Semi-supervised Stacked Label Consistent Autoencoder for Reconstruction and Analysis of Biomedical Signals

Semi-Supervised Autoencoder for Chemical Gas Classification with FTIR Spectrum

Chemical warfare agents pose a serious threat due to their extreme toxicity, necessitating swift the identification of chemical gases and individual responses to the identified threats. Fourier transform infrared (FTIR) spectroscopy offers a method for remote material analysis, particularly in detecting colorless and odorless chemical agents. In this paper, we propose a deep neural network utilizing a semi-supervised autoencoder (SSAE) for the classification of chemical gases based on FTIR spectra. In contrast to traditional methods, the SSAE concurrently trains an autoencoder and a classifier attached to a latent vector of the autoencoder, enhancing feature extraction for classification. The SSAE was evaluated on laboratory-collected FTIR spectra, demonstrating a superior classification performance compared to existing methods. The efficacy of the SSAE lies in its ability to generate denser cluster distributions in latent vectors, thereby enhancing gas classification. This study established a consistent experimental environment for hyperparameter optimization, offering valuable insights into the influence of latent vectors on classification performance.

The role of compression in large scale data transfer and storage of typical biomedical signals at hospitals

In modern hospitals, monitoring patients' vital signs and other biomedical signals is standard practice. With the advent of data-driven healthcare, Internet of medical things, wearable technologies, and machine learning, we expect this to accelerate and to be used in new and promising ways, including early warning systems and precision diagnostics. Hence, we see an ever-increasing need for retrieving, storing, and managing the large amount of biomedical signal data generated. The popularity of standards, such as HL7 FHIR for interoperability and data transfer, have also resulted in their use as a data storage model, which is inefficient. This article raises concern about the inefficiency of using FHIR for storage of biomedical signals and instead highlights the possibility of a sustainable storage based on data compression. Most reported efforts have focused on ECG signals; however, many other typical biomedical signals are understudied. In this article, we are considering arterial blood pressure, photoplethysmography, and respiration. We focus on simple lossless compression with low implementation complexity, low compression delay, and good compression ratios suitable for wide adoption. Our results show that it is easy to obtain a compression ratio of 2.7:1 for arterial blood pressure, 2.9:1 for photoplethysmography, and 4.1:1 for respiration.

Compressed sensing framework for BCG signals based on the optical fiber sensor

A compressed sensing (CS) framework is built for ballistocardiography (BCG) signals, which contains two parts of an optical fiber sensor-based heart monitoring system with a CS module and an end-to-end deep learning-based reconstruction algorithm. The heart monitoring system collects BCG data, and then compresses and transmits the data through the CS module at the sensing end. The deep learning-based algorithm reconstructs compressed data at the received end. To evaluate results, three traditional CS reconstruction algorithms and a deep learning method are adopted as references to reconstruct the compressed BCG data with different compression ratios (CRs). Results show that our framework can reconstruct signals successfully when the CR grows from 50% to 95% and outperforms other methods at high CRs. The mean absolute error (MAE) of the estimated heartbeat rate (HR) is lower than 1 bpm when the CR is below 95%. The proposed CS framework for BCG signals can be integrated into the IoMT system, which has great potential in health care for both medical and home use.

Multi-channel EEG emotion recognition through residual graph attention neural network

In this paper, a novel EEG emotion recognition method based on residual graph attention neural network is proposed. The method constructs a three-dimensional sparse feature matrix according to the relative position of electrode channels, and inputs it into the residual network to extract high-level abstract features containing electrode spatial position information. At the same time, the adjacency matrix representing the connection relationship of electrode channels is constructed, and the time-domain features of multi-channel EEG are modeled using graph. Then, the graph attention neural network is utilized to learn the intrinsic connection relationship between EEG channels located in different brain regions from the adjacency matrix and the constructed graph structure data. Finally, the high-level abstract features extracted from the two networks are fused to judge the emotional state. The experiment is carried out on DEAP data set. The experimental results show that the spatial domain information of electrode channels and the intrinsic connection relationship between different channels contain salient information related to emotional state, and the proposed model can effectively fuse these information to improve the performance of multi-channel EEG emotion recognition.

EEG seizure detection: concepts, techniques, challenges, and future trends

A central nervous system disorder is usually referred to as epilepsy. In epilepsy brain activity becomes abnormal, leading to times of abnormal behavior or seizures, and at times loss of awareness. Consequently, epilepsy patients face problems in daily life due to precautions they must take to adapt to this condition, particularly when they use heavy equipment, e.g., vehicle derivation. Epilepsy studies rely primarily on electroencephalography (EEG) signals to evaluate brain activity during seizures. It is troublesome and time-consuming to manually decide the location of seizures in EEG signals. The automatic detection framework is one of the principal tools to help doctors and patients take appropriate precautions. This paper reviews the epilepsy mentality disorder and the types of seizure, preprocessing operations that are performed on EEG data, a generally extracted feature from the signal, and a detailed view on classification procedures used in this problem and provide insights on the difficulties and future research directions in this innovative theme. Therefore, this paper presents a review of work on recent methods for the epileptic seizure process along with providing perspectives and concepts to researchers to present an automated EEG-based epileptic seizure detection system using IoT and machine learning classifiers for remote patient monitoring in the context of smart healthcare systems. Finally, challenges and open research points in EEG seizure detection are investigated.

Learning matrix factorization with scalable distance metric and regularizer

Matrix factorization has always been an encouraging field, which attempts to extract discriminative features from high-dimensional data. However, it suffers from negative generalization ability and high computational complexity when handling large-scale data. In this paper, we propose a learnable deep matrix factorization via the projected gradient descent method, which learns multi-layer low-rank factors from scalable metric distances and flexible regularizers. Accordingly, solving a constrained matrix factorization problem is equivalently transformed into training a neural network with an appropriate activation function induced from the projection onto a feasible set. Distinct from other neural networks, the proposed method activates the connected weights not just the hidden layers. As a result, it is proved that the proposed method can learn several existing well-known matrix factorizations, including singular value decomposition, convex, nonnegative and semi-nonnegative matrix factorizations. Finally, comprehensive experiments demonstrate the superiority of the proposed method against other state-of-the-arts.

ANet: Autoencoder-Based Local Field Potential Feature Extractor for Evaluating an Antidepressant Effect in Mice After Administering Kratom Leaf Extracts

Kratom (KT) typically exerts antidepressant (AD) effects. However, evaluating which form of KT extracts possesses AD properties similar to the standard AD fluoxetine (flu) remained challenging. Here, we adopted an autoencoder (AE)-based anomaly detector called ANet to measure the similarity of mice's local field potential (LFP) features that responded to KT leave extracts and AD flu. The features that responded to KT syrup had the highest similarity to those that responded to the AD flu at 87.11 ± 0.25%. This finding presents the higher feasibility of using KT syrup as an alternative substance for depressant therapy than KT alkaloids and KT aqueous, which are the other candidates in this study. Apart from the similarity measurement, we utilized ANet as a multi-task AE and evaluated the performance in discriminating multi-class LFP responses corresponding to the effect of different KT extracts and AD flu simultaneously. Furthermore, we visualized learned latent features among LFP responses qualitatively and quantitatively as t-SNE projection and maximum mean discrepancy distance, respectively. The classification results reported the accuracy and F1-score of 90.11 ± 0.11% and 90.08 ± 0.00%. In summary, the outcomes of this research might help therapeutic design devices for an alternative substance profile evaluation, such as Kratom-based form, in real-world applications.

Multi-Hierarchical Fusion to Capture the Latent Invariance for Calibration-Free Brain-Computer Interfaces

Brain-computer interfaces (BCI) based motor imagery (MI) has become a research hotspot for establishing a flexible communication channel for patients with apoplexy or degenerative pathologies. Accurate decoding of motor imagery electroencephalography (MI-EEG) signals, while essential for effective BCI systems, is still challenging due to the significant noise inherent in the EEG signals and the lack of informative correlation between the signals and brain activities. The application of deep learning for EEG feature representation has been rarely investigated, nevertheless bringing improvements to the performance of motor imagery classification. This paper proposes a deep learning decoding method based on multi-hierarchical representation fusion (MHRF) on MI-EEG. It consists of a concurrent framework constructed of bidirectional LSTM (Bi-LSTM) and convolutional neural network (CNN) to fully capture the contextual correlations of MI-EEG and the spectral feature. Also, the stacked sparse autoencoder (SSAE) is employed to concentrate these two domain features into a high-level representation for cross-session and subject training guidance. The experimental analysis demonstrated the efficacy and practicality of the proposed approach using a public dataset from BCI competition IV and a private one collected by our MI task. The proposed approach can serve as a robust and competitive method to improve inter-session and inter-subject transferability, adding anticipation and prospective thoughts to the practical implementation of a calibration-free BCI system.

Representative Task Self-Selection for Flexible Clustered Lifelong Learning

Consider the lifelong machine learning paradigm whose objective is to learn a sequence of tasks depending on previous experiences, e.g., knowledge library or deep network weights. However, the knowledge libraries or deep networks for most recent lifelong learning models are of prescribed size and can degenerate the performance for both learned tasks and coming ones when facing with a new task environment (cluster). To address this challenge, we propose a novel incremental clustered lifelong learning framework with two knowledge libraries: feature learning library and model knowledge library, called Flexible Clustered Lifelong Learning (FCL3). Specifically, the feature learning library modeled by an autoencoder architecture maintains a set of representation common across all the observed tasks, and the model knowledge library can be self-selected by identifying and adding new representative models (clusters). When a new task arrives, our FCL3 model firstly transfers knowledge from these libraries to encode the new task, i.e., effectively and selectively soft-assigning this new task to multiple representative models over feature learning library. Then: 1) the new task with a higher outlier probability will be judged as a new representative, and used to redefine both feature learning library and representative models over time; or 2) the new task with lower outlier probability will only refine the feature learning library. For model optimization, we cast this lifelong learning problem as an alternating direction minimization problem as a new task comes. Finally, we evaluate the proposed framework by analyzing several multitask data sets, and the experimental results demonstrate that our FCL3 model can achieve better performance than most lifelong learning frameworks, even batch clustered multitask learning models.

MIN2Net: End-to-End Multi-Task Learning for Subject-Independent Motor Imagery EEG Classification

OBJECTIVE

Advances in the motor imagery (MI)-based brain-computer interfaces (BCIs) allow control of several applications by decoding neurophysiological phenomena, which are usually recorded by electroencephalography (EEG) using a non-invasive technique. Despite significant advances in MI-based BCI, EEG rhythms are specific to a subject and various changes over time. These issues point to significant challenges to enhance the classification performance, especially in a subject-independent manner.

METHODS

To overcome these challenges, we propose MIN2Net, a novel end-to-end multi-task learning to tackle this task. We integrate deep metric learning into a multi-task autoencoder to learn a compact and discriminative latent representation from EEG and perform classification simultaneously.

RESULTS

This approach reduces the complexity in pre-processing, results in significant performance improvement on EEG classification. Experimental results in a subject-independent manner show that MIN2Net outperforms the state-of-the-art techniques, achieving an F1-score improvement of 6.72 %, and 2.23 % on the SMR-BCI, and OpenBMI datasets, respectively.

CONCLUSION

We demonstrate that MIN2Net improves discriminative information in the latent representation.

SIGNIFICANCE

This study indicates the possibility and practicality of using this model to develop MI-based BCI applications for new users without calibration.

Pacing Electrocardiogram Detection With Memory-Based Autoencoder and Metric Learning

Remote ECG diagnosis has been widely used in the clinical ECG workflow. Especially for patients with pacemaker, in the limited information of patient's medical history, doctors need to determine whether the patient is wearing a pacemaker and also diagnose other abnormalities. An automatic detection pacing ECG method can help cardiologists reduce the workload and the rates of misdiagnosis. In this paper, we propose a novel autoencoder framework that can detect the pacing ECG from the remote ECG. First, we design a memory module in the traditional autoencoder. The memory module is to record and query the typical features of the training pacing ECG type. The framework does not directly feed features of the encoder into the decoder but uses the features to retrieve the most relevant items in the memory module. In the training process, the memory items are updated to represent the latent features of the input pacing ECG. In the detection process, the reconstruction data of the decoder is obtained by the fusion features in the memory module. Therefore, the reconstructed data of the decoder tends to be close to the pacing ECG. Meanwhile, we introduce an objective function based on the idea of metric learning. In the context of pacing ECG detection, comparing the error of objective function of the input data and reconstructed data can be used as an indicator of detection. According to the objective function, if the input data does not belong to pacing ECG, the objective function may get a large error. Furthermore, we introduce a new database named the pacing ECG database including 800 patients with a total of 8,000 heartbeats. Experimental results demonstrate that our method achieves an average F1-score of 0.918. To further validate the generalization of the proposed method, we also experiment on a widely used MIT-BIH arrhythmia database.

Deep Learning for Reconstructing Low-Quality FTIR and Raman Spectra─A Case Study in Microplastic Analyses

Herein we report on a deep-learning method for the removal of instrumental noise and unwanted spectral artifacts in Fourier transform infrared (FTIR) or Raman spectra, especially in automated applications in which a large number of spectra have to be acquired within limited time. Automated batch workflows allowing only a few seconds per measurement, without the possibility of manually optimizing measurement parameters, often result in challenging and heterogeneous datasets. A prominent example of this problem is the automated spectroscopic measurement of particles in environmental samples regarding their content of microplastic (MP) particles. Effective spectral identification is hampered by low signal-to-noise ratios and baseline artifacts as, again, spectral post-processing and analysis must be performed in automated measurements, without adjusting specific parameters for each spectrum. We demonstrate the application of a simple autoencoding neural net for reconstruction of complex spectral distortions, such as high levels of noise, baseline bending, interferences, or distorted bands. Once trained on appropriate data, the network is able to remove all unwanted artifacts in a single pass without the need for tuning spectra-specific parameters and with high computational efficiency. Thus, it offers great potential for monitoring applications with a large number of spectra and limited analysis time with availability of representative data from already completed experiments.

An Attention Based CNN-LSTM Approach for Sleep-Wake Detection With Heterogeneous Sensors

In this article, we propose an attention based convolutional neural network long short-term memory (CNN-LSTM) approach for sleep-wake detection with heterogeneous sensor data, i.e., acceleration and heart rate variability (HRV). Since the three-dimensional acceleration data was sampled with a high frequency, we firstly design a CNN-LSTM structure to effectively learn latent features from the acceleration. Meanwhile, considering the unique format of the HRV data, some effective features are extracted based on domain knowledge. Next, we design a unified architecture to efficiently merge the features learned by CNN-LSTM approach from the acceleration and the extracted features from the HRV, which enables us to make full use of all the available information from these two heterogeneous sources. Taking into consideration that these two heterogeneous sources may have distinct contributions for the sleep and wake states, we propose an attention network to dynamically adjust the importance of features from the two sources. Real-world experiments have been conducted to verify the effectiveness of the proposed approach for sleep-wake detection. The results demonstrate that the proposed method outperforms all existing approaches for sleep-wake classification. In the evaluation of leave-one-subject-out (LOSO) cross-validation which is more challenging and practical, the proposed method achieves remarkable improvements ranging from 5% to 46% over the benchmark approaches.

A Hierarchical Discriminative Sparse Representation Classifier for EEG Signal Detection

Classification of electroencephalogram (EEG) signal data plays a vital role in epilepsy detection. Recently sparse representation-based classification (SRC) methods have achieved the good performance in EEG signal automatic detection, by which the EEG signals are sparsely represented using a few active coefficients in the dictionary and classified according to the reconstruction criteria. However, most of SRC learn a linear dictionary for encoding, and cannot extract enough information and nonlinear relationship of data for classification. To solve this problem, a hierarchical discriminative sparse representation classification model (called HD-SRC) for EEG signal detection is proposed. Based on the framework of neural network, HD-SRC learns the hierarchical nonlinear transformation and maps the signal data into the nonlinear transformed space. Through incorporating this idea into label consistent K singular value decomposition (LC-KSVD) at the top layer of neural network, HD-SRC seeks discriminative representation together with dictionary, while minimizing errors of classification, reconstruction and discriminative sparse-code for pattern classification. By learning the hierarchical feature mapping and discriminative dictionary simultaneously, more discriminative information of data can be exploited. In the experiment the proposed model is evaluated on the Bonn EEG database, and the results show it obtains satisfactory classification performance in multiple EEG signal detection tasks.

Epileptic Seizure Recognition Using Reduced Deep Convolutional Stack Autoencoder and Improved Kernel RVFLN From EEG Signals

In this paper, reduced deep convolutional stack autoencoder (RDCSAE) and improved kernel random vector functional link network (IKRVFLN) are combined to recognize the epileptic seizure using both the multichannel scalp and single-channel electroencephalogram (EEG) signals. The novel RDCSAE structure is designed to extract the most discriminative unsupervised features from EEG signals and fed into the proposed supervised IKRVFLN classifier to train efficiently by reducing the mean-square error cost function for recognizing the epileptic seizure activity with promising accuracy. The proposed RDCSAE-IKRVFLN algorithm is tested over the benchmark Boston Children's Hospital multichannel scalp EEG (sEEG) and Boon University, Germany single-channel EEG databases. The less computational complexity, higher learning speed, better model generalization, accurate epileptic seizure recognition, remarkable classification accuracy, negligible false positive rate per hour (FPR/h) and short event recognition time are the main advantages of the proposed RDCSAE-IKRVFLN method over reduced deep convolutional neural network (RDCNN), RDCSAE and RDCSAE-KRVFLN methods. The proposed RDCSAE-IKRVFLN method is implemented in a high-speed reconfigurable field-programmable gate array (FPGA) hardware environment to design a computer-aided-diagnosis (CAD) system for automatic epileptic seizure diagnosis. The simplicity, feasibility, and practicability of the proposed method validate the stable and reliable performances of epilepsy detection and recognition.

A Recent Investigation on Detection and Classification of Epileptic Seizure Techniques Using EEG Signal

The benefits of early detection and classification of epileptic seizures in analysis, monitoring and diagnosis for the realization and actualization of computer-aided devices and recent internet of medical things (IoMT) devices can never be overemphasized. The success of these applications largely depends on the accuracy of the detection and classification techniques employed. Several methods have been investigated, proposed and developed over the years. This paper investigates various seizure detection algorithms and classifications in the last decade, including conventional techniques and recent deep learning algorithms. It also discusses epileptiform detection as one of the steps towards advanced diagnoses of disorders of consciousness (DOCs) and their understanding. A performance comparison was carried out on the different algorithms investigated, and their advantages and disadvantages were explored. From our survey, much attention has recently been paid to exploring the efficacy of deep learning algorithms in seizure detection and classification, which are employed in other areas such as image processing and classification. Hybrid deep learning has also been explored, with CNN-RNN being the most popular.

A Deep Learning Approach for Automatic Seizure Detection in Children With Epilepsy

Over the last few decades, electroencephalogram (EEG) has become one of the most vital tools used by physicians to diagnose several neurological disorders of the human brain and, in particular, to detect seizures. Because of its peculiar nature, the consequent impact of epileptic seizures on the quality of life of patients made the precise diagnosis of epilepsy extremely essential. Therefore, this article proposes a novel deep-learning approach for detecting seizures in pediatric patients based on the classification of raw multichannel EEG signal recordings that are minimally pre-processed. The new approach takes advantage of the automatic feature learning capabilities of a two-dimensional deep convolution autoencoder (2D-DCAE) linked to a neural network-based classifier to form a unified system that is trained in a supervised way to achieve the best classification accuracy between the ictal and interictal brain state signals. For testing and evaluating our approach, two models were designed and assessed using three different EEG data segment lengths and a 10-fold cross-validation scheme. Based on five evaluation metrics, the best performing model was a supervised deep convolutional autoencoder (SDCAE) model that uses a bidirectional long short-term memory (Bi-LSTM) – based classifier, and EEG segment length of 4 s. Using the public dataset collected from the Children’s Hospital Boston (CHB) and the Massachusetts Institute of Technology (MIT), this model has obtained 98.79 ± 0.53% accuracy, 98.72 ± 0.77% sensitivity, 98.86 ± 0.53% specificity, 98.86 ± 0.53% precision, and an F1-score of 98.79 ± 0.53%, respectively. Based on these results, our new approach was able to present one of the most effective seizure detection methods compared to other existing state-of-the-art methods applied to the same dataset.

Multi-label Arrhythmia Classification from Fixed-length Compressed ECG Segments in Real-time Wearable ECG Monitoring

Recently, classification from compressed physiological signals in compressed sensing has been successfully applied to cardiovascular disease monitoring. However, in real-time wearable electrocardiogram (ECG) monitoring, it is very difficult to directly obtain the heartbeats information from compressed ECG signals. Thus arrhythmia classification from compressed ECG signals has to be handled in fixed-length segments instead of individual heartbeats. An inevitable issue is that a fixed-length ECG segment may contain multiple different types of arrhythmia. As a result, it is not appropriate to represent the multi-type real arrhythmia with a single label. In this paper, we first introduce multiple labels into fixed-length compressed ECG segments to challenge the arrhythmia classification issue. Then, we propose a deep learning model, which can directly classify multiple different types of arrhythmia from fixed-length compressed ECG segments with the advantages of low time cost for data processing and relatively high classification accuracy at a high compression ratio. Experimental results on the MIT-BIH arrhythmia database show that the exact match rate of our proposed method has reached 96.03% at CR(Compression Ratio)=70%, 94.99% at CR=80% and 93.19% at CR=90%.

Arrhythmias Classification Using Short-Time Fourier Transform and GAN Based Data Augmentation

Lacking sufficient training samples of different heart rhythms is a common bottleneck to obtain arrhythmias classification models with high accuracy using artificial neural networks. To solve this problem, we propose a novel data augmentation method based on short-time Fourier transform (STFT) and generative adversarial network (GAN) to obtain evenly distributed samples in the training dataset. Firstly, the one-dimensional electrocardiogram (ECG) signals with a fixed length of 6 s are subjected to STFT to obtain the coefficient matrices, and then the matrices of different heart rhythm samples are used to train GAN models respectively. The generated matrices are later employed to augment the training dataset of classification models based on four convolutional neural networks (CNNs). The result shows that the performances of the classification networks are all improved after we adopt the data enhancement strategy. The proposed method is helpful in augmentation and classification of biomedical signals, especially in detecting multiple arrhythmias, since adequate training samples are usually inaccessible in these studies.

Semi-supervised labelling of the femur in a whole-body post-mortem CT database using deep learning

A deep learning pipeline was developed and used to localize and classify a variety of implants in the femur contained in whole-body post-mortem computed tomography (PMCT) scans. The results provide a proof-of-principle approach for labelling content not described in medical/autopsy reports. The pipeline, which incorporated residual networks and an autoencoder, was trained and tested using n = 450 full-body PMCT scans. For the localization component, Dice scores of 0.99, 0.96, and 0.98 and mean absolute errors of 3.2, 7.1, and 4.2 mm were obtained in the axial, coronal, and sagittal views, respectively. A regression analysis found the orientation of the implant to the scanner axis and also the relative positioning of extremities to be statistically significant factors. For the classification component, test cases were properly labelled as nail (N⁺), hip replacement (H⁺), knee replacement (K⁺) or without-implant (I^-) with an accuracy >97%. The recall for I^- and H⁺ cases was 1.00, but fell to 0.82 and 0.65 for cases with K⁺ and N⁺. This semi-automatic approach provides a generalized structure for image-based labelling of features, without requiring time-consuming segmentation.

EEG-based image classification via a region-level stacked bi-directional deep learning framework

BACKGROUND

As a physiological signal, EEG data cannot be subjectively changed or hidden. Compared with other physiological signals, EEG signals are directly related to human cortical activities with excellent temporal resolution. After the rapid development of machine learning and artificial intelligence, the analysis and calculation of EEGs has made great progress, leading to a significant boost in performances for content understanding and pattern recognition of brain activities across the areas of both neural science and computer vision. While such an enormous advance has attracted wide range of interests among relevant research communities, EEG-based classification of brain activities evoked by images still demands efforts for further improvement with respect to its accuracy, generalization, and interpretation, yet some characters of human brains have been relatively unexplored.

METHODS

We propose a region-level stacked bi-directional deep learning framework for EEG-based image classification. Inspired by the hemispheric lateralization of human brains, we propose to extract additional information at regional level to strengthen and emphasize the differences between two hemispheres. The stacked bi-directional long short-term memories are used to capture the dynamic correlations hidden from both the past and the future to the current state in EEG sequences.

RESULTS

Extensive experiments are carried out and our results demonstrate the effectiveness of our proposed framework. Compared with the existing state-of-the-arts, our framework achieves outstanding performances in EEG-based classification of brain activities evoked by images. In addition, we find that the signals of Gamma band are not only useful for achieving good performances for EEG-based image classification, but also play a significant role in capturing relationships between the neural activations and the specific emotional states.

CONCLUSIONS

Our proposed framework provides an improved solution for the problem that, given an image used to stimulate brain activities, we should be able to identify which class the stimuli image comes from by analyzing the EEG signals. The region-level information is extracted to preserve and emphasize the hemispheric lateralization for neural functions or cognitive processes of human brains. Further, stacked bi-directional LSTMs are used to capture the dynamic correlations hidden in EEG data. Extensive experiments on standard EEG-based image classification dataset validate that our framework outperforms the existing state-of-the-arts under various contexts and experimental setups.

Energy-Efficient Intelligent ECG Monitoring for Wearable Devices

Wearable intelligent ECG monitoring devices can perform automatic ECG diagnosis in real time and send out alert signal together with abnormal ECG signal for doctor's further analysis. This provides a means for the patient to identify their heart problem as early as possible and go to doctors for medical treatment. For such system the key requirements include high accuracy and low power consumption. However, the existing wearable intelligent ECG monitoring schemes suffer from high power consumption in both ECG diagnosis and transmission in order to achieve high accuracy. In this work, we have proposed an energy-efficient wearable intelligent ECG monitor scheme with two-stage end-to-end neural network and diagnosis-based adaptive compression. Compared to the state-of-the-art schemes, it significantly reduces the power consumption in ECG diagnosis and transmission while maintaining high accuracy.

Cellular frustration algorithms for anomaly detection applications

Cellular frustrated models have been developed to describe how the adaptive immune system works. They are composed by independent agents that continuously pair and unpair depending on the information that one sub-set of these agents display. The emergent dynamics is sensitive to changes in the displayed information and can be used to detect anomalies, which can be important to accomplish the immune system main function of protecting the host. Therefore, it has been hypothesized that these models could be adequate to model the immune system activation. Likewise it has been hypothesized that these models could provide inspiration to develop new artificial intelligence algorithms for data mining applications. However, computational algorithms do not need to follow strictly the immunological reality. Here, we investigate efficient implementation strategies of these immune inspired ideas for anomaly detection applications and use real data to compare the performance of cellular frustration algorithms with standard implementations of one-class support vector machines and deep autoencoders. Our results demonstrate that more efficient implementations of cellular frustration algorithms are possible and also that cellular frustration algorithms can be advantageous for semi-supervised anomaly detection applications given their robustness and accuracy.

A Real-Time Arrhythmia Heartbeats Classification Algorithm Using Parallel Delta Modulations and Rotated Linear-Kernel Support Vector Machines

Real-time wearable electrocardiogram monitoring sensor is one of the best candidates in assisting cardiovascular disease diagnosis. In this paper, we present a novel real-time machine learning system for Arrhythmia classification. The system is based on the parallel Delta modulation and QRS/PT wave detection algorithms. We propose a patient dependent rotated linear-kernel support vector machine classifier that combines the global and local classifiers, with three types of feature vectors extracted directly from the Delta modulated bit-streams. The performance of the proposed system is evaluated using the MIT-BIH Arrhythmia database. According to the AAMI standard, two binary classifications are performed and evaluated, which are supraventricular ectopic beat (SVEB) versus the rest four classes, and ventricular ectopic beat (VEB) versus the rest. For SVEB classification, the preferred SkP-32 method's F1 score, sensitivity, specificity, and positive predictivity value are 0.83, 79.3%, 99.6%, and 88.2%, respectively, and for VEB classification, the numbers are 0.92%, 92.8%, 99.4%, and 91.6%, respectively. The results show that the performance of our proposed approach is comparable to that of published research. The proposed low-complexity algorithm has the potential to be implemented as an on-sensor machine learning solution.

A Novel Electrocardiogram Biometric Identification Method Based on Temporal-Frequency Autoencoding

For good performance, most existing electrocardiogram (ECG) identification methods still need to adopt a denoising process to remove noise interference beforehand. This specific signal preprocessing technique requires great efforts for algorithm engineering and is usually complicated and time-consuming. To more conveniently remove the influence of noise interference and realize accurate identification, a novel temporal-frequency autoencoding based method is proposed. In particular, the raw data is firstly transformed into the wavelet domain, where multi-level time-frequency representation is achieved. Then, a prior knowledge-based feature selection is proposed and applied to the transformed data to discard noise components and retain identity-related information simultaneously. Afterward, the stacked sparse autoencoder is introduced to learn intrinsic discriminative features from the selected data, and Softmax classifier is used to perform the identification task. The effectiveness of the proposed method is evaluated on two public databases, namely, ECG-ID and Massachusetts Institute of Technology-Biotechnology arrhythmia (MIT-BIH-AHA) databases. Experimental results show that our method can achieve high multiple-heartbeat identification accuracies of 98.87%, 92.3%, and 96.82% on raw ECG signals which are from the ECG-ID (Two-recording), ECG-ID (All-recording), and MIT-BIH-AHA database, respectively, indicating that our method can provide an efficient way for ECG biometric identification.

SAE+LSTM: A New Framework for Emotion Recognition From Multi-Channel EEG

EEG-based automatic emotion recognition can help brain-inspired robots in improving their interactions with humans. This paper presents a novel framework for emotion recognition using multi-channel electroencephalogram (EEG). The framework consists of a linear EEG mixing model and an emotion timing model. Our proposed framework considerably decomposes the EEG source signals from the collected EEG signals and improves classification accuracy by using the context correlations of the EEG feature sequences. Specially, Stack AutoEncoder (SAE) is used to build and solve the linear EEG mixing model and the emotion timing model is based on the Long Short-Term Memory Recurrent Neural Network (LSTM-RNN). The framework was implemented on the DEAP dataset for an emotion recognition experiment, where the mean accuracy of emotion recognition achieved 81.10% in valence and 74.38% in arousal, and the effectiveness of our framework was verified. Our framework exhibited a better performance in emotion recognition using multi-channel EEG than the compared conventional approaches in the experiments.

Serial electrocardiography to detect newly emerging or aggravating cardiac pathology: a deep-learning approach

BACKGROUND

Serial electrocardiography aims to contribute to electrocardiogram (ECG) diagnosis by comparing the ECG under consideration with a previously made ECG in the same individual. Here, we present a novel algorithm to construct dedicated deep-learning neural networks (NNs) that are specialized in detecting newly emerging or aggravating existing cardiac pathology in serial ECGs.

METHODS

We developed a novel deep-learning method for serial ECG analysis and tested its performance in detection of heart failure in post-infarction patients, and in the detection of ischemia in patients who underwent elective percutaneous coronary intervention. Core of the method is the repeated structuring and learning procedure that, when fed with 13 serial ECG difference features (intra-individual differences in: QRS duration; QT interval; QRS maximum; T-wave maximum; QRS integral; T-wave integral; QRS complexity; T-wave complexity; ventricular gradient; QRS-T spatial angle; heart rate; J-point amplitude; and T-wave symmetry), dynamically creates a NN of at most three hidden layers. An optimization process reduces the possibility of obtaining an inefficient NN due to adverse initialization.

RESULTS

Application of our method to the two clinical ECG databases yielded 3-layer NN architectures, both showing high testing performances (areas under the receiver operating curves were 84% and 83%, respectively).

CONCLUSIONS

Our method was successful in two different clinical serial ECG applications. Further studies will investigate if other problem-specific NNs can successfully be constructed, and even if it will be possible to construct a universal NN to detect any pathologic ECG change.

An autoencoder based formulation for compressed sensing reconstruction

This work proposes a new formulation for image reconstruction based on the autoencoder framework. The work follows the adaptive approach used in prior dictionary and transform learning based reconstruction techniques. Existing autoencoder based reconstructions are non-adaptive; they are trained on a separate training set and applied on another. In this work, the autoencoder is learnt from the patches of the image it is reconstructing. Experimental studies on MRI reconstruction shows that the proposed method outperforms state-of-the-art methods in dictionary learning, transform learning and (non-adaptive) autoencoder based approaches.

Stacked sparse autoencoder networks and statistical shape models for automatic staging of distal femur trochlear dysplasia

BACKGROUND

The quantitative morphological analysis of the trochlear region in the distal femur and the precise staging of the potential dysplastic condition constitute a key point for the use of personalized treatment options for the patella-femoral joint. In this paper, we integrated statistical shape models (SSM), able to represent the individual morphology of the trochlea by means of a set of parameters and stacked sparse autoencoder (SSPA) networks, which exploit the parameters to discriminate among different levels of abnormalities.

METHODS

Two datasets of distal femur reconstructions were obtained from CT scans, including pathologic and physiologic shapes. Both of them were processed to compute SSM of healthy and dysplastic trochlear regions. The parameters obtained by the 3D-3D reconstruction of a femur shape were fed into a trained SSPA classifier to automatically establish the membership to one of three clinical conditions, namely, healthy, mild dysplasia, and severe dysplasia of the trochlea. The validation was performed on a subset of the shapes not used in the construction of the SSM, by verifying the occurrence of a correct classification.

RESULTS

A major finding of the work is that SSM are able to represent anomalies of the trochlear geometry by means of specific eigenmodes of variation and to model the interplay between morphologic features related to dysplasia. Exploiting the patient-specific morphing parameters of SSM, computed by means of a 3D-3D reconstruction, SSPA is demonstrated to outperform traditional discriminant analysis in classifying healthy, mild, and severe trochlear dysplasia providing 99%, 97%, and 98% accuracy for each of the three classes, respectively (discriminant analysis accuracy: 85%, 89%, and 77%).

CONCLUSIONS

From a clinical point of view, this paper contributes to support the increasing role of SSM, integrated with deep learning techniques, in diagnostics and therapy definition as quantitative and advanced visualization tools.

Deep Learning on 1-D Biosignals: a Taxonomy-based Survey

OBJECTIVES

 Deep learning models such as convolutional neural networks (CNNs) have been applied successfully to medical imaging, but biomedical signal analysis has yet to fully benefit from this novel approach. Our survey aims at (i) reviewing deep learning techniques for biosignal analysis in computer- aided diagnosis; and (ii) deriving a taxonomy for organizing the growing number of applications in the field.

METHODS

 A comprehensive literature research was performed using PubMed, Scopus, and ACM. Deep learning models were classified with respect to the (i) origin, (ii) dimension, and (iii) type of the biosignal as input to the deep learning model; (iv) the goal of the application; (v) the size and (vi) type of ground truth data; (vii) the type and (viii) schedule of learning the network; and (ix) the topology of the model.

RESULTS

 Between January 2010 and December 2017, a total 71 papers were published on the topic. The majority (n = 36) of papers are on electrocariography (ECG) signals. Most applications (n = 25) aim at detection of patterns, while only a few (n = 6) at predection of events. Out of 36 ECG-based works, many (n = 17) relate to multi-lead ECG. Other biosignals that have been identified in the survey are electromyography, phonocardiography, photoplethysmography, electrooculography, continuous glucose monitoring, acoustic respiratory signal, blood pressure, and electrodermal activity signal, while ballistocardiography or seismocardiography have yet to be analyzed using deep learning techniques. In supervised and unsupervised applications, CNNs and restricted Boltzmann machines are the most and least frequently used, (n = 34) and (n = 15), respectively.

CONCLUSION

 Our key-code classification of relevant papers was used to cluster the approaches that have been published to date and demonstrated a large variability of research with respect to data, application, and network topology. Future research is expected to focus on the standardization of deep learning architectures and on the optimization of the network parameters to increase performance and robustness. Furthermore, application-driven approaches and updated training data from mobile recordings are needed.

Graph structured autoencoder

In this work, we introduce the graph regularized autoencoder. We propose three variants. The first one is the unsupervised version. The second one is tailored for clustering, by incorporating subspace clustering terms into the autoencoder formulation. The third is a supervised label consistent autoencoder suitable for single label and multi-label classification problems. Each of these has been compared with the state-of-the-art on benchmark datasets. The problems addressed here are image denoising, clustering and classification. Our proposed methods excel of the existing techniques in all of the problems.

Resolution enhancement of wide-field interferometric microscopy by coupled deep autoencoders

Wide-field interferometric microscopy is a highly sensitive, label-free, and low-cost biosensing imaging technique capable of visualizing individual biological nanoparticles such as viral pathogens and exosomes. However, further resolution enhancement is necessary to increase detection and classification accuracy of subdiffraction-limited nanoparticles. In this study, we propose a deep-learning approach, based on coupled deep autoencoders, to improve resolution of images of L-shaped nanostructures. During training, our method utilizes microscope image patches and their corresponding manual truth image patches in order to learn the transformation between them. Following training, the designed network reconstructs denoised and resolution-enhanced image patches for unseen input.