Signal Folding for Efficient Classification of Near-Cyclostationary Biological Signals

The classification of biological signals is important in detecting abnormal conditions in observed biological subjects. The classifiers are trained on feature vectors, which often constitute the parameters of the observed time series data models. Since the feature extraction is usually the most time-consuming step in training a classifier, in this paper, signal folding and the associated folding operator are introduced to reduce the variability in near-cyclostationary biological signals so that these signals can be represented by models that have a lower order. This leads to a substantial reduction in computational complexity, so the classifier can be learned an order of magnitude faster and still maintain its decision accuracy. The performance of different classifiers involving signal folding as a pre-processing step is studied for sleep apnea detection in one-lead ECG signals assuming ARIMA modeling of the time series data. It is shown that the R-peak-based folding of ECG segments has superior performance to other more general, similarity based signal folding methods. The folding order can be optimized for the best classification accuracy. However, signal folding requires precise scaling and alignment of the created signal fragments.

834 A robust deep learning approach for precisely segmenting cells in multiplex tissue images

BackgroundMultiplex images are becoming pivotal in tissue pathology because they provide positional location and multidimensional phenotype of every cell. The heterogeneity of cells, morphologies, and densities makes the identification of the millions of cells in a tissue slice challenging. There is an urgent need for a robust, yet flexible, algorithm to automatically demarcate each cell that accurately defines cellular boundaries. We have developed a method to extend a DL nuclear identification algorithm beyond the nucleus and to the outer boundary of the cell using biological signals from multiplex panels.MethodsAll image analysis was performed in the Visiopharm image analysis platform. Three human observers provided ground truth (GT) annotations by outlining cells in predefined areas each containing ~30 cells in six different images from two different multiplex instruments: mIF = 8-plex via Vectra Polaris from Akoya and IMC = 13-plex via Hyperion from Fluidigm. Images were subsequently segmented by different AI methods: Machine Learning Nuclear Detection (ML), Deep Learning Nuclear Detection (DL), and DL that incorporates biological signals (DL+). Each set of computer-generated annotations was compared to GT using common evaluation metrics DICE, Precision and Sensitivity.ResultsOverall, we found a high degree of concordance between the computer-generated and human annotations (DICE = 0.73±0.08, n=12) and between imaging modalities (mIF: 0.76±0.07; IMC: 0.71±0.08; n=6). Comparison of DICE scores for the AI methods indicated a superior delineation of cell boundaries using the DL+ method (DL+: 0.79±0.07; ML: 0.74±0.08; DL: 0.74±0.03;). Precision, which compares true vs false positive annotated regions to GT, was also high for all images (0.77±0.11) (mIF: 0.76±0.10; IMC: 0.78±0.11). Sensitivity, which compares true positives vs false negative annotated regions GT, was also high for all images (0.77±0.09) (mIF: 0.76±0.09; IMC: 0.79±0.09).ConclusionsWe developed a flexible DL based strategy that enables the most comprehensive segmentation of cells in multiplex tissue images. Each AI approach shows a high concordance with segmentation annotations from human observers as measured by the industry standards DICE, Precision and Sensitivity. The DL+ method did achieve the highest DICE score indicating a more accurate delineation of cell boundaries. Expectedly, precision and sensitivity metrics are similar between all methods while DICE Coefficient better accounts for the annotations at the cell edge. The DL+ cell segmentation algorithm will yield an improved accuracy when phenotyping cells in downstream analysis as the precise biomarker composition is more accurately contained within each cell.

The Influence of Brain Activity on the Interactive Process through Biofeedback Mechanisms in Virtual Reality Environments

This work focuses on the development of a software link interface tool between the Looxid Link Device coupled to the HTC Vive Pro VR HeadSets and the Unity platform, to generate real-time interactivity in virtual reality applications. The software incorporates a dynamic and parameterizable algorithm to be used as a core-engine in the real-time Biofeedback process, recognizing the values of the biological signals registered in each of the EEG channels of the Looxid Link device. The values of EEG frequencies detected in real time can be used to generate elements of interactivity, with different frequencies and intensities.

Improving Driving Safety by Detecting Negative Emotions with Biological Signals: Which Is the Best?

There is ample evidence confirming the adverse effects of negative emotions such as anger, fear, and anxiety on drivers’ performance. Also, effectiveness of biological signals in emotion recognition has been confirmed. Therefore, developing advanced driver-assistance systems based on biological signals to detect negative emotions can play a major role in improving driving safety. However, since recording signals, data analysis, as well as design and implementation of a system based on one or more biological signals take time and are costly, it is necessary to conduct appropriate preliminary studies on the efficiency of these signals in identifying negative emotions. The purpose of this study was to explore the efficiency of four biological signals including electrocardiogram (ECG), electromyogram (EMG), electrodermal activity (EDA), and electroencephalogram (EEG) in detecting negative emotions while driving. To this end, a series of scenarios were designed to arouse negative emotions in the driving simulator environment. A total of 43 individuals participated in the experiments, during which the four signals were recorded. Next, we extracted 58 features from the collected data for analysis. Then, multi-layer perceptron and radial basis function neural networks were implemented using the features of each of these signals separately. Afterward, the four evaluation criteria of accuracy, sensitivity, specificity, and precision were calculated for the signals. Finally, TOPSIS was used to rank the signals. ECG and EDA signals, with 88% and 90% accuracy, respectively, were found to be the best signals in detecting negative emotions during driving.

The Implementation and Evaluation of Individual Preference in Robot Facial Expression Based on Emotion Estimation Using Biological Signals

Recently, robot services have been widely applied in many fields. To provide optimum service, it is essential to maintain good acceptance of the robot for more effective interaction with users. Previously, we attempted to implement facial expressions by synchronizing an estimated human emotion on the face of a robot. The results revealed that the robot could present different perceptions according to individual preferences. In this study, we considered individual differences to improve the acceptance of the robot by changing the robot’s expression according to the emotion of its interacting partner. The emotion was estimated using biological signals, and the robot changed its expression according to three conditions: synchronized with the estimated emotion, inversely synchronized, and a funny expression. During the experiment, the participants provided feedback regarding the robot’s expression by choosing whether they “like” or “dislike” the expression. We investigated individual differences in the acceptance of the robot expression using the Semantic Differential scale method. In addition, logistic regression was used to create a classification model by considering individual differences based on the biological data and feedback from each participant. We found that the robot expression based on inverse synchronization when the participants felt a negative emotion could result in impression differences among individuals. Then, the robot’s expression was determined based on the classification model, and the Semantic Differential scale on the impression of the robot was compared with the three conditions. Overall, we found that the participants were most accepting when the robot expression was calculated using the proposed personalized method.

A hierarchical approach to removal of unwanted variation for large-scale metabolomics data

Liquid chromatography-mass spectrometry-based metabolomics studies are increasingly applied to large population cohorts, which run for several weeks or even years in data acquisition. This inevitably introduces unwanted intra- and inter-batch variations over time that can overshadow true biological signals and thus hinder potential biological discoveries. To date, normalisation approaches have struggled to mitigate the variability introduced by technical factors whilst preserving biological variance, especially for protracted acquisitions. Here, we propose a study design framework with an arrangement for embedding biological sample replicates to quantify variance within and between batches and a workflow that uses these replicates to remove unwanted variation in a hierarchical manner (hRUV). We use this design to produce a dataset of more than 1000 human plasma samples run over an extended period of time. We demonstrate significant improvement of hRUV over existing methods in preserving biological signals whilst removing unwanted variation for large scale metabolomics studies. Our tools not only provide a strategy for large scale data normalisation, but also provides guidance on the design strategy for large omics studies.

Dynamical Analysis of Biological Signals with the 0–1 Test: A Case Study of the PhotoPlethysmoGraphic (PPG) Signal

The 0–1 test distinguishes between regular and chaotic dynamics for a deterministic system using a time series as a starting point without appealing to any state space reconstruction method. A modification of the 0–1 test allows for the determination of a more comprehensive range of signal dynamic behaviors, particularly in the field of biological signals. We report the results of applying the test and study with more details the PhotoPlethysmoGraphic (PPG) signal behavior from different healthy young subjects, although its use is extensible to other biological signals. While mainly used for heart rate and blood oxygen saturation monitoring, the PPG signal contains extensive physiological dynamics information. We show that the PPG signal, on a healthy young individual, is predominantly quasi-periodic on small timescales (short span of time concerning the dominant frequency). However, on large timescales, PPG signals yield an aperiodic behavior that can be firmly chaotic or a prior transition via an SNA (Strange Nonchaotic Attractor). The results are based on the behavior of well-known time series that are random, chaotic, aperiodic, periodic, and quasi-periodic.