Showing papers in "Biomedical Signal Processing and Control in 2021"

Application of deep learning techniques for detection of COVID-19 cases using chest X-ray images: A comprehensive study.

Abstract: The emergence of Coronavirus Disease 2019 (COVID-19) in early December 2019 has caused immense damage to health and global well-being. Currently, there are approximately five million confirmed cases and the novel virus is still spreading rapidly all over the world. Many hospitals across the globe are not yet equipped with an adequate amount of testing kits and the manual Reverse Transcription-Polymerase Chain Reaction (RT-PCR) test is time-consuming and troublesome. It is hence very important to design an automated and early diagnosis system which can provide fast decision and greatly reduce the diagnosis error. The chest X-ray images along with emerging Artificial Intelligence (AI) methodologies, in particular Deep Learning (DL) algorithms have recently become a worthy choice for early COVID-19 screening. This paper proposes a DL assisted automated method using X-ray images for early diagnosis of COVID-19 infection. We evaluate the effectiveness of eight pre-trained Convolutional Neural Network (CNN) models such as AlexNet, VGG-16, GoogleNet, MobileNet-V2, SqueezeNet, ResNet-34, ResNet-50 and Inception-V3 for classification of COVID-19 from normal cases. Also, comparative analyses have been made among these models by considering several important factors such as batch size, learning rate, number of epochs, and type of optimizers with an aim to find the best suited model. The models have been validated on publicly available chest X-ray images and the best performance is obtained by ResNet-34 with an accuracy of 98.33%. This study will be useful for researchers to think for the design of more effective CNN based models for early COVID-19 detection.

Deep learning for motor imagery EEG-based classification: A review

Abstract: Objectives The availability of large and varied Electroencephalogram (EEG) datasets, rapidly advances and inventions in deep learning techniques, and highly powerful and diversified computing systems have all permitted to easily analyzing those datasets and discovering vital information within. However, the classification process of EEG signals and discovering vital information should be robust, automatic, and with high accuracy. Motor Imagery (MI) EEG has attracted us due to its significant applications in daily life. Methods This paper attempts to achieve those goals throughout a systematic review of the state-of-the-art studies within this field of research. The process began by intensely surfing the well-known specialized digital libraries and, as a result, 40 related papers were gathered. The papers were scrutinized upon multiple noteworthy technical issues, among them deep neural network architecture, input formulation, number of MI EEG tasks, and frequency range of interest. Outcomes Deep neural networks build robust and automated systems for the classification of MI EEG recordings by exploiting the whole input data throughout learning salient features. Specifically, convolutional neural networks (CNN) and hybrid-CNN (h-CNN) are the dominant architectures with high performance in comparison to public datasets with other types of architectures. The MI related datasets, input formulation, frequency ranges, and preprocessing and regularization methods were also reviewed. Inferences This review gives the required preliminaries in developing MI EEG-based BCI systems. The review process of the published articles in the last five years aims to help in choosing the appropriate deep neural network architecture and other hyperparameters for developing those systems.

A Fully Automated Deep Learning-based Network For Detecting COVID-19 from a New And Large Lung CT Scan Dataset.

Abstract: This paper aims to propose a high-speed and accurate fully-automated method to detect COVID-19 from the patient's chest CT scan images. We introduce a new dataset that contains 48,260 CT scan images from 282 normal persons and 15,589 images from 95 patients with COVID-19 infections. At the first stage, this system runs our proposed image processing algorithm that analyzes the view of the lung to discard those CT images that inside the lung is not properly visible in them. This action helps to reduce the processing time and false detections. At the next stage, we introduce a novel architecture for improving the classification accuracy of convolutional networks on images containing small important objects. Our architecture applies a new feature pyramid network designed for classification problems to the ResNet50V2 model so the model becomes able to investigate different resolutions of the image and do not lose the data of small objects. As the infections of COVID-19 exist in various scales, especially many of them are tiny, using our method helps to increase the classification performance remarkably. After running these two phases, the system determines the condition of the patient using a selected threshold. We are the first to evaluate our system in two different ways on Xception, ResNet50V2, and our model. In the single image classification stage, our model achieved 98.49% accuracy on more than 7996 test images. At the patient condition identification phase, the system correctly identified almost 234 of 245 patients with high speed. Our dataset is accessible at https://github.com/mr7495/COVID-CTset .

Automated Atrial Fibrillation Detection using a Hybrid CNN-LSTM Network on Imbalanced ECG Datasets

Abstract: Atrial fibrillation is a heart arrhythmia strongly associated with other heart-related complications that can increase the risk of strokes and heart failure. Manual electrocardiogram (ECG) interpretation for its diagnosis is tedious, time-consuming, requires high expertise, and suffers from inter- and intra-observer variability. Deep learning techniques could be exploited in order for robust arrhythmia detection models to be designed. In this paper, we propose a novel hybrid neural model utilizing focal loss, an improved version of cross-entropy loss, to deal with training data imbalance. ECG features initially extracted via a Convolutional Neural Network (CNN) are input to a Long Short-Term Memory (LSTM) model for temporal dynamics memorization and thus, more accurate classification into the four ECG rhythm types, namely normal (N), atrial fibrillation (AFIB), atrial flutter (AFL) and AV junctional rhythm (J). The model was trained on the MIT-BIH Atrial Fibrillation Database and achieved a sensitivity of 97.87%, and specificity of 99.29% using a ten-fold cross-validation strategy. The proposed model can aid clinicians to detect common atrial fibrillation in real-time on routine screening ECG.

Deep learning architecture based on segmented fundus image features for classification of diabetic retinopathy

Abstract: Diabetic retinopathy is ophthalmological distress, diabetic patients suffer due to clots, lesions, or haemorrhage formation in the light-sensitive region of the retina. Blocking of vessels leads, due to the increase of blood sugar leads to the formation of new vessel growth, which gives rise to mesh-like structures. Assessing the branching retinal vasculature is an important aspect for ophthalmologists for efficient diagnosis. The fundus scans of the eye are first subjected to pre-processing, followed by segmentation. To extract the branching blood vessels, the technique of maximal principal curvature has been applied, which utilizes the maximum Eigenvalues of the Hessian matrix. Adaptive histogram equalization and the morphological opening, are performed post to that, to enhance and eliminate falsely segmented regions. The proliferation of optical nerves was observed much greater in diabetic or affected patients than in healthy ones. We have used a convolution neural network (CNN) to train the classifier for performing classification. The CNN, constructed for classification, comprises a combination of squeeze and excitation and bottleneck layers, one for each class, and a convolution and pooling layer architecture for classification between the two classes. For the performance evaluation of the proposed algorithm, we use the dataset DIARETDB1 (standard Diabetic Retinopathy Dataset) and the dataset provided by a medical institution, comprised of fundus scans of both affected and normal retinas. Experimental results show that the proposed algorithm provides improved results, when compared to traditional schemes. The model yielded an accuracy of 98.7 % and a precision of 97.2 % while evaluated on the DIARETDB1 dataset.

An image quality enhancement scheme employing adolescent identity search algorithm in the NSST domain for multimodal medical image fusion

Abstract: The fast-developing Image fusion technique has become a necessary one in every field. Analyzing the efficiency of various fusion technologies analytically and objectively are spotted as an essentially required processes. Further, Image fusion becomes an inseparable technique in the medical field, since the role of medical images in diagnosing and identifying diseases becomes a crucial task for the radiologists and doctors at its early stage. Different modalities used in clinical applications offer unique information, unlike any other in any form. To diagnose diseases with high accuracy, clinicians require data from more than one modality. Multimodal image fusion has received wide popularity in the medical field since it enhances the accuracy of the clinical diagnosis thereby fusing the complementary information present in more than one image. Obtaining optimal value along with a reduction in cost and time in multimodal medical image fusions are a critical one. Here, in this paper a new multi-modality algorithm for medical image fusion based on the Adolescent Identity Search Algorithm (AISA) for the Non-Subsampled Shearlet Transform is proposed to obtain image optimization and to reduce the computational cost and time. The NSST is a multi-directional and multi-dimensional example of a multiscale and multi-directional wavelet transform. The input source image is decomposed into the NSST subbands at the initial stage. The boundary measure is modulated by the Adolescent Identity Search Algorithm (AISA) that fuses the sub-band in the NSST thereby reducing the complexity and increasing the computational speed. The proposed method is tested under different real-time disease datasets such as Glioma, mild Alzheimer's, and Encephalopathy with hypertension that includes similar pairs of images and analyzed different evaluation measures such as Entropy, standard deviation, structural similarity index measure,Mutual information, Average gradient, Xydeas and Petrovic metric, Peak-signal to-noise-ratio, processing time. The experimental findings and discussions indicate that the proposed algorithm outperforms other approaches and offers high quality fused images for an accurate diagnosis.

Fusion of convolution neural network, support vector machine and Sobel filter for accurate detection of COVID-19 patients using X-ray images.

Abstract: The coronavirus (COVID-19) is currently the most common contagious disease which is prevalent all over the world. The main challenge of this disease is the primary diagnosis to prevent secondary infections and its spread from one person to another. Therefore, it is essential to use an automatic diagnosis system along with clinical procedures for the rapid diagnosis of COVID-19 to prevent its spread. Artificial intelligence techniques using computed tomography (CT) images of the lungs and chest radiography have the potential to obtain high diagnostic performance for Covid-19 diagnosis. In this study, a fusion of convolutional neural network (CNN), support vector machine (SVM), and Sobel filter is proposed to detect COVID-19 using X-ray images. A new X-ray image dataset was collected and subjected to high pass filter using a Sobel filter to obtain the edges of the images. Then these images are fed to CNN deep learning model followed by SVM classifier with ten-fold cross validation strategy. This method is designed so that it can learn with not many data. Our results show that the proposed CNN-SVM with Sobel filter (CNN-SVM + Sobel) achieved the highest classification accuracy, sensitivity and specificity of 99.02%, 100% and 95.23%, respectively in automated detection of COVID-19. It showed that using Sobel filter can improve the performance of CNN. Unlike most of the other researches, this method does not use a pre-trained network. We have also validated our developed model using six public databases and obtained the highest performance. Hence, our developed model is ready for clinical application.

ECG arrhythmia classification by using a recurrence plot and convolutional neural network

Abstract: Cardiovascular diseases affect approximately 50 million people worldwide; thus, heart disease prevention is one of the most important tasks of any health care system. Despite the high popularity electrocardiography, superior automatic electrocardiography (ECG) signal analysis methods are required. The aim of this research was to design a new deep learning method for effectively classifying arrhythmia by using 2-second segments of 2D recurrence plot images of ECG signals. In the first stage, the noise and ventricular fibrillation (VF) categories were distinguished. In the second stage, the atrial fibrillation (AF), normal, premature AF, and premature VF categories were distinguished. Models were trained and tested using ECG databases publicly available at the website of PhysioNet. The MIT-BIH Arrhythmia Database, Creighton University Ventricular Tachyarrhythmia Database, MIT-BIH Atrial Fibrillation Database, and MIT-BIH Malignant Ventricular Ectopy Database were used to compare six types of arrhythmia. Testing accuracies of up to 95.3 % ± 1.27 % and 98.41 % ± 0.11 % were achieved for arrhythmia detection in the first and second stage, respectively, after five-fold cross-validation. In conclusion, this study provides clinicians with an advanced methodology for detecting and discriminating between different arrhythmia types.

MH-COVIDNet: Diagnosis of COVID-19 using Deep Neural Networks and Meta-heuristic-based Feature Selection on X-ray Images.

Abstract: COVID-19 is a disease that causes symptoms in the lungs and causes deaths around the world. Studies are ongoing for the diagnosis and treatment of this disease, which is defined as a pandemic. Early diagnosis of this disease is important for human life. This process is progressing rapidly with diagnostic studies based on deep learning. Therefore, to contribute to this field, a deep learning-based approach that can be used for early diagnosis of the disease is proposed in our study. In this approach, a data set consisting of 3 classes of COVID19, normal and pneumonia lung X-ray images was created, with each class containing 364 images. Pre-processing was performed using the image contrast enhancement algorithm on the prepared data set and a new data set was obtained. Feature extraction was completed from this data set with deep learning models such as AlexNet, VGG19, GoogleNet, and ResNet. For the selection of the best potential features, two metaheuristic algorithms of binary particle swarm optimization and binary gray wolf optimization were used. After combining the features obtained in the feature selection of the enhancement data set, they were classified using SVM. The overall accuracy of the proposed approach was obtained as 99.38%. The results obtained by verification with two different metaheuristic algorithms proved that the approach we propose can help experts during COVID-19 diagnostic studies.

DepHNN: A novel hybrid neural network for electroencephalogram (EEG)-based screening of depression

Abstract: Depression is a psychological disorder characterized by the continuous occurrence of bad mood state. It is critical to understand that this disorder is severely affecting people of multiple age groups across the world. This illness is now considered as a global issue and its early diagnosis will be effective in saving the lives of many people. This mental disorder can be diagnosed with the help of Electroencephalogram (EEG) signals as an analysis of these signals can indicate the prevailing mental state of the patients. This paper elaborates on the advantages of a fully automated Depression Detection System, as manual analysis of the EEG signal is very time consuming, tedious and it requires a lot of experience. This research paper presents a novel EEG based computer-aided (CAD) Hybrid Neural Network that can be identified as DepHNN (Depression Hybrid Neural Network) for depression screening. The proposed method uses Convolutional Neural Network (CNN) for temporal learning, windowing and long-short term memory (LSTM) architectures for the sequence learning process. In this model, EEG signals have been obtained from 21 drug-free, symptomatic depressed, and 24 normal patients using neuroscan. The model has less time and minimized computation complexity as it uses the windowing technique. It has attained an accuracy of 99.10% with mean absolute error (MAE) of 0.2040. The results show that the developed hybrid CNN-LSTM model is accurate, less complex, and useful in detecting depression using EEG signals.

A new transfer learning based approach to magnification dependent and independent classification of breast cancer in histopathological images

Abstract: The visual analysis of histopathological images is the gold standard for diagnosing breast cancer, yet a strenuous and an intricate task that requires years of pathologist training. Therefore, automating this task using computer-aided diagnosis (CAD) is highly expected. This paper proposes a new transfer learning-based approach to automated classification of breast cancer from histopathological images, including magnification dependent (MD) and magnification independent (MI) binary and eight-class classifications. We apply the deep neural network ResNet-18 to this problem, which is pre-trained on ImageNet, a large dataset of common images. We then design our transfer learning method to refine the network on histopathological images. Our transfer learning method is based on block-wise fine-tuning strategy; in which we make the last two residual blocks of the deep network model more domain-specific to our target data. It substantially helps to avoid over-fitting and speed up the training. Furthermore, we strengthen the adaptability of the proposed approach by using global contrast normalization (GCN) based on the target’s data values and three-fold data augmentation on training data. The experimental results of MD and MI binary and eight-class classifications on the publicly available BreaKHis dataset demonstrate that our approach is promising and effective, outperforming recent state-of-the-art MD and MI counterparts by a fair margin.

Real-time COVID-19 diagnosis from X-Ray images using deep CNN and extreme learning machines stabilized by chimp optimization algorithm

Abstract: Real-time detection of COVID-19 using radiological images has gained priority due to the increasing demand for fast diagnosis of COVID-19 cases. This paper introduces a novel two-phase approach for classifying chest X-ray images. Deep Learning (DL) methods fail to cover these aspects since training and fine-tuning the model's parameters consume much time. In this approach, the first phase comes to train a deep CNN working as a feature extractor, and the second phase comes to use Extreme Learning Machines (ELMs) for real-time detection. The main drawback of ELMs is to meet the need of a large number of hidden-layer nodes to gain a reliable and accurate detector in applying image processing since the detective performance remarkably depends on the setting of initial weights and biases. Therefore, this paper uses Chimp Optimization Algorithm (ChOA) to improve results and increase the reliability of the network while maintaining real-time capability. The designed detector is to be benchmarked on the COVID-Xray-5k and COVIDetectioNet datasets, and the results are verified by comparing it with the classic DCNN, Genetic Algorithm optimized ELM (GA-ELM), Cuckoo Search optimized ELM (CS-ELM), and Whale Optimization Algorithm optimized ELM (WOA-ELM). The proposed approach outperforms other comparative benchmarks with 98.25 % and 99.11 % as ultimate accuracy on the COVID-Xray-5k and COVIDetectioNet datasets, respectively, and it led relative error to reduce as the amount of 1.75 % and 1.01 % as compared to a convolutional CNN. More importantly, the time needed for training deep ChOA-ELM is only 0.9474 milliseconds, and the overall testing time for 3100 images is 2.937 s.

An improved method for soft tissue modeling

Abstract: The technique of force and haptic reappearance is an effective method to solve the shortage of haptic presence and improve the medical robots' practicability. Soft tissue models, the core of force-haptic reappearance systems, play a decisive role in its performance. The establishment of realistic soft tissue models can improve the system's authenticity and efficiency and better realize the representation of force and touch in the interaction process. At present, there exists a contradiction between timeliness and accuracy in soft tissue modeling. This paper combined the finite element method with the mass-spring model. We estimated the mass-spring model's parameters with the finite element method by neglecting the damping coefficient and obtained the relationship between the elastic coefficients. Then, according to the real measurement data of soft tissue in literature and the stress-strain curve obtained from real measurement, the values of a , A , e , k s σ were determined. Through those methods above, an improved soft tissue model was obtained. Through our comparison experiments, the improved spring particle model has a high degree of data fitting, and the force value under the same displacement is smaller. Moreover, the improved model's force-displacement curve in the large deformation stage is still very close to the measurement curve of the volume, which cannot be achieved by the empirical spring particle model. These comparative experiments show that the improved finite element-based particle spring model can better consider the timeliness and accuracy and simulate the soft tissue more accurately.

EEG-based emotion recognition using tunable Q wavelet transform and rotation forest ensemble classifier

Abstract: Emotion recognition by artificial intelligence (AI) is a challenging task. A wide variety of research has been done, which demonstrated the utility of audio, imagery, and electroencephalography (EEG) data for automatic emotion recognition. This paper presents a new automated emotion recognition framework, which utilizes electroencephalography (EEG) signals. The proposed method is lightweight, and it consists of four major phases, which include: a reprocessing phase, a feature extraction phase, a feature dimension reduction phase, and a classification phase. A discrete wavelet transforms (DWT) based noise reduction method, which is hereby named multi scale principal component analysis (MSPCA), is utilized during the pre-processing phase, where a Symlets-4 filter is utilized for noise reduction. A tunable Q wavelet transform (TQWT) is utilized as feature extractor. Six different statistical methods are used for dimension reduction. In the classification step, rotation forest ensemble (RFE) classifier is utilized with different classification algorithms such as k-Nearest Neighbor (k-NN), support vector machine (SVM), artificial neural network (ANN), random forest (RF), and four different types of the decision tree (DT) algorithms. The proposed framework achieves over 93 % classification accuracy with RFE + SVM. The results clearly show that the proposed TQWT and RFE based emotion recognition framework is an effective approach for emotion recognition using EEG signals.

Automatic classification of breast cancer histopathological images based on deep feature fusion and enhanced routing

Abstract: Automatic classification of breast cancer histopathological images is of great application value in breast cancer diagnosis. Convolutional neural network (CNN) usually highlights semantics, while capsule network (CapsNet) focuses on detailed information about the position and posture. Combining these information can obtain more discriminative features which is useful to improve the classification performance. In the paper, breast cancer histopathological image classification based on deep feature fusion and enhanced routing (FE-BkCapsNet) is proposed to take advantages of CNN and CapsNet. First, a novel structure with dual channels which can extract convolution features and capsule features simultaneously, integrate sematic features and spatial features into new capsules to obtain more discriminative information is designed. Then, routing coefficients are optimized indirectly and adaptively by modifying the loss function and embedding the routing process into entire optimization process. The proposed method FE-BkCapsNet was tested on a public dataset BreaKHis. Experimental results (40×: 92.71%, 100×: 94.52%, 200×: 94.03%, 400×: 93.54) demonstrate that the proposed method is efficient for breast cancer classification in clinical settings.

Deep learning models for cuffless blood pressure monitoring from PPG signals using attention mechanism

Abstract: Hypertension or high blood pressure is a major health problem worldwide and primary risk factor for cardiovascular disease. Blood pressure is one of the four vital signs that provides important information regarding patients’ cardiovascular system conditions. Continuous and regular blood pressure monitoring is essential for early diagnosis and prevention of cardiovascular disease. Considering the existing invasive or cuff-based blood pressuring monitoring techniques in clinical practice, several studies have identified motivation and advantages of a new non-invasive and cuffless blood pressuring measurement technique using Photoplethysmogram (PPG) signals. In this study, we propose several systolic and diastolic blood pressure estimation models using recurrent neural networks with bidirectional connections and attention mechanism utilising only PPG signals. The models were evaluated on PPG and blood pressure signals derived from the Multiparameter Intelligent Monitoring in Intensive Care II database. In the process, 22 characteristic features were extracted from the PPG waveform followed by various dimensionality reduction techniques to eliminate redundancies and reduce computational complexity. The proposed models were evaluated on both the 22-feature set and the reduced input feature vector, respectively. The models were compared with four machine learning techniques commonly used in the literature. Experimental results demonstrated that the proposed models could capture the non-linear relationship between the PPG features and blood pressure with high accuracy and outperformed the conventional machine learning methods on both datasets. The results for all the proposed models were acceptable by the global standards set by the Association for the Advancement of Medical Instrumentation for cuffless blood pressure estimation.

Schizophrenia detection technique using multivariate iterative filtering and multichannel EEG signals

Abstract: A new approach for extension of univariate iterative filtering (IF) for decomposing a signal into intrinsic mode functions (IMFs) or oscillatory modes is proposed for multivariate multi-component signals. Additionally the paper proposes a method to detect schizophrenia (Sz), based on analysing multi-channel electroencephalogram (EEG) signals. Using proposed multivariate iterative filtering (MIF), multi-channel EEG data are decomposed into multivariate IMFs (MIMFs). Depends on mean frequency, IMFs are grouped in order to separate EEG rhythms (delta, theta, alpha, beta, gamma) from EEG signals. The features, such as Hjorth parameters are extracted from EEG rhythms. Extracted features are ranked using student t -test and most discriminant 30 features are used for classification. Different classifier such as K-nearest neighbours (K-NN), linear discriminant analysis (LDA), support vector machine (SVM) with diffident kernels are considered to classify Sz and healthy EEG patterns. The proposed method is employed to evaluate 19-channel EEG signals recorded from 14 paranoid Sz patients and 14 healthy subjects. We have achieved highest accuracy of 98.9% using the SVM (Cubic) classifier. Sensitivity, specificity, positive predictive value (PPV), and area under ROC curve (AUC) of the same classifier are 99.0%, 98.8%, 98.4% and 0.999 respectively. Proposed approach for MIF is computationally efficient as compared to other multivariate signal decomposition algorithms. This paper presents a framework for decomposing multivariate signals efficiently and builds a model for detecting Sz accurately.

Hybrid deep neural network using transfer learning for EEG motor imagery decoding

Abstract: A major challenge in motor imagery (MI) of electroencephalogram (EEG) based brain–computer interfaces (BCIs) is the individual differences for different people. That the classification model should be retrained from scratch for a new subject often leads to unnecessary time consumption. In this paper, a “brain-ID” framework based on the hybrid deep neural network with transfer learning (HDNN-TL) is proposed to deal with individual differences of 4-class MI task. An end-to-end HDNN is developed to learn the common features of MI signal. HDNN consists of convolutional neural network (CNN) and Long Short-Term Memory (LSTM) which are utilized to decode the spatial and temporal features of the MI signal simultaneously. To deal with the EEG individual differences problem, transfer learning technique is implemented to fine-tune the followed fully connected (FC) layer to accommodate new subject with fewer training data. The classification performance on BCI competition IV dataset 2a by the proposed HDNN-TL in terms of kappa value is 0.8. We compared HDNN-TL, HDNN and other state-of-art methods and the experimental results demonstrate that the proposed method can get a satisfying result for new subjects with less time and fewer training data in MI task.

Towards effective classification of brain hemorrhagic and ischemic stroke using CNN

Abstract: Brain stroke is one of the most leading causes of worldwide death and requires proper medical treatment. Therefore, in this paper, our aim is to classify brain computed tomography (CT) scan images into hemorrhagic stroke, ischemic stroke and normal. Our newly proposed convolutional neural network (CNN) model utilizes image fusion and CNN approaches. Initially, some preprocessing operations have been employed by using multi-focus image fusion in order to improve the quality of CT images. Further, preprocessed images are fed into the newly proposed 13 layers CNN architecture for stroke classification. The robustness of our CNN method has been checked by conducting two experiments on two different datasets. In the first experiment, CT image dataset is partitioned into 20% testing and 80% training sets, while in the second experiment, 10 fold cross-validation of the image dataset has been performed. The classification accuracy obtained by our method on dataset 1 in the first experiment is 98.33% and in the second experiment, it is 98.77%, while in dataset 2 accuracy obtained in experiment 1 and 2 is 92.22% and 93.33% respectively. All the experiments have been conducted on the real CT image dataset which we have been collected from Himalayan Institute of Medical Sciences (HIMS), Dehradun, India. The results obtained by the proposed method have also been compared with AlexNet and ResNet50 where results show improvement over these CNN architectures.

Multi-class multi-label ophthalmological disease detection using transfer learning based convolutional neural network

Abstract: Fundus imaging is a retinal image modality for capturing anatomical structures and abnormalities in the human eye. Fundus images are the primary tool for observation and detection of a wide range of ophthalmological diseases. Changes in and around the anatomical structures like blood vessels, optic disc, fovea, and macula indicate the presence of disease like diabetic retinopathy, glaucoma, age-related macular degeneration (AMD), myopia, hypertension, and cataract. The patient may be suffering from more than one ophthalmological disease observed in either or both the eyes. Two models are proposed for multi-class multi-label fundus images classification of ophthalmological diseases using transfer learning based convolutional neural network (CNN) approaches. Ocular Disease Intelligent Recognition (ODIR) database having fundus images of left and right eye of patients for eight categories is used for experimentation. Four different pre-trained CNN architectures with two different optimizers are used and it is observed that VGG16 pre-trained architecture with SGD optimizer performs better for multi-class multi-label fundus images classification on ODIR database.

DCT & DWT based watermarking scheme for medical information security

Abstract: In order to secure the exchanged medical images in telemedicine, we propose in this work two blind watermarking approaches for the medical images protection. In the first scheme a combination of DCT and Schur decomposition is performed. In order to obtain a good compromise between robustness and imperceptibility, the integration is performed in the medium frequencies of the image. In the second scheme, the combination of DWT and Schur decomposition provide a more robust watermark distribution. Imperceptibility and robustness experimental results shows that the proposed methods maintain a high quality watermarked images and are very robust against several conventional attacks. These schemes allow the protection of the patient's information and thus ensure the confidentiality of personal data.

Automatic diagnosis of glaucoma using two-dimensional Fourier-Bessel series expansion based empirical wavelet transform

Abstract: Glaucoma is an eye disease in which fluid within the eye rises and puts pressure on optic nerves. This fluid pressure slowly damages the optic nerves, and if it is left untreated, it may lead to permanent vision loss. So the detection of glaucoma is necessary for on-time treatment. This paper presents a method, namely two dimensional Fourier-Bessel series expansion based empirical wavelet transform (2D-FBSE-EWT), which uses the Fourier-Bessel series expansion (FBSE) spectrum of order zero and order one for boundaries detection. 2D-FBSE-EWT method is also studied on multi-frequency scale during boundaries detection in FBSE spectrum. In multi-frequency scale based 2D-FBSE-EWT analysis, three frequency scales full, half, and quarter are used. These methods are used for the decomposition of fundus images into sub-images. For glaucoma detection from sub-images, two methods are used: (1) proposed method-1, which is a conventional machine learning (ML) based method and (2) proposed method-2, which is an ensemble ResNet-50 based method. The ensemble is done using operations like maxima, minima, averages, and fusion. Proposed method-1 has provided best result with order one 2D-FBSE-EWT at full scale. In Proposed method-2, order one 2D-FBSE-EWT at full scale with fusion ensemble method provides better accuracy as compared to other ensemble methods. Our proposed methods have outperformed all the compared methods used for glaucoma detection.

Electroencephalography-based motor imagery classification using temporal convolutional network fusion

Abstract: Motor imagery electroencephalography (MI-EEG) signals are generated when a person imagines a task without actually performing it. In recent studies, MI-EEG has been used in the rehabilitation process of paralyzed patients, therefore, decoding MI-EEG signals accurately is an important task, and it is difficult task due to the low signal-to-noise ratio and the variation of brain waves between subjects. Deep learning techniques such as the convolution neural network (CNN) have shown an impact in extracting meaningful features to improve the accuracy of classification. In this paper, we propose TCNet-Fusion, a fixed hyperparameter-based CNN model that utilizes multiple techniques, such as temporal convolutional networks (TCNs), separable convolution, depth-wise convolution, and the fusion of layers. This model outperforms other fixed hyperparameter-based CNN models while remaining similar to those that utilize variable hyperparameter networks, which are networks that change their hyperparameters based on each subject, resulting in higher accuracy than fixed networks. It also uses less memory than variable networks. The EEG signal undergoes two successive 1D convolutions, first along with the time domain, then channel-wise. Then, we obtain an image-like representation, which is fed to the main TCN. During experimentation, the model achieved a classification accuracy of 83.73 % on the four-class MI of the BCI Competition IV-2a dataset, and an accuracy of 94.41 % on the High Gamma Dataset.

An Attention-based Bi-LSTM Method for Visual Object Classification via EEG

Abstract: Background and Objective Despite many models have been proposed for brain visual perception and content understanding via electroencephalograms (EEGs), due to the lack of research on the inherent temporal relationship, EEG-based visual object classification still demands the improvement on its accuracy and computation complexity. Methods To take full advantage of the uneven visual feature saturation between time segments, an end-to-end attention-based Bi-LSTM Method is proposed, named Bi-LSTM-AttGW. Two attention strategies are introduced to Bi-LSTM framework. The attention gate replaces the forget gate in traditional LSTM. It is only relevant to the historical cell state, and not related to the current input. Hence, the attention gate can greatly reduce the number of training parameters. Moreover, the attention weighting method is applied to Bi-LSTM output, and it can explore the most decisive information. Results The best classification accuracy achieved by Bi-LSTM-AttGW model is 99.50%. Compared with the state-of-art algorithms and baseline models, the proposed method has great advantages in classification performance and computational complexity. Considering brain region level contribution on visual cognition task, we also verify our method using EEG signals collected from the frontal and occipital regions, that are highly correlated with visual perception tasks. Conclusions The results show promise towards the idea that human brain activity related to visual recognition can be more effectively decoded by neural networks with neural mechanism. The experimental results not only could provide strong support for the modularity theory about the brain cognitive function, but show the superiority of the proposed Bi-LSTM model with attention mechanism again.

A novel approach based on Three-scale image decomposition and Marine predators algorithm for multi-modal medical image fusion

Abstract: Multi-modal medical image fusion not only creates an image that preserves important information from the input images but also significantly improves in quality. This work contributes significantly to improving the ability of the physician to diagnose. So far, there have been many proposed approaches to improve efficiency for medical image fusion. However, some existing approaches still have certain drawbacks. The first drawback is that some vital information such as edges may be lost in the output image because of the high-frequency component fusion rules’ inefficiency. The second drawback is that the fused images often have low contrast because they have applied an average rule for the low-frequency components. In this study, a novel approach is introduced to overcome the aforementioned drawbacks, and it includes the following main steps. Firstly, the three-scale decomposition (TSD) technique is introduced to obtain the base and detail components. Secondly, a rule base on local energy function using the Kirsch compass operator is applied to fusing detail layers, which helps the output image preserve important information. Thirdly, the Marine predators algorithm (MPA) is utilized to fuse base layers by optimal parameters, allowing the output image to have good quality. To verify the proposed approach's effectiveness, we have utilized five state-of-the-art medical image fusion approaches and six image quality metrics for comparison. Experimental results show that the proposed approach significantly enhanced the fused image’s quality and preserved edge information.

Energy efficient ECG classification with spiking neural network

Abstract: Heart disease is one of the top ten threats to global health in 2019 according to the WHO. Continuous monitoring of ECG on wearable devices can detect abnormality in the user’s heartbeat early, thereby significantly increasing the chance of early intervention which is known to be the key to saving lives. In this paper, we present a set of inter-patient ECG classification methods that use convolutional (CNNs) and spiking neural networks (SNNs). We focused on inter-patient heartbeat classification, in which the model is trained over several patients and then used to infer that for patients not used in training. Raw heartbeat data is used in this paper because most wearable devices cannot deal with complex data preprocessing. A two-steps convolutional neural network testing method is proposed for saving power. For even greater energy-saving, a spiking neural network is also proposed. The latter is obtained from converting the trained CNN model with a less than one percent accuracy drop. The average power of a two-classes SNN is 0.077 W, or 0 . 0074 × that of previously proposed neural network-based solutions.

Multi-modal brain image fusion based on multi-level edge-preserving filtering

Abstract: Recently, multi-modal medical imaging technology and its collaborative diagnosis technology are developing rapidly. The application of medical image fusion technology in medical diagnosis becomes more important. In this paper, a multi-modal medical image fusion algorithm based on multi-level edge-preserving filtering (MLEPF) decomposition model is proposed. Firstly, an MLEPF model based on weighted mean curvature filtering is presented and used to decompose the multi-modal medical image into three types of layers: fine-structure (FS), coarse-structure (CS), and base (BS) layers. Secondly, a gradient domain pulse-coupled neural network (PCNN) fusion strategy is used to merge the FS and CS layers, and an energy attribute fusion strategy is used to merge the BS layers. Finally, the fused image is obtained by combining the three types of fused layers. The experiments are performed on six different disease datasets and one normal dataset, which contains more than 100 image pairs. Qualitative and quantitative evaluation testify that the proposed algorithm is superior to some excellent algorithms and can achieve close result to some state-of-the-art algorithms.

X-Ray and CT-Scan-Based Automated Detection and Classification of Covid-19 Using Convolutional Neural Networks (CNN).

Abstract: Covid-19 (Coronavirus Disease-2019) is the most recent coronavirus-related disease that has been announced as a pandemic by the World Health Organization (WHO). Furthermore, it has brought the whole planet to a halt as a result of the worldwide introduction of lockdown and killed millions of people. While this virus has a low fatality rate, the problem is that it is highly infectious, and as a result, it has infected a large number of people, putting a strain on the healthcare system, hence, Covid-19 identification in patients has become critical. The goal of this research is to use X-rays images and computed tomography(CT) images to introduce a deep learning strategy based on the Convolutional Neural Network (CNN) to automatically detect and identify the Covid-19 disease. We have implemented two different classifications using CNN(binary classification and multiclass classification). A total of 3,877 image datasets from CTs and X-rays were utilised to train the model in binary classification, with 1,917 images from Covid-19 infected individuals among them. The experimental results for binary classification show an overall accuracy of 99.64%, recall(or sensitivity) of 99.58%, the precision of 99.56%, F1-score of 99.59%, and ROC of 100%. For multiple classifications, the model was trained using a total of 6,077 images, with 1,917 images of Covid-19 infected people, 1,960 images of normal healthy people, and 2,200 images of pneumonia infected people. The experimental results for multiple classifications show an accuracy of 98.28%, recall(or sensitivity) of 98.25%, the precision of 98.22%, F1-score of 98.23%, and ROC of 99.87%. On the currently available dataset, the model produced the desired results, and it can assist healthcare workers in quickly detecting Covid-19 positive patients.

A new automated CNN deep learning approach for identification of ECG congestive heart failure and arrhythmia using constant-Q non-stationary Gabor transform

Abstract: Electrocardiogram (ECG) is an important noninvasive diagnostic method for interpretation and identification of various kinds of heart diseases. In this work, a new Deep Learning (DL) approach is proposed for automated identification of Congestive Heart Failure (CHF) and Arrhythmia (ARR) with high accuracy and low computational requirements. This study introduces, for the first time, a new ECG diagnosis algorithm that combines Convolutional Neural Network (CNN) with the Constant-Q Non-Stationary Gabor Transform (CQ-NSGT). The CQ-NSGT algorithm is investigated to transform the 1-D ECG signal into 2-D time-frequency representation that will be fed to a pre-trained CNN model, called AlexNet. Extracted features with the AlexNet architecture are used as relevant features to be discriminated by a Multi-Layer Perceptron (MLP) technique into three different cases, namely CHF, ARR, and Normal Sinus Rhythm (NSR). The performance of the proposed CNN with CQ-NSGT is compared versus CNN with Continuous Wavelet Transform (CWT), revealing the effectiveness of the CQ-NSGT algorithm. The proposed approach is examined with real ECG records, and the experimental results show the superior performance of the proposed approach over other existing techniques in terms of accuracy 98.82%, sensitivity 98.87%, specificity 99.21%, and precision 99.20%. This demonstrates the effectiveness of the proposed system in enhancing the ECG diagnosis accuracy.

A robust deep learning approach for automatic classification of seizures against non-seizures

Abstract: Identifying epileptic seizures through analysis of the electroencephalography (EEG) signal becomes a standard method for the diagnosis of epilepsy. Manual seizure identification on EEG by trained neurologists is time-consuming, labor-intensive and a reliable automatic seizure/non-seizure classification method is needed. One of the challenges in automatic seizure/non-seizure classification is that seizure morphologies exhibit considerable variabilities. In order to capture essential seizure patterns, this paper leverages an attention mechanism and a bidirectional long short-term memory (BiLSTM) to exploit both spatial and temporal discriminating features and overcome seizure variabilities. The attention mechanism captures spatial features according to the contributions of different brain regions to seizures. BiLSTM extracts discriminating temporal features in forward and backward directions. Cross-validation experiments and cross-patient experiments over the noisy data of CHB-MIT were performed. We obtained average sensitivity of 87.30%, specificity of 88.30% and precision of 88.29% in cross-validation experiments, higher than using the current state-of-the-art methods, and the standard deviations were lower. These results indicate that our approach performs well against current state-of-the-art methods and is more robust across patients.