Achieved Accuracy Values Research Articles

A Deep Learning-Based System to Assist Radiologists in Detecting COVID-19 Disease from Chest Computed Tomography Images

The COVID-19 pandemic has had a significant negative impact on the world in various ways. In an effort to mitigate the negative effects of the pandemic, this study proposes a deep learning approach for the automatic detection of COVID-19 from chest computed tomography (CT) images. This would enable healthcare professionals to more efficiently identify the presence of the virus and provide appropriate care and support to infected individuals. The proposed deep learning approach is based on binary classification and utilizes members of the pre-trained EfficientNet model family. These models were trained on a dataset of real patient images, called the EFSCH-19 dataset, to classify chest CT images as positive or negative for COVID-19. The results of the predictions made on the test images showed that all models achieved accuracy values of over 98%. Among these models, the EfficientNet-B2 model performed the best, with an accuracy of 99.75%, sensitivity of 99.50%, specificity of 100%, and an F1 score of 99.75%. In addition to the high accuracy achieved in the classification of chest CT images using the proposed pre-trained deep learning models, the gradient-weighted class activation mapping (Grad-CAM) method was also applied to further understand and interpret the model's predictions.

Glaucoma Detection in Retinal Fundus Images Based on Deep Transfer Learning and Fuzzy Aggregation Operators

The early diagnosis of the glaucoma disease in the eye is crucial to avoid vision loss. This paper proposes an efficient computer-aided detection (CAD) system for diagnosing glaucoma based on fundus images, deep transfer learning and fuzzy aggregation operators. Specifically, the proposed CAD system includes three stages: (1) Detection of the region of interest of the optic disc using an efficient deep learning network, (2) Classification of images based on different pre-trained deep convolutional neural networks and support vector machines, and (3) Use of fuzzy aggregation operators to fuse the predictions of glaucoma classifiers. We used three popular yet robust aggregators: ordered weighted averaging (OWA) operator, weighted power mean (WPM), and exponential mean (EXM). We assessed the efficacy of the proposed glaucoma CAD system on three public datasets: DRISHTI-GS1, RIM-ONE, and REFUGE. The proposed conjunctive OWA aggregation method (Conj-OWA) achieves the best glaucoma classification results. Specifically, it achieves accuracy values of 90.2%, 97.8%, and 94.3% and area under the curve (AUC) values of 95.3%, 99.8%, and 96.2%, respectively, on DRISHTI-GS1, RIM-ONE, and REFUGE databases.

Sign language recognition by means of common spatial patterns: An analysis.

Currently there are around 466 million hard of hearing people and this amount is expected to grow in the coming years. Despite the efforts that have been made, there is a communication barrier between deaf and hard of hearing signers and non-signers in environments without an interpreter. Different approaches have been developed lately to try to deal with this issue. In this work, we present an Argentinian Sign Language (LSA) recognition system which uses hand landmarks extracted from videos of the LSA64 dataset in order to distinguish between different signs. Different features are extracted from the signals created with the hand landmarks values, which are first transformed by the Common Spatial Patterns (CSP) algorithm. CSP is a dimensionality reduction algorithm and it has been widely used for EEG systems. The features extracted from the transformed signals have been then used to feed different classifiers, such as Random Forest (RF), K-Nearest Neighbors (KNN) or Multilayer Perceptron (MLP). Several experiments have been performed from which promising results have been obtained, achieving accuracy values between 0.90 and 0.95 on a set of 42 signs.

Evaluation of 1D and 2D Deep Convolutional Neural Networks for Driving Event Recognition.

Driving event detection and driver behavior recognition have been widely explored for many purposes, including detecting distractions, classifying driver actions, detecting kidnappings, pricing vehicle insurance, evaluating eco-driving, and managing shared and leased vehicles. Some systems can recognize the main driving events (e.g., accelerating, braking, and turning) by using in-vehicle devices, such as inertial measurement unit (IMU) sensors. In general, feature extraction is a commonly used technique to obtain robust and meaningful information from the sensor signals to guarantee the effectiveness of the subsequent classification algorithm. However, a general assessment of deep neural networks merits further investigation, particularly regarding end-to-end models based on Convolutional Neural Networks (CNNs), which combine two components, namely feature extraction and the classification parts. This paper primarily explores supervised deep-learning models based on 1D and 2D CNNs to classify driving events from the signals of linear acceleration and angular velocity obtained with the IMU sensors of a smartphone placed in the instrument panel of the vehicle. Aggressive and non-aggressive behaviors can be recognized by monitoring driving events, such as accelerating, braking, lane changing, and turning. The experimental results obtained are promising since the best classification model achieved accuracy values of up to 82.40%, and macro- and micro-average F1 scores, respectively, equal to 75.36% and 82.40%, thus, demonstrating high performance in the classification of driving events.

Alzheimer's Disease Diagnosis With Brain Structural MRI Using Multiview-Slice Attention and 3D Convolution Neural Network.

Numerous artificial intelligence (AI) based approaches have been proposed for automatic Alzheimer's disease (AD) prediction with brain structural magnetic resonance imaging (sMRI). Previous studies extract features from the whole brain or individual slices separately, ignoring the properties of multi-view slices and feature complementarity. For this reason, we present a novel AD diagnosis model based on the multiview-slice attention and 3D convolution neural network (3D-CNN). Specifically, we begin by extracting the local slice-level characteristic in various dimensions using multiple sub-networks. Then we proposed a slice-level attention mechanism to emphasize specific 2D-slices to exclude the redundancy features. After that, a 3D-CNN was employed to capture the global subject-level structural changes. Finally, all these 2D and 3D features were fused to obtain more discriminative representations. We conduct the experiments on 1,451 subjects from ADNI-1 and ADNI-2 datasets. Experimental results showed the superiority of our model over the state-of-the-art approaches regarding dementia classification. Specifically, our model achieves accuracy values of 91.1 and 80.1% on ADNI-1 for AD diagnosis and mild cognitive impairment (MCI) convention prediction, respectively.

Riparian ecosystems mapping at fine scale: a density approach based on multi‐temporal UAV photogrammetric point clouds

AbstractIn recent years, numerous directives worldwide have addressed the conservation and restoration of riparian corridors, activities that rely on continuous vegetation mapping to understand its volumetric features and health status. Mapping riparian corridors requires not only fine‐scale resolution but also the coverage of relatively large areas. The use of Unmanned Aerial Vehicles (UAV) allows for meeting both conditions, although the cost‐effectiveness of their use is highly influenced by the type of sensor mounted on them. Few works have so far investigated the use of photogrammetric sensors for individual tree crown detection, despite being cheaper than the most common Light Detection and Ranging (LiDAR) ones. This work aims to improve the individual crown detection from UAV‐photogrammetric datasets in a twofold way. Firstly, the effectiveness of a new approach that has already achieved interesting results in LiDAR applications was tested for photogrammetric point clouds. The test was carried out by comparing the accuracy achieved by the new approach, which is based on the point density features of the analysed dataset, with those related to the more common local maxima and textural methods. The results indicated the potentiality of the density‐based method, which achieved accuracy values (0.76 F‐score) consistent with the traditional methods (0.49–0.80 F‐score range) but was less affected by under‐ and over‐fitting. Secondly, the potential improvement of working on intra‐annual multi‐temporal datasets was assessed by applying the density‐based approach to seven different scenarios, three of which were constituted by single‐epoch datasets and the remaining given by the joining of the others. The F‐score increased from 0.67 to 0.76 when passing from single‐ to multi‐epoch datasets, aligning with the accuracy achieved by the new method when applied to LiDAR data. The results demonstrate the potential of multi‐temporal acquisitions when performing individual crown detection from photogrammetric data.

Sentiment analysis on Bangla text using extended lexicon dictionary and deep learning algorithms

Sentiment analysis (SA) is a subset of natural language processing (NLP) research. In the case of categorical weighted based dictionary with rule-based sentiment score generation, no work in SA has been done yet in Bangla language using deep learning (DL) approaches. This paper proposes DL models for SA on Bangla text using an extended lexicon data dictionary (LDD). We implement the rule-based method Bangla text sentiment score (BTSC) algorithm for extracting polarity from large texts. These polarities are then fed into the neural network along with the preprocessed text as training samples. The preprocessed texts are formatted as a vectorization of words of unique numbers of pre-trained word embedding models. Word2Vec matrix with top highest probability word is applied on embedding layer as a weighted matrix to fit the DL models. This paper also presents a remarkably detailed analysis of selective DL models with some fine tuning. The fine-tuning includes the use of drop out, optimizer regularization, learning rate, multiple layers, filters, attention mechanism, capsule layers, transformer with progressive training along with validation and testing accuracy, precision, recall and F1-score. Experimental results indicate that the proposed new long short-term memory (LSTM) models are highly accurate in performing SA tasks. For our proposed hierarchical attention based LSTM (HAN-LSTM), Dynamic routing based capsule neural network with Bi-LSTM (D-CAPSNET-Bi-LSTM) and bidirectional encoder representations from Transformers (BERT) with LSTM (BERT-LSTM) model we achieved accuracy values of 78.52%, 80.82% and 84.18% respectively.

Sequence Alignment Ensemble With a Single Neural Network for Sequence Labeling

Sequence labeling, in which a class or label is assigned to each token in a given input order, is a fundamental task in natural language processing. Many advanced neural network architectures have recently been proposed to solve the sequential labeling problem affecting this task. By contrast, only a few approaches have been proposed to address the sequential ensemble problem. In this paper, we resolve the sequential ensemble problem by applying the sequential alignment method in a proposed ensemble framework. Specifically, we propose a simple but efficient ensemble candidate generation framework with which multiple heterogeneous systems can easily be prepared from a single neural sequence labeling network. To evaluate the proposed framework, experiments were conducted with part-of-speech (POS) tagging and dependency label prediction problems. The results indicate that the proposed framework achieved accuracy values that were higher by 0.19 and 0.33 than those achieved by the hard-voting method on the Penn-treebank POS-tagged and Universal dependency-tagged datasets, respectively.

A deep learning approach for detecting pneumonia in chest X-rays

Pneumonia causes the death of many children every year and constitutes a certain proportion of the world population. Chest X-rays are primarily used to diagnose this disease, but even for a trained radiologist, chest X-rays are not easy to interpret. In this study, a model for pneumonia detection trained on digital chest X-ray images is presented to assist radiologists in their decision-making processes. The study is carried out on the Phyton platform by using deep learning models, which have been widely preferred recently. In this study, a deep learning framework for pneumonia classification with four different CNN models is proposed. Three of them are pre-trained models, MobileNet, ResNet and AlexNet and the other is the recommended CNN Model. These models are evaluated by comparing them with each other according to their performance. The experimental performance of the proposed deep learning framework is evaluated on the basis of precision, recall and f1-score. The models achieved accuracy values of 93%, 97%, 97% and 86%, respectively. It is clear that the proposed ResNet model achieves the highest results compared to the others.

GraphSynergy: a network-inspired deep learning model for anticancer drug combination prediction.

To develop an end-to-end deep learning framework based on a protein-protein interaction (PPI) network to make synergistic anticancer drug combination predictions. We propose a deep learning framework named Graph Convolutional Network for Drug Synergy (GraphSynergy). GraphSynergy adapts a spatial-based Graph Convolutional Network component to encode the high-order topological relationships in the PPI network of protein modules targeted by a pair of drugs, as well as the protein modules associated with a specific cancer cell line. The pharmacological effects of drug combinations are explicitly evaluated by their therapy and toxicity scores. An attention component is also introduced in GraphSynergy, which aims to capture the pivotal proteins that play a part in both PPI network and biomolecular interactions between drug combinations and cancer cell lines. GraphSynergy outperforms the classic and state-of-the-art models in predicting synergistic drug combinations on the 2 latest drug combination datasets. Specifically, GraphSynergy achieves accuracy values of 0.7553 (11.94% improvement compared to DeepSynergy, the latest published drug combination prediction algorithm) and 0.7557 (10.95% improvement compared to DeepSynergy) on DrugCombDB and Oncology-Screen datasets, respectively. Furthermore, the proteins allocated with high contribution weights during the training of GraphSynergy are proved to play a role in view of molecular functions and biological processes, such as transcription and transcription regulation. The introduction of topological relations between drug combination and cell line within the PPI network can significantly improve the capability of synergistic drug combination identification.

Multi-model and multi-slice ensemble learning architecture based on 2D convolutional neural networks for Alzheimer's disease diagnosis.

Alzheimer's Disease (AD) is a chronic neurodegenerative disease without effective medications or supplemental treatments. Thus, predicting AD progression is crucial for clinical practice and medical research. Due to limited neuroimaging data, two-dimensional convolutional neural networks (2D CNNs) have been commonly adopted to differentiate among cognitively normal subjects (CN), people with mild cognitive impairment (MCI), and AD patients. Therefore, this paper proposes an ensemble learning (EL) architecture based on 2D CNNs, using a multi-model and multi-slice ensemble. First, the top 11 coronal slices of grey matter density maps for AD versus CN classifications were selected. Second, the discriminator of a generative adversarial network, VGG16, and ResNet50 were trained with the selected slices, and the majority voting scheme was used to merge the multi-slice decisions of each model. Afterwards, those three classifiers were used to construct an ensemble model. Multi-slice ensemble learning was designed to obtain spatial features, while multi-model integration reduced the prediction error rate. Finally, transfer learning was used in domain adaptation to refine those CNNs, moving them from working solely with AD versus CN classifications to being applicable to other tasks. This ensemble approach achieved accuracy values of 90.36%, 77.19%, and 72.36% when classifying AD versus CN, AD versus MCI, and MCI versus CN, respectively. Compared with other state-of-the-art 2D studies, the proposed approach provides an effective, accurate, automatic diagnosis along the AD continuum. This technique may enhance AD diagnostics when the sample size is limited.

Auditory Attention Detection via Cross-Modal Attention.

Humans show a remarkable perceptual ability to select the speech stream of interest among multiple competing speakers. Previous studies demonstrated that auditory attention detection (AAD) can infer which speaker is attended by analyzing a listener's electroencephalography (EEG) activities. However, previous AAD approaches perform poorly on short signal segments, more advanced decoding strategies are needed to realize robust real-time AAD. In this study, we propose a novel approach, i.e., cross-modal attention-based AAD (CMAA), to exploit the discriminative features and the correlation between audio and EEG signals. With this mechanism, we hope to dynamically adapt the interactions and fuse cross-modal information by directly attending to audio and EEG features, thereby detecting the auditory attention activities manifested in brain signals. We also validate the CMAA model through data visualization and comprehensive experiments on a publicly available database. Experiments show that the CMAA achieves accuracy values of 82.8, 86.4, and 87.6% for 1-, 2-, and 5-s decision windows under anechoic conditions, respectively; for a 2-s decision window, it achieves an average of 84.1% under real-world reverberant conditions. The proposed CMAA network not only achieves better performance than the conventional linear model, but also outperforms the state-of-the-art non-linear approaches. These results and data visualization suggest that the CMAA model can dynamically adapt the interactions and fuse cross-modal information by directly attending to audio and EEG features in order to improve the AAD performance.

Mapping European Rice Paddy Fields Using Yearly Sequences of Spaceborne Radar Reflectivity: A Case Study in Italy

Although a vast literature exists on satellite-based mapping of rice paddy fields in Asia, where most of the global production takes place, little has been produced so far that focuses on the European context. Detection and mapping methods that work well in the Asian context will not offer the same performance in Europe, where different seasonal cycles, environmental contexts, and rice varieties make distinctive features dissimilar to the Asian case. In this context, water management is a key clue; watering practices are distinctive for rice with respect to other crops, and within rice there exist diverse cultivation practices including organic and non-organic approaches. In this paper, we focus on satellite-observed water management to identify rice paddy fields cultivated with a traditional agricultural approach. Building on established research results, and guided by the output of experiments on real-world cases, a new method for analyzing time-series of Sentinel-1 data has been developed, which can identify traditional rice fields with a high degree of reliability. Typical watering practices for traditional rice cultivation leave distinctive marks on the yearly sequence of spaceborne radar reflectivity that are identified by the proposed classifier. The method is tested on a small sample of rice paddy fields, built by direct collection of ground reference information. Automated setting of parameters was sufficient to achieve accuracy values beyond 90%, and scanning of a range of values led to touch full score on an independent test set. This work is a part of a broader initiative to build space-based tools for collecting additional pieces of evidence to support food chain traceability; the whole system will consider various parameters, whose analysis procedures are still at their early stages of development.

Fast High–Resolution Mass Spectrometry and Chemometrics for Evaluation of Sensory Parameters of Commercial Coffee Blends

This study presents a fast method to estimate sensory parameters of commercial capsules of roasted coffee using flow injection analysis coupled to high–resolution mass spectrometry (FIA–HRMS) as an alternative to traditional sensory analysis, which is a laborious and subjective method. Over 25 types of coffee capsules were studied. The samples were partitioned into an aqueous and organic extract, which were analyzed by FIA–HRMS in the positive and negative ionization modes. Data fusion of such mass spectra was performed to explore the complementary information of sample preparation and ionization conditions. Orthogonalized partial least square discriminant analysis (OPLS–DA) models were built and trained to determine the type of capsule and to estimate important coffee parameters (e.g., acidity, bitterness, body, intensity, and roasting level), achieving accuracy values higher than 91.1%. In addition, variable importance in projection (VIP) scores enabled assignment of the elemental composition and, in some cases, putative identification of compounds in coffee (e.g., caffeine, caffearine, and quinides) that exhibited an important role in class discrimination.

Optimal school site selection in Urban areas using deep neural networks

Finding an optimal location for a future facility amidst existing sites is a challenging task—and potentially has numerous applications in resource planning. Formally, given a set of candidate sites S and existing sites E, the Optimal Site Selection (OSS) problem aims to find the optimal location $$s \in S$$ that serves maximum customers. In this paper, we study the OSS problem for predicting school location from a set of available sites in the Urban areas. We leverage deep neural networks to solve this problem in a resource and time-efficient way—using a dataset of 47,926 schools across 36 districts of the Punjab province. The dataset includes numerous features such as no. of teachers, no. of classrooms, and no. of sections associated with each school site. We also incorporate a novel feature called site influence that gives the expected number of potential students for a particular school. Our proposed model performs exceedingly well with the addition of this feature. We approach this problem by defining class labels based on a given school’s capacity—and further train a deep neural network using the aforementioned features to achieve accuracy values up to 82%. Overall, this study facilitates optimal school site selection in the Urban areas, thereby, adding to communal prosperity and growth.

Computational identification of chemical compounds with potential anti-Chagas activity using a classification tree

ABSTRACT Chagas disease is endemic to 21 Latin American countries and is a great public health problem in that region. Current chemotherapy remains unsatisfactory; consequently the need to search for new drugs persists. Here we present a new approach to identify novel compounds with potential anti-chagasic action. A large dataset of 584 compounds, obtained from the Drugs for Neglected Diseases initiative, was selected to develop the computational model. Dragon software was used to calculate the molecular descriptors and WEKA software to obtain the classification tree. The best model shows accuracy greater than 93.4% for the training set; the tree was also validated using a 10-fold cross-validation procedure and through a test set, achieving accuracy values over 90.5% and 92.2%, correspondingly. The values of sensitivity and specificity were around 90% in all series; also the false alarm rate values were under 10.5% for all sets. In addition, a simulated ligand-based virtual screening for several compounds recently reported as promising anti-chagasic agents was carried out, yielding good agreement between predictions and experimental results. Finally, the present work constitutes an example of how this rational computer-based method can help reduce the cost and increase the rate in which novel compounds are developed against Chagas disease.

Multi-task Learning for Intrusion Detection and Analysis of Computer Network Traffic

Accurate identification of malicious computer network traffic is a challenging task for a number of reasons. This is especially highlighted when a new type of attack is launched because the amount of available data that belongs to this attack can be scarce. Having small amounts of such data makes understanding the behaviour of traffic and building models to accurately discover it more difficult. In this paper we present a novel classification method based on multi-task learning for the accurate identification of malicious network traffic even when little amount of training data is available. We show the effectiveness of our method by carrying out several experiments and comparisons with existing methods using open source data. Our results show that our method outperforms those methods especially when training data is scarce. Particularly, it achieves accuracy values of 98.51% and 99.76% on two computer network traffic dataset settings, whereas a start-ofthe-art algorithm achieves accuracy values of 93.56% and 96.25% on the same settings.

Estimating sex and age from a face: a forensic approach using machine learning based on photo-anthropometric indexes of the Brazilian population.

The facial analysis permits many investigations, some of the most important of which are craniofacial identification, facial recognition, and age and sex estimation. In forensics, photo-anthropometry describes the study of facial growth and allows the identification of patterns in facial skull development, for example, by using a group of cephalometric landmarks to estimate anthropological information. Previous works presented, as indirect applications, the use of photo-anthropometric measurements to estimate anthropological information such as age and sex. In several areas, automation of manual procedures has achieved advantages over and similar measurement confidence as a forensic expert. This manuscript presents an approach using photo-anthropometric indexes, generated from frontal faces cephalometric landmarks of the Brazilian population, to create an artificial neural network classifier that allows the estimation of anthropological information, in this specific case age and sex. This work is focused on four tasks: (i) sex estimation on ages from 5 to 22 years old, evaluating the interference of age on sex estimation; (ii) age estimation from photo-anthropometric indexes for four age intervals (1 year, 2 years, 4 years, and 5 years); (iii) age group estimation for thresholds of over 14 and over 18 years old; and; (iv) the provision of a new data set, available for academic purposes only, with a large and complete set of facial photo-anthropometric points marked and checked by forensic experts, measured from over 18,000 faces of individuals from Brazil over the last 4 years. The proposed binary classifier obtained significant results, using this new data set, for the sex estimation of individuals over 14 years old, achieving accuracy values higher than 0.85 by the F1 measure. For age estimation, the accuracy results are 0.72 for the F1 measure with an age interval of 5 years. For the age group estimation, the F1 measures of accuracy are higher than 0.93 and 0.83 for thresholds of 14 and 18 years, respectively.

Towards an approach using grammars for automatic classification of masses in mammograms

AbstractApproximately 15% of all cancer deaths among women worldwide is due to breast cancer. Mammography is one of the most useful methods for the early detection of this disease. Over the last decade, several papers were published reporting the usage of different computer‐aided diagnosis systems using pattern recognition techniques as a second opinion to obtain a more accurate diagnosis. However, the theory of formal languages has not been explored in this field. In this context, the main contribution of this study is to present the usage of a new syntactic approach that is able to classify breast masses found in mammograms as benign or malignant. The experimental tests were performed using a dataset that contains 111 images from different sources. The grammar‐based classifiers achieved accuracy values ranging from 89% to 100% depending on the features and the model employed. Furthermore, to achieve a feature dimension reduction, a feature selection technique based on the Gini importance of each feature was employed. Additionally, we compared the obtained results with the grammar‐based classifiers to the more traditional classifiers used in this research area, such as artificial neural networks, support vector machines, k‐nearest neighbors, and random forest. The best result achieved by the grammar‐based classifiers was approximately 10% higher, in terms of accuracy, than the best results produced by the traditional classifiers, showing the strength of this grammatical approach.

Multi-Task Siamese Network for Retinal Artery/Vein Separation via Deep Convolution Along Vessel

Vascular tree disentanglement and vessel type classification are two crucial steps of the graph-based method for retinal artery-vein (A/V) separation. Existing approaches treat them as two independent tasks and mostly rely on ad hoc rules (e.g. change of vessel directions) and hand-crafted features (e.g. color, thickness) to handle them respectively. However, we argue that the two tasks are highly correlated and should be handled jointly since knowing the A/V type can unravel those highly entangled vascular trees, which in turn helps to infer the types of connected vessels that are hard to classify based on only appearance. Therefore, designing features and models isolatedly for the two tasks often leads to a suboptimal solution of A/V separation. In view of this, this paper proposes a multi-task siamese network which aims to learn the two tasks jointly and thus yields more robust deep features for accurate A/V separation. Specifically, we first introduce Convolution Along Vessel (CAV) to extract the visual features by convolving a fundus image along vessel segments, and the geometric features by tracking the directions of blood flow in vessels. The siamese network is then trained to learn multiple tasks: i) classifying A/V types of vessel segments using visual features only, and ii) estimating the similarity of every two connected segments by comparing their visual and geometric features in order to disentangle the vasculature into individual vessel trees. Finally, the results of two tasks mutually correct each other to accomplish final A/V separation. Experimental results demonstrate that our method can achieve accuracy values of 94.7%, 96.9%, and 94.5% on three major databases (DRIVE, INSPIRE, WIDE) respectively, which outperforms recent state-of-the-arts.