application when the jump is made. The first idea of this work is to carry out a research work on the tools and methods to detect a vertical jump in a video. This investigation was carried out in this order: a phase of analysis of the different tools for their detection, another phase of analysis of the ways to detect the jump with the chosen tool, a phase of tests to verify that the preestablished precision requirements are reached, and finally, an analysis of the results to establish some guidelines for the performance of the vertical jump. The results obtained from the research are that first, a vertical jump is detectable in a video by analyzing contours and their movements in the Y axis of the frames, the best position to detect it in a video is to stand sideways to the camera, and finally, the best distance to stand from the camera to detect the jump is 120cm. As the final result obtained, the user is offered a system to add videos of vertical jumps through an Android application and that they are processed automatically without their intervention. This processing is responsible for obtaining all the respective measurements to the jump and showing them to the user in a simple way. Finally, a system is provided to the user to view their history of validated jumps together with the measurements and relevant aspects of each of the vertical jumps.