Deep Neural Networks and Transfer Learning on a Multivariate Physiological Signal Dataset

Physiological Noise Filtering in Functional Near-Infrared Spectroscopy Signals Using Wavelet Transform and Long-Short Term Memory Networks

Activated channels of functional near-infrared spectroscopy are typically identified using the desired hemodynamic response function (dHRF) generated by a trial period. However, this approach is not possible for an unknown trial period. In this paper, an innovative method not using the dHRF is proposed, which extracts fluctuating signals during the resting state using maximal overlap discrete wavelet transform, identifies low-frequency wavelets corresponding to physiological noise, trains them using long-short term memory networks, and predicts/subtracts them during the task session. The motivation for prediction is to maintain the phase information of physiological noise at the start time of a task, which is possible because the signal is extended from the resting state to the task session. This technique decomposes the resting state data into nine wavelets and uses the fifth to ninth wavelets for learning and prediction. In the eighth wavelet, the prediction error difference between the with and without dHRF from the 15-s prediction window appeared to be the largest. Considering the difficulty in removing physiological noise when the activation period is near the physiological noise, the proposed method can be an alternative solution when the conventional method is not applicable. In passive brain-computer interfaces, estimating the brain signal starting time is necessary.

Detection of Corneal Ulcer Using a Genetic Algorithm-Based Image Selection and Residual Neural Network

Corneal ulcer is one of the most devastating eye diseases causing permanent damage. There exist limited soft techniques available for detecting this disease. In recent years, deep neural networks (DNN) have significantly solved numerous classification problems. However, many samples are needed to obtain reasonable classification performance using a DNN with a huge amount of layers and weights. Since collecting a data set with a large number of samples is usually a difficult and time-consuming process, very large-scale pre-trained DNNs, such as the AlexNet, the ResNet and the DenseNet, can be adapted to classify a dataset with a small number of samples, through the utility of transfer learning techniques. Although such pre-trained DNNs produce successful results in some cases, their classification performances can be low due to many parameters, weights and the emergence of redundancy features that repeat themselves in many layers in som cases. The proposed technique removes these unnecessary features by systematically selecting images in the layers using a genetic algorithm (GA). The proposed method has been tested on ResNet on a small-scale dataset which classifies corneal ulcers. According to the results, the proposed method significantly increased the classification performance compared to the classical approaches.

Artificial Intelligence for Personalized Genetics and New Drug Development: Benefits and Cautions

As the global health care system grapples with steadily rising costs, increasing numbers of admissions, and the chronic defection of doctors and nurses from the profession, appropriate measures need to be put in place to reverse this course before it is too late [...].

Advances in Multivariate and Multiscale Physiological Signal Analysis

Physiological systems are characterized by complex dynamics and nonlinear behaviors due to their intricate structural organization and regulatory mechanisms [...].

A Short Survey on Deep Learning for Multimodal Integration: Applications, Future Perspectives and Challenges

Deep learning has achieved state-of-the-art performances in several research applications nowadays: from computer vision to bioinformatics, from object detection to image generation. In the context of such newly developed deep-learning approaches, we can define the concept of multimodality. The objective of this research field is to implement methodologies which can use several modalities as input features to perform predictions. In this, there is a strong analogy with respect to what happens with human cognition, since we rely on several different senses to make decisions. In this article, we present a short survey on multimodal integration using deep-learning methods. In a first instance, we comprehensively review the concept of multimodality, describing it from a two-dimensional perspective. First, we provide, in fact, a taxonomical description of the multimodality concept. Secondly, we define the second multimodality dimension as the one describing the fusion approaches in multimodal deep learning. Eventually, we describe four applications of multimodal deep learning to the following fields of research: speech recognition, sentiment analysis, forensic applications and image processing.

A Survey on Physiological Signal-Based Emotion Recognition

Physiological signals are the most reliable form of signals for emotion recognition, as they cannot be controlled deliberately by the subject. Existing review papers on emotion recognition based on physiological signals surveyed only the regular steps involved in the workflow of emotion recognition such as pre-processing, feature extraction, and classification. While these are important steps, such steps are required for any signal processing application. Emotion recognition poses its own set of challenges that are very important to address for a robust system. Thus, to bridge the gap in the existing literature, in this paper, we review the effect of inter-subject data variance on emotion recognition, important data annotation techniques for emotion recognition and their comparison, data pre-processing techniques for each physiological signal, data splitting techniques for improving the generalization of emotion recognition models and different multimodal fusion techniques and their comparison. Finally, we discuss key challenges and future directions in this field.

A Machine Learning Perspective on fNIRS Signal Quality Control Approaches

Despite a rise in the use of functional Near Infra-Red Spectroscopy (fNIRS) to study neural systems, fNIRS signal processing is not standardized and is highly affected by empirical and manual procedures. At the beginning of any signal processing procedure, Signal Quality Control (SQC) is critical to prevent errors and unreliable results. In fNIRS analysis, SQC currently relies on applying empirical thresholds to handcrafted Signal Quality Indicators (SQIs). In this study, we use a dataset of fNIRS signals (N = 1,340) recorded from 67 subjects, and manually label the signal quality of a subset of segments (N = 548) to investigate the pitfalls of current practices while exploring the opportunities provided by Deep Learning approaches. We show that SQIs statistically discriminate signals with bad quality, but the identification by means of empirical thresholds lacks sensitivity. Alternatively to manual thresholding, conventional machine learning models based on the SQIs have been proven more accurate, with end-to-end approaches, based on Convolutional Neural Networks, capable of further improving the performance. The proposed approach, based on machine learning, represents a more objective SQC for fNIRS and moves towards the use of fully automated and standardized procedures.

Current trends and opportunities in the methodology of electrodermal activity measurement

Electrodermal activity (EDA) has been measured in the laboratory since the late 1800s. Although the influence of sudomotor nerve activity and the sympathetic nervous system on EDA is well established, the mechanisms underlying EDA signal generation are not completely understood. Owing to simplicity of instrumentation and modern electronics, these measurements have recently seen a transfer from the laboratory to wearable devices, sparking numerous novel applications while bringing along both challenges and new opportunities. In addition to developments in electronics and miniaturization, current trends in material technology and manufacturing have sparked innovations in electrode technologies, and trends in data science such as machine learning and sensor fusion are expanding the ways that measurement data can be processed and utilized. Although challenges remain for the quality of wearable EDA measurement, ongoing research and developments may shorten the quality gap between wearable EDA and standardized recordings in the laboratory. In this topical review, we provide an overview of the basics of EDA measurement, discuss the challenges and opportunities of wearable EDA, and review recent developments in instrumentation, material technology, signal processing, modeling and data science tools that may advance the field of EDA research and applications over the coming years.

Improving the Efficacy of Deep-Learning Models for Heart Beat Detection on Heterogeneous Datasets

Deep learning (DL) has greatly contributed to bioelectric signal processing, in particular to extract physiological markers. However, the efficacy and applicability of the results proposed in the literature is often constrained to the population represented by the data used to train the models. In this study, we investigate the issues related to applying a DL model on heterogeneous datasets. In particular, by focusing on heart beat detection from electrocardiogram signals (ECG), we show that the performance of a model trained on data from healthy subjects decreases when applied to patients with cardiac conditions and to signals collected with different devices. We then evaluate the use of transfer learning (TL) to adapt the model to the different datasets. In particular, we show that the classification performance is improved, even with datasets with a small sample size. These results suggest that a greater effort should be made towards the generalizability of DL models applied on bioelectric signals, in particular, by retrieving more representative datasets.

Acquisition and Processing of Brain Signals

We live within a context of unprecedented opportunities for brain research, with a flourishing of novel sensing technologies and methodological approaches [...].