Deep Classification Research Articles

A Hierarchical Approach toward Prediction of Human Biological Age from Masked Facial Image Leveraging Deep Learning Techniques

The lifestyle of humans has changed noticeably since the contagious COVID-19 disease struck globally. People should wear a face mask as a protective measure to curb the spread of the contagious disease. Consequently, real-world applications (i.e., electronic customer relationship management) dealing with human ages extracted from face images must migrate to a robust system proficient to estimate the age of a person wearing a face mask. In this paper, we proposed a hierarchical age estimation model from masked facial images in a group-to-specific manner rather than a single regression model because age progression across different age groups is quite dissimilar. Our intention was to squeeze the feature space among limited age classes so that the model could fairly discern age. We generated a synthetic masked face image dataset over the IMDB-WIKI face image dataset to train and validate our proposed model due to the absence of a benchmark masked face image dataset with real age annotations. We somewhat mitigated the data sparsity problem of the large public IMDB-WIKI dataset using off-the-shelf down-sampling and up-sampling techniques as required. The age estimation task was fully modeled like a deep classification problem, and expected ages were formulated from SoftMax probabilities. We performed a classification task by deploying multiple low-memory and higher-accuracy-based convolutional neural networks (CNNs). Our proposed hierarchical framework demonstrated marginal improvement in terms of mean absolute error (MAE) compared to the one-off model approach for masked face real age estimation. Moreover, this research is perhaps the maiden attempt to estimate the real age of a person from his/her masked face image.

A Metaheuristic Approach to Map Driving Pattern for Analyzing Driver Behavior Using Big Data Analysis

The modern-day influx of vehicular traffic along with rapid expansion of roadways has made the selection of the best driver based on driving best practices an imperative, thus optimizing cost and ensuring safe arrival at the destination. A key factor in this is the analysis of driver behavior based on driver activities by monitoring adherence to the features comprising the established driving principles. In general, indiscriminate use of features to predict driver performance can increase process complexity due to inclusion of redundant features. An effective knowledge-based approach with a reduced set of features can help attune the driver behavior and improve driving patterns. Hence, a Deep Mutual Invariance Feature Classification (DMIFC) model has been proposed in this study for predicting driver performance to recommend the best driver. To achieve this, first, the driver behavior is broken down into various features corresponding to a simulated driving dataset and subjected to preprocessing to reduce the noise and form a redundant dataset. Thereafter, a Mutual Invariance Scale Feature Selection (MISFS) filter is used to select the relational features by calculating the spectral variance weight between mutual features. The observed mutual features are promoted to create a dominant pattern to estimate the Max feature-pattern generation using Driver Activity Intense Rate (DAIR). The features are then selected for classification based on the DAIR weightage. Additionally, the Interclass-ReLU (Rectified Linear Unit) is used to generate activation functions to produce logical neurons. The logical neurons are further optimized with Multiperceptron Radial Basis Function Networks (MP-RBFNs) to enable better classification of driver features for best prediction results. The proposed system was found to improve the driver pattern prediction accuracy and enable optimal recommendations of driving principles to the driver.

Comparison of Deep Learning Models for the Classification of Noctilucent Cloud Images

Optically thin layers of tiny ice particles near the summer mesopause, known as noctilucent clouds, are of significant interest within the aeronomy and climate science communities. Ground-based optical cameras mounted at various locations in the arctic regions collect the dataset during favorable summer times. In this paper, first, we compare the performances of various deep learning-based image classifiers against a baseline machine learning model trained with support vector machine (SVM) algorithm to identify an effective and lightweight model for the classification of noctilucent clouds. The SVM classifier is trained with histogram of oriented gradient (HOG) features, and deep learning models such as SqueezeNet, ShuffleNet, MobileNet, and Resnet are fine-tuned based on the dataset. The dataset includes images observed from different locations in northern Europe with varied weather conditions. Second, we investigate the most informative pixels for the classification decision on test images. The pixel-level attributions calculated using the guide back-propagation algorithm are visualized as saliency maps. Our results indicate that the SqueezeNet model achieves an F1 score of 0.95. In addition, SqueezeNet is the lightest model used in our experiments, and the saliency maps obtained for a set of test images correspond better with relevant regions associated with noctilucent clouds.

NEURAL NETWORK SEGMENTATION IN MRIs IMAGES FOR BRAIN TUMORs SEGMENTATION

Individual of the most difficult challenges in medical image processing is detecting brain tumors. The challenge is challenging to complete since the photographs have a lots of variety, as brain tumors exist in a variety of shapes and textures. Tumors can appear in a variety of areas, and the locations of a tumor can reveal information about the sort of cells that are creating it, which can help with further diagnosis. The picture intensities of tumor and non-tumor images can overlap, making it challenging for any model to make accurate predictions from raw images. Magnetic Resonances Imaging (MRI)sis a common imaging tool for assessing these tumors, but the vast amount of data produced by MRI makes manual segmentation impossible in a reasonable length of time, limiting the use of exact quantitative values in clinical practice. As a result, approaches for automatic and reliable segmentation are necessary. Because it gives Relevant information for the diagnosis, Monitoring, and therapy of brain tumors, segmentation is individual of the most important processes in interpreting medical pictures. Ins this papers, we offers a Deep Neural Network-based automatic segmentation and classification approach.

Smart Healthcare: Disease Prediction Using the Cuckoo-Enabled Deep Classifier in IoT Framework

The Internet of Things (IoT) is commonly employed to detect different kinds of diseases in the health sector. Presently, disease detection is performed using MRI images, X-rays, CT scans, and so on for diagnosing the diseases. The manual detection process is found to be time-consuming and may result in detection errors that affect the diagnosis. Hence, there is a need for an automatic system for which the deep learning methods gain a major interest. Hence, the idea to combine deep learning and disease prediction to effectively predict the disease is initiated. In this research, the deep learning method is combined with deep learning for the effective prediction of diseases, where the IoT network is employed in the data collection from the patients. The proposed cuckoo-based deep convolutional long-short term memory (deep convLSTM) classifier is employed for disease prediction, where the cuckoo search optimization is utilized for tuning the deep convLSTM classifier. The proposed method is compared with the conventional methods, and it achieved a training percentage of 97.591%, 95.874%, and 97.094%, respectively, for accuracy, sensitivity, and specificity. The comparative analysis proved that the proposed method obtained higher accuracy than other methods.

Adversarial Robustness of Deep Reinforcement Learning Based Dynamic Recommender Systems.

Adversarial attacks, e.g., adversarial perturbations of the input and adversarial samples, pose significant challenges to machine learning and deep learning techniques, including interactive recommendation systems. The latent embedding space of those techniques makes adversarial attacks challenging to detect at an early stage. Recent advance in causality shows that counterfactual can also be considered one of the ways to generate the adversarial samples drawn from different distribution as the training samples. We propose to explore adversarial examples and attack agnostic detection on reinforcement learning (RL)-based interactive recommendation systems. We first craft different types of adversarial examples by adding perturbations to the input and intervening on the casual factors. Then, we augment recommendation systems by detecting potential attacks with a deep learning-based classifier based on the crafted data. Finally, we study the attack strength and frequency of adversarial examples and evaluate our model on standard datasets with multiple crafting methods. Our extensive experiments show that most adversarial attacks are effective, and both attack strength and attack frequency impact the attack performance. The strategically-timed attack achieves comparative attack performance with only 1/3 to 1/2 attack frequency. Besides, our white-box detector trained with one crafting method has the generalization ability over several other crafting methods.

Chin Electromyogram, an Effectual and Useful Biosignal for the Diagnosis of Obstructive Sleep Apnea

Background and Objective: Obstructive sleep apnea (OSA) is among the critical sleep disorders, and researchers have been investigating its novel diagnostic methods. Polysomnography signals' complexity, difficult visual interpretation, and the need for an efficient algorithm based on simpler signals have made the study of sleep apnea a compelling issue. In this study, the accuracy of chin electromyogram in the diagnosis of OSA was evaluated.
 Materials and Methods: The amplitude variation and power spectral density (PSD) of chin electromyograms of 100 patients during apnea and before-after apnea occurrences (non-apnea) periods were compared after complete processing of the raw signal. Two-dimensional (2D) spectrograms related to the specified periods were extracted and fed into the residual neural network (ResNet). The network performance was reported by model evaluation parameters.
 Results: The results showed that OSA event influences the patient's chin muscle and increases the amplitude variances and power spectrum of the chin electromyogram. The ResNet-50 deep model classified the dataset of this sleep disorder with about 97% accuracy, which was higher than previous studies in this field.
 Conclusion: Chin electromyogram can be introduced as a practical and useful biosignal for accurate OSA diagnosis with a deep classifier without the need for current specialized equipment and multiple vital signals.

DGCyTOF: Deep learning with graphic cluster visualization to predict cell types of single cell mass cytometry data.

Single-cell mass cytometry, also known as cytometry by time of flight (CyTOF) is a powerful high-throughput technology that allows analysis of up to 50 protein markers per cell for the quantification and classification of single cells. Traditional manual gating utilized to identify new cell populations has been inadequate, inefficient, unreliable, and difficult to use, and no algorithms to identify both calibration and new cell populations has been well established. A deep learning with graphic cluster (DGCyTOF) visualization is developed as a new integrated embedding visualization approach in identifying canonical and new cell types. The DGCyTOF combines deep-learning classification and hierarchical stable-clustering methods to sequentially build a tri-layer construct for known cell types and the identification of new cell types. First, deep classification learning is constructed to distinguish calibration cell populations from all cells by softmax classification assignment under a probability threshold, and graph embedding clustering is then used to identify new cell populations sequentially. In the middle of two-layer, cell labels are automatically adjusted between new and unknown cell populations via a feedback loop using an iteration calibration system to reduce the rate of error in the identification of cell types, and a 3-dimensional (3D) visualization platform is finally developed to display the cell clusters with all cell-population types annotated. Utilizing two benchmark CyTOF databases comprising up to 43 million cells, we compared accuracy and speed in the identification of cell types among DGCyTOF, DeepCyTOF, and other technologies including dimension reduction with clustering, including Principal Component Analysis (PCA), Factor Analysis (FA), Independent Component Analysis (ICA), Isometric Feature Mapping (Isomap), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP) with k-means clustering and Gaussian mixture clustering. We observed the DGCyTOF represents a robust complete learning system with high accuracy, speed and visualization by eight measurement criteria. The DGCyTOF displayed F-scores of 0.9921 for CyTOF1 and 0.9992 for CyTOF2 datasets, whereas those scores were only 0.507 and 0.529 for the t-SNE+k-means; 0.565 and 0.59, for UMAP+ k-means. Comparison of DGCyTOF with t-SNE and UMAP visualization in accuracy demonstrated its approximately 35% superiority in predicting cell types. In addition, observation of cell-population distribution was more intuitive in the 3D visualization in DGCyTOF than t-SNE and UMAP visualization. The DGCyTOF model can automatically assign known labels to single cells with high accuracy using deep-learning classification assembling with traditional graph-clustering and dimension-reduction strategies. Guided by a calibration system, the model seeks optimal accuracy balance among calibration cell populations and unknown cell types, yielding a complete and robust learning system that is highly accurate in the identification of cell populations compared to results using other methods in the analysis of single-cell CyTOF data. Application of the DGCyTOF method to identify cell populations could be extended to the analysis of single-cell RNASeq data and other omics data.

Grading diabetic retinopathy and prostate cancer diagnostic images with deep quantum ordinal regression

Although for many diseases there is a progressive diagnosis scale, automatic analysis of grade-based medical images is quite often addressed as a binary classification problem, missing the finer distinction and intrinsic relation between the different possible stages or grades. Ordinal regression (or classification) considers the order of the values of the categorical labels and thus takes into account the order of grading scales used to assess the severity of different medical conditions. This paper presents a quantum-inspired deep probabilistic learning ordinal regression model for medical image diagnosis that takes advantage of the representational power of deep learning and the intrinsic ordinal information of disease stages. The method is evaluated on two different medical image analysis tasks: prostate cancer diagnosis and diabetic retinopathy grade estimation on eye fundus images. The experimental results show that the proposed method not only improves the diagnosis performance on the two tasks but also the interpretability of the results by quantifying the uncertainty of the predictions in comparison to conventional deep classification and regression architectures. The code and datasets are available at https://github.com/stoledoc/DQOR.

Deep Neural Network-Based Business Data Classification in Intelligent Business Management

The purpose of this paper is to explore how intelligent data mining technology can be used to improve the customer service capability of commercial companies. Based on extensive research on commercial business, this paper uses data mining and machine learning techniques to build an overall framework for applying intelligent technologies to business improvement, and uses multilayer perceptrons and integrated learning algorithms to build classifiers for customer segmentation; uses association rule mining to assist commercial companies in business decisions; uses clustering algorithms and visualization techniques to further analyze claims cases and assist in commercial fraud detection. The multilayer perceptron classification makes the classification of commercial customers more detailed and reasonable, and the company’s business staff can sell products in a more targeted manner; association rule mining greatly improves the quality and efficiency of the company’s management’s decision making.

CPRO: Competitive Poor and Rich Optimizer-Enabled Deep Learning Model and Holoentropy Weighted-Power K-Means Clustering for Brain Tumor Classification Using MRI

A brain tumor is a collection of irregular and needless cell development in the brain region, and it is considered a life-threatening disease. Therefore, early level segmentation and brain tumor detection with Magnetic Resonance Imaging (MRI) is more important to save the patient’s life. Moreover, MRI is more effective in identifying patients with brain tumors since the recognition of this modality is moderately larger than considering other imaging modalities. The classification of brain tumors is the most important, difficult task in medical imaging systems because of size, appearance and shape variations. In this paper, Competitive Poor and Rich Optimization (CPRO)-based Deep Quantum Neural Network (Deep QNN) is proposed for brain tumor classification. Additionally, the pre-processing process assists in eradicating noises and uses image intensity to eliminate the artifacts. The significant features are extracted from pre-processed image to perform a productive classification process. The Deep QNN classifier is employed for classifying the brain tumor regions. Besides, the Deep QNN classifier is trained by the developed CPRO approach, which is newly designed by integrating Poor and Rich Optimization (PRO) and Competitive Swarm Optimizer (CSO). The developed brain tumor detection model outperformed other existing models with accuracy, sensitivity and specificity of 94.44%, 97.60% and 93.78%.

A novel deep classifier framework for automated molecular subtyping of breast carcinoma using immunohistochemistry image analysis

Breast carcinoma has various subtypes based on the genetic factors involved in the pathogenesis of the malignancy. Identifying the exact subtype and providing targeted treatment to the patient can improve the survival chances. Molecular subtyping through immunohistochemistry analysis is a pathology procedure to determine the subtype of breast cancer. The existing manual procedure is tedious and involves assessing the status of the four vital molecular biomarkers present in the tumor tissues. In this paper, a deep learning-based framework for automated molecular subtyping of breast cancer is proposed. Digital slide images of the four biomarkers are separately processed by the proposed framework. In the preprocessing stage, the non-informative background regions from the images are separated. The patches extracted from the foreground regions are classified into target classes using convolutional neural network models trained for this purpose. Classification results are post-processed to predict the status of all the four biomarkers. The predictions for the individual biomarkers are finally consolidated as per clinical guidelines to determine the subtype of the cancer. The proposed system is evaluated for the performance of individual biomarker status prediction and patient-level subtype classification.For patient-level evaluation of biomarkers ER, PR, K67, and HER2, the proposed method gives F1 Scores 1.00, 1.00, 0.90, and 0.94 respectively, whereas for molecular subtyping an F1 score of 0.89 is obtained. In both these aspects, the proposed framework has given significant results that show the effectiveness of our approach.

IDC: quantitative evaluation benchmark of interpretation methods for deep text classification models

Recent advances in deep neural networks have achieved outstanding success in natural language processing tasks. Interpretation methods that provide insight into the decision-making process of these models have received an influx of research attention because of the success and the black-box nature of the deep text classification models. Evaluation of these methods has been based on changes in classification accuracy or prediction confidence when removing important words identified by these methods. There are no measurements of the actual difference between the predicted important words and humans’ interpretation of ground truth because of the lack of interpretation ground truth. A large publicly available interpretation ground truth has the potential to advance the development of interpretation methods. Manual labeling important words for each document to create a large interpretation ground truth is very time-consuming. This paper presents (1) IDC, a new benchmark for quantitative evaluation of interpretation methods for deep text classification models, and (2) evaluation of six interpretation methods using the benchmark. The IDC benchmark consists of: (1) Three methods that generate three pseudo-interpretation ground truth datasets. (2) Three performance metrics: interpretation recall, interpretation precision, and Cohen’s kappa inter-agreement. Findings: IDC-generated interpretation ground truth agrees with human annotators on sampled movie reviews. IDC identifies Layer-wise Relevance Propagation and the gradient-by-input methods as the winning interpretation methods in this study.

Radiology report generation for proximal femur fractures using deep classification and language generation models

Proximal femur fractures represent a major health concern, and substantially contribute to the morbidity of elderly. Correct classification and diagnosis of hip fractures has a significant impact on mortality, costs and hospital stay. In this paper, we present a method and empirical validation for automatic subclassification of proximal femur fractures and Dutch radiological report generation that does not rely on manually curated data. The fracture classification model was trained on 11,000 X-ray images obtained from 5000 electronic health records in a general hospital. To generate the Dutch reports, we first trained an embedding model on 20,000 radiological reports of pelvic region fractures, and used its embeddings in the report generation model. We trained the report generation model on the 5000 radiological reports associated with the fracture cases. Our report generation model is on par with state-of-the-art in terms of BLEU and ROUGE scores. This is promising, because in contrast to those earlier works, our approach does not require manual preprocessing of either images or the reports. This boosts the applicability of automatic clinical report generation in practice. A quantitative and qualitative user study among medical students found no significant difference in provenance of real and generated reports. A qualitative, in-depth clinical relevance study with medical domain experts showed that from a human perspective the quality of the generated reports approximates the quality of the original reports and highlights challenges in creating sufficiently detailed and versatile training data for automatic radiology report generation.

Adaptive deep learning for deep COVID-19 diagnosis

PurposeCOVID-19 has occurred in more than 150 countries and causes a huge impact on the health of many people. The main purpose of this work is, COVID-19 has occurred in more than 150 countries and causes a huge impact on the health of many people. The COVID-19 diagnosis is required to detect at the beginning stage and special attention should be given to them. The fastest way to detect the COVID-19 infected patients is detecting through radiology and radiography images. The few early studies describe the particular abnormalities of the infected patients in the chest radiograms. Even though some of the challenges occur in concluding the viral infection traces in X-ray images, the convolutional neural network (CNN) can determine the patterns of data between the normal and infected X-rays that increase the detection rate. Therefore, the researchers are focusing on developing a deep learning-based detection model.Design/methodology/approachThe main intention of this proposal is to develop the enhanced lung segmentation and classification of diagnosing the COVID-19. The main processes of the proposed model are image pre-processing, lung segmentation and deep classification. Initially, the image enhancement is performed by contrast enhancement and filtering approaches. Once the image is pre-processed, the optimal lung segmentation is done by the adaptive fuzzy-based region growing (AFRG) technique, in which the constant function for fusion is optimized by the modified deer hunting optimization algorithm (M-DHOA). Further, a well-performing deep learning algorithm termed adaptive CNN (A-CNN) is adopted for performing the classification, in which the hidden neurons are tuned by the proposed DHOA to enhance the detection accuracy. The simulation results illustrate that the proposed model has more possibilities to increase the COVID-19 testing methods on the publicly available data sets.FindingsFrom the experimental analysis, the accuracy of the proposed M-DHOA–CNN was 5.84%, 5.23%, 6.25% and 8.33% superior to recurrent neural network, neural networks, support vector machine and K-nearest neighbor, respectively. Thus, the segmentation and classification performance of the developed COVID-19 diagnosis by AFRG and A-CNN has outperformed the existing techniques.Originality/valueThis paper adopts the latest optimization algorithm called M-DHOA to improve the performance of lung segmentation and classification in COVID-19 diagnosis using adaptive K-means with region growing fusion and A-CNN. To the best of the authors’ knowledge, this is the first work that uses M-DHOA for improved segmentation and classification steps for increasing the convergence rate of diagnosis.

A computer-aided brain tumor diagnosis by adaptive fuzzy active contour fusion model and deep fuzzy classifier

A computer-aided brain tumor diagnosis by adaptive fuzzy active contour fusion model and deep fuzzy classifier

A Sparsity-Limitation-Based High-Dimensional Distribution Searching Algorithm for Adversarial Attack

Deep neural networks-based image classification systems could suffer from adversarial attack algorithms, which generate input examples by adding deliberately crafted yet imperceptible noise to original inputs. To reduce the impact on human visual sense and to ensure adversarial attack ability, the input image needs to be modified by pixels in considerable iterations which is time consuming. By using sparse mapping network to map the input into a higher dimensional space, searching space of adversarial perturbation distribution is enlarged to better acquire perturbation information. Taking both searching speed and searching effectiveness into consideration, sparsity limitation is introduced to suppress unnecessary neurons during parameter updating process. Based on different eye sensitivity of different colors, maps of each color channel are disturbed by perturbations with different strengths to reduce visual perception. Numerical experiments confirm that compared with the state-of-the-art adversarial attack algorithms, the proposed SparseAdv performs a relatively high attack ability, better imperceptible visualization, and faster generation speed.

Robustness verification of ReLU networks via quadratic programming

Neural networks are known to be sensitive to adversarial perturbations. To investigate this undesired behavior we consider the problem of computing the distance to the decision boundary (DtDB) from a given sample for a deep neural net classifier. In this work we present a procedure where we solve a convex quadratic programming (QP) task to obtain a lower bound on the DtDB. This bound is used as a robustness certificate of the classifier around a given sample. We show that our approach provides better or competitive results in comparison with a wide range of existing techniques.

An intelligent railway surveillance framework based on recognition of object and railway track using deep learning

In high speed railways, the intelligent railway safety system is necessary to avoid the accidents due to collision between trains and obstacles on the railway track. The unceasing research work is being performed to reinforce the railway safety and to diminish the accident rates. The rapid development in the field of deep learning has prompted new research opportunities in this area. In this paper, a novel and efficient approach is proposed to recognize the objects (obstacles) on the railway track ahead the train using deep classifier network. The 2-D Singular Spectrum Analysis (SSA) is utilized as decomposition tool that decomposes the image in useful components. That component is further applied to the deep classifier network. The obstacle recognition performance is enhanced by the combination of 2D-SSA and deep network. This method also presents a novel measure to identify the railway tracks. In addition, the performance of this approach is analyzed under different illumination conditions using OSU thermal pedestrian benchmark database. This system can be a tremendous support to curtail rail accidental rate and monetary loads. The results of proposed approach present good accuracy as well as can effectively recognize the objects (obstacles) on the railway track which helps to the railway safety. It also achieves a better performance with 85.2% accuracy, 84.5% precision and 88.6% recall.

Deep learning representations to support COVID-19 diagnosis on CT slices.

Introduction: The coronavirus disease 2019 (COVID-19) has become a significant public health problem worldwide. In this context, CT-scan automatic analysis has emerged as a COVID-19 complementary diagnosis tool allowing for radiological finding characterization, patient categorization, and disease follow-up. However, this analysis depends on the radiologist’s expertise, which may result in subjective evaluations.Objective: To explore deep learning representations, trained from thoracic CT-slices, to automatically distinguish COVID-19 disease from control samples.Materials and methods: Two datasets were used: SARS-CoV-2 CT Scan (Set-1) and FOSCAL clinic’s dataset (Set-2). The deep representations took advantage of supervised learning models previously trained on the natural image domain, which were adjusted following a transfer learning scheme. The deep classification was carried out: (a) via an end-to-end deep learning approach and (b) via random forest and support vector machine classifiers by feeding the deep representation embedding vectors into these classifiers.Results: The end-to-end classification achieved an average accuracy of 92.33% (89.70% precision) for Set-1 and 96.99% (96.62% precision) for Set-2. The deep feature embedding with a support vector machine achieved an average accuracy of 91.40% (95.77% precision) and 96.00% (94.74% precision) for Set-1 and Set-2, respectively.Conclusion: Deep representations have achieved outstanding performance in the identification of COVID-19 cases on CT scans demonstrating good characterization of the COVID-19 radiological patterns. These representations could potentially support the COVID-19 diagnosis in clinical settings.