Entropy

Research

Jump to: Review

12 pages, 2625 KiB

Open AccessArticle

Information Theoretic-Based Interpretation of a Deep Neural Network Approach in Diagnosing Psychogenic Non-Epileptic Seizures

by Sara Gasparini, Maurizio Campolo, Cosimo Ieracitano, Nadia Mammone, Edoardo Ferlazzo, Chiara Sueri, Giovanbattista Gaspare Tripodi, Umberto Aguglia and Francesco Carlo Morabito

Entropy 2018, 20(2), 43; https://doi.org/10.3390/e20020043 - 23 Jan 2018

Cited by 35 | Viewed by 5163

Abstract

The use of a deep neural network scheme is proposed to help clinicians solve a difficult diagnosis problem in neurology. The proposed multilayer architecture includes a feature engineering step (from time-frequency transformation), a double compressing stage trained by unsupervised learning, and a classification [...] Read more.

The use of a deep neural network scheme is proposed to help clinicians solve a difficult diagnosis problem in neurology. The proposed multilayer architecture includes a feature engineering step (from time-frequency transformation), a double compressing stage trained by unsupervised learning, and a classification stage trained by supervised learning. After fine-tuning, the deep network is able to discriminate well the class of patients from controls with around 90% sensitivity and specificity. This deep model gives better classification performance than some other standard discriminative learning algorithms. As in clinical problems there is a need for explaining decisions, an effort has been carried out to qualitatively justify the classification results. The main novelty of this paper is indeed to give an entropic interpretation of how the deep scheme works and reach the final decision. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Figure 1
(A) The flowchart of the method: (a) The 19-channels electroencephalography (EEG) recording is partitioned into M = 20 non-overlapping epochs (of 5 s width), (b) given an epoch, the time frequency map (TFM) is estimated over each channel. The i-th TFM (i = 1, …, 19) is partitioned into three sub-bands (<math display="inline"> <semantics> <mrow> <mi>S</mi> <msub> <mi>B</mi> <mi>j</mi> </msub> </mrow> </semantics> </math> j = 1, 2, 3); then, the mean (µ), standard deviation (σ), and skewness (ν) of the wavelet coefficients are evaluated for each SB and for the whole TFM. Once the TFMs are computed on the M = 20 epochs, a database of 20 × 12 × 19 data (#epochs × #features × #channels) is generated, (c) The vectors of 228 features are the input of a 2-stacked autoencoders (SAE) architecture. The last softmax layer performs the 2-way classification task (CNT-PNES); (B) The two Autoencodes (AE) implemented: the first AE compresses the 228 input features to 50 parameters (encoder stage) and then attempts to reconstruct the input (decoder stage); whereas, the second AE compresses the 50 features output of the first AE to 20 latent parameters. The compressed representations H1 (50 × 1) and H2 (20 × 1) (indicated in red and green, respectively) are used in the stacked autoencoders architecture. Full article ">Figure 1 Cont.
(A) The flowchart of the method: (a) The 19-channels electroencephalography (EEG) recording is partitioned into M = 20 non-overlapping epochs (of 5 s width), (b) given an epoch, the time frequency map (TFM) is estimated over each channel. The i-th TFM (i = 1, …, 19) is partitioned into three sub-bands (<math display="inline"> <semantics> <mrow> <mi>S</mi> <msub> <mi>B</mi> <mi>j</mi> </msub> </mrow> </semantics> </math> j = 1, 2, 3); then, the mean (µ), standard deviation (σ), and skewness (ν) of the wavelet coefficients are evaluated for each SB and for the whole TFM. Once the TFMs are computed on the M = 20 epochs, a database of 20 × 12 × 19 data (#epochs × #features × #channels) is generated, (c) The vectors of 228 features are the input of a 2-stacked autoencoders (SAE) architecture. The last softmax layer performs the 2-way classification task (CNT-PNES); (B) The two Autoencodes (AE) implemented: the first AE compresses the 228 input features to 50 parameters (encoder stage) and then attempts to reconstruct the input (decoder stage); whereas, the second AE compresses the 50 features output of the first AE to 20 latent parameters. The compressed representations H1 (50 × 1) and H2 (20 × 1) (indicated in red and green, respectively) are used in the stacked autoencoders architecture. Full article ">Figure 2
Time frequency representation of the psychogenic non-epileptic seizures (PNES) and healthy control (CNT). Each epoch of the 19-channels electroencephalography (EEG) is transformed in a time frequency map (TFM); then, the mean over the 19 channels, over the subjects and over the epochs is evaluated coming up with a single TFM per class. (a) TFM averaged over the 19 channels, the 20 epochs, and the six PNES subjects; (b) TFM averaged over 19 channels, the 20 epochs, and the 10 CNT subjects. Full article ">Figure 3
Softmax output representation of PNES (a) and CNT (b) for the 20 leave-one-out testing sessions carried out for every subject. Each bin represents the output estimated by the softmax layer ranged between 0 and 1 (1 correct classification; 0 misclassification). The red dotted line is the average output level of the network, evaluated over the 20 sessions. Full article ">Figure 4
Entropy representation of PNES (red dots) and CNT (blue dots) evaluated at the outputs of the hidden nodes of the two compressed representations. (a) Entropy values related to PNES and CNT features extracted from the first AE (50 × 1). At this stage, the entropies of the two classes are comparable; (b) Entropy values related to PNES and CNT features extracted from the second AE (20 × 1). At this stage, the entropies decrease and they are different for the two classes and generally greater for PNES than CNT. Full article ">

19 pages, 1072 KiB

Open AccessFeature PaperArticle

Characterizing Normal and Pathological Gait through Permutation Entropy

by Massimiliano Zanin, David Gómez-Andrés, Irene Pulido-Valdeolivas, Juan Andrés Martín-Gonzalo, Javier López-López, Samuel Ignacio Pascual-Pascual and Estrella Rausell

Entropy 2018, 20(1), 77; https://doi.org/10.3390/e20010077 - 19 Jan 2018

Cited by 18 | Viewed by 6091

Abstract

Cerebral palsy is a physical impairment stemming from a brain lesion at perinatal time, most of the time resulting in gait abnormalities: the first cause of severe disability in childhood. Gait study, and instrumental gait analysis in particular, has been receiving increasing attention [...] Read more.

Cerebral palsy is a physical impairment stemming from a brain lesion at perinatal time, most of the time resulting in gait abnormalities: the first cause of severe disability in childhood. Gait study, and instrumental gait analysis in particular, has been receiving increasing attention in the last few years, for being the complex result of the interactions between different brain motor areas and thus a proxy in the understanding of the underlying neural dynamics. Yet, and in spite of its importance, little is still known about how the brain adapts to cerebral palsy and to its impaired gait and, consequently, about the best strategies for mitigating the disability. In this contribution, we present the hitherto first analysis of joint kinematics data using permutation entropy, comparing cerebral palsy children with a set of matched control subjects. We find a significant increase in the permutation entropy for the former group, thus indicating a more complex and erratic neural control of joints and a non-trivial relationship between the permutation entropy and the gait speed. We further show how this information theory measure can be used to train a data mining model able to forecast the child’s condition. We finally discuss the relevance of these results in clinical applications and specifically in the design of personalized medicine interventions. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Figure 1
Probability distributions of the Permutation Entropy (PE), as calculated in control subjects and Cerebral Palsy (CP) patients, the latter including aggregated and disaggregated (w.r.t. the Gross Motor Function Classification System (GMFCS) scale) results. Rows, from top to bottom, respectively correspond to pelvis, hip, knee, ankle and forefoot; columns, from left to right, to the abduction-adduction, sagittal and rotational axes. Full article ">Figure 2
Forest plots showing the beta coefficients of linear mixed models comparing gait PE values according to the patient’s GMFCS level. Squares represent the mean value of each beta coefficient and horizontal lines the corresponding <math display="inline"> <semantics> <mrow> <mn>95</mn> <mo>%</mo> </mrow> </semantics> </math> bias-corrected and accelerated bootstrap intervals. See main text and <a href="#sec4dot3-entropy-20-00077" class="html-sec">Section 4.3</a> for details. Full article ">Figure 3
Permutation entropy as a function of the normalized walking speed, for control subjects (black dots) and CP patients (red dots). Each panel corresponds to the same joint/axis as in <a href="#entropy-20-00077-f001" class="html-fig">Figure 1</a>. Full article ">Figure 4
Forest plots showing the beta coefficients of linear mixed models comparing gait PE values according to normalized walking speed (left panel), condition (using healthy as the reference, central panel) and the interaction of normalized walking speed with condition (right panel). The magnitudes of the effects are indicated on the X axis. Squares represent the mean values of each beta coefficient and horizontal lines the corresponding <math display="inline"> <semantics> <mrow> <mn>95</mn> <mo>%</mo> </mrow> </semantics> </math> bias-corrected and accelerated bootstrap intervals. See <a href="#sec4dot3-entropy-20-00077" class="html-sec">Section 4.3</a> for further details. Full article ">Figure 5
Forest plots showing the beta coefficients of linear mixed models that measure the effect of PE on the Gait Deviation Index (GDI, left), the Global Profile Score (GPS, center) and elements of the Movement Analysis Profile (MAP, right). The magnitudes of the effects are indicated in the X axis. Squares represent the mean values of each beta coefficient and horizontal lines the corresponding <math display="inline"> <semantics> <mrow> <mn>95</mn> <mo>%</mo> </mrow> </semantics> </math> bias-corrected and accelerated bootstrap intervals. See <a href="#sec4dot3-entropy-20-00077" class="html-sec">Section 4.3</a> for further details. Full article ">Figure 6
(Top) Multi-scale PE, for control subjects (black lines) and CP patients (red lines), as a function of the down-sampling <math display="inline"> <semantics> <mi>υ</mi> </semantics> </math>; see <a href="#sec4dot2dot2-entropy-20-00077" class="html-sec">Section 4.2.2</a> for details. Each panel corresponds to the same joint/axis as in <a href="#entropy-20-00077-f001" class="html-fig">Figure 1</a>. (Bottom) <math display="inline"> <semantics> <mo>Δ</mo> </semantics> </math>MSE for all joint/axis pairs; see Equation (<a href="#FD1-entropy-20-00077" class="html-disp-formula">1</a>). Full article ">Figure 7
Classifying patients according to their gait entropy. The left panel depicts the Receiver Operating Characteristic (ROC) curve (blue solid line), obtained through a random forest model; the dashed grey line represents the result obtained by a random classification. The right panel depicts the drop in the Area Under the Curve (AUC) when individual features (joint/axis pairs) are deleted from the dataset; the higher the value, the more important is the considered feature. Full article ">Figure 8
Random forest classification of the patients’ GMFCS stage according to the PE of the joint time series. The left panel shows the importance of the individual features. The higher the value in the X axis of the left panel, the more important the corresponding feature is for an accurate classification. Importance is estimated according to the increase in the classification error when this feature is randomly permuted. The right panels show the adjusted class probability for healthy, GMFCS I, GMFCS II, GMFCS III and GMFCS IV stages according to an RF classification. In the X axis, values of PE of hip flexion (upper plot) and ankle flexion (lower plot) are shown. Different values presented in the split are shown by a small mark in the axis. The left Y axis of the panel indicates the adjusted class probability, while the right one shows the proportion of cycles of the different classes. Full article ">

3283 KiB

Open AccessArticle

Normalised Mutual Information of High-Density Surface Electromyography during Muscle Fatigue

by Adrian Bingham, Sridhar P. Arjunan, Beth Jelfs and Dinesh K. Kumar

Entropy 2017, 19(12), 697; https://doi.org/10.3390/e19120697 - 20 Dec 2017

Cited by 15 | Viewed by 6240

Abstract

This study has developed a technique for identifying the presence of muscle fatigue based on the spatial changes of the normalised mutual information (NMI) between multiple high density surface electromyography (HD-sEMG) channels. Muscle fatigue in the tibialis anterior (TA) during isometric contractions at [...] Read more.

This study has developed a technique for identifying the presence of muscle fatigue based on the spatial changes of the normalised mutual information (NMI) between multiple high density surface electromyography (HD-sEMG) channels. Muscle fatigue in the tibialis anterior (TA) during isometric contractions at 40% and 80% maximum voluntary contraction levels was investigated in ten healthy participants (Age range: 21 to 35 years; Mean age = 26 years; Male = 4, Female = 6). HD-sEMG was used to record 64 channels of sEMG using a 16 by 4 electrode array placed over the TA. The NMI of each electrode with every other electrode was calculated to form an NMI distribution for each electrode. The total NMI for each electrode (the summation of the electrode’s NMI distribution) highlighted regions of high dependence in the electrode array and was observed to increase as the muscle fatigued. To summarise this increase, a function, M(k), was defined and was found to be significantly affected by fatigue and not by contraction force. The technique discussed in this study has overcome issues regarding electrode placement and was used to investigate how the dependences between sEMG signals within the same muscle change spatially during fatigue. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

2674 KiB

Open AccessArticle

Entropy and Compression Capture Different Complexity Features: The Case of Fetal Heart Rate

by João Monteiro-Santos, Hernâni Gonçalves, João Bernardes, Luís Antunes, Mohammad Nozari and Cristina Costa-Santos

Entropy 2017, 19(12), 688; https://doi.org/10.3390/e19120688 - 14 Dec 2017

Cited by 9 | Viewed by 4534

Abstract

Entropy and compression have been used to distinguish fetuses at risk of hypoxia from their healthy counterparts through the analysis of Fetal Heart Rate (FHR). Low correlation that was observed between these two approaches suggests that they capture different complexity features. This study [...] Read more.

Entropy and compression have been used to distinguish fetuses at risk of hypoxia from their healthy counterparts through the analysis of Fetal Heart Rate (FHR). Low correlation that was observed between these two approaches suggests that they capture different complexity features. This study aims at characterizing the complexity of FHR features captured by entropy and compression, using as reference international guidelines. Single and multi-scale approaches were considered in the computation of entropy and compression. The following physiologic-based features were considered: FHR baseline; percentage of abnormal long (%abLTV) and short (%abSTV) term variability; average short term variability; and, number of acceleration and decelerations. All of the features were computed on a set of 68 intrapartum FHR tracings, divided as normal, mildly, and moderately-severely acidemic born fetuses. The correlation between entropy/compression features and the physiologic-based features was assessed. There were correlations between compressions and accelerations and decelerations, but neither accelerations nor decelerations were significantly correlated with entropies. The %abSTV was significantly correlated with entropies (ranging between ?0.54 and ?0.62), and to a higher extent with compression (ranging between ?0.80 and ?0.94). Distinction between groups was clearer in the lower scales using entropy and in the higher scales using compression. Entropy and compression are complementary complexity measures. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Figure 1
Multiscale analysis of tracings using Approximate entropy (ApEn) with tolerance 0.2. Plotted 95% Confidence Intervals of the mean for each group in each scale. (MA: mildly acidemic fetuses; MSA: moderate-to-severe acidemic fetuses; N: the normal range). Full article ">Figure 2
Multiscale analysis of tracings using Sample entropy (SampEn) with tolerance 0.2. Plotted 95% Confidence Intervals of the mean for each group in each scale. Full article ">Figure 3
Multiscale analysis of tracings using compressor Brotli with maximum level of compression. Plotted 95% Confidence Intervals of the mean for each group in each scale. Full article ">Figure 4
Multiscale analysis of tracings using compressor Gzip with maximum level of compression. Plotted 95% Confidence Intervals of the mean for each group in each scale. Full article ">Figure 5
Multiscale analysis of tracings using compressor Paq8l with maximum level of compression. Plotted 95% Confidence Intervals of the mean for each group in each scale. Full article ">Figure 6
Multiscale analysis of tracings using compressor Bzip2 with maximum level of compression. Plotted 95% Confidence Intervals of the mean for each group in each scale. Full article ">

848 KiB

Open AccessArticle

Altered Brain Complexity in Women with Primary Dysmenorrhea: A Resting-State Magneto-Encephalography Study Using Multiscale Entropy Analysis

by Intan Low, Po-Chih Kuo, Yu-Hsiang Liu, Cheng-Lin Tsai, Hsiang-Tai Chao, Jen-Chuen Hsieh, Li-Fen Chen and Yong-Sheng Chen

Entropy 2017, 19(12), 680; https://doi.org/10.3390/e19120680 - 11 Dec 2017

Cited by 11 | Viewed by 6501

Abstract

How chronic pain affects brain functions remains unclear. As a potential indicator, brain complexity estimated by entropy-based methods may be helpful for revealing the underlying neurophysiological mechanism of chronic pain. In this study, complexity features with multiple time scales and spectral features were [...] Read more.

How chronic pain affects brain functions remains unclear. As a potential indicator, brain complexity estimated by entropy-based methods may be helpful for revealing the underlying neurophysiological mechanism of chronic pain. In this study, complexity features with multiple time scales and spectral features were extracted from resting-state magnetoencephalographic signals of 156 female participants with/without primary dysmenorrhea (PDM) during pain-free state. Revealed by multiscale sample entropy (MSE), PDM patients (PDMs) exhibited loss of brain complexity in regions associated with sensory, affective, and evaluative components of pain, including sensorimotor, limbic, and salience networks. Significant correlations between MSE values and psychological states (depression and anxiety) were found in PDMs, which may indicate specific nonlinear disturbances in limbic and default mode network circuits after long-term menstrual pain. These findings suggest that MSE is an important measure of brain complexity and is potentially applicable to future diagnosis of chronic pain. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Figure 1
Examples of time series data for an 8-s epoch over a range of representative scale factors (τ) from fine to coarse scales: 1, 5, 30, 60 and 90. The x-axis represents time and the y-axis represents the estimated brain activity index. Full article ">Figure 2
Significant multiscale sample entropy (MSE) group differences from resting-state MEG brain signals during pain-free state. MSE values of PDMs were mostly lower than those of CONs. Significant MSE group differences were found at larger scale factors (SFs) and mainly in the right hemisphere. (a) Histogram of the number of regions showing significant group differences (permutation testing with 5000 iterations, p < 0.05, two-tailed) from SFs 1 to 10: SFs 1~50 (0), 51~60 (4), 61–70 (8), 71~80 (6), 81~90 (10), and 91~100 (14). Top three: SF 92 (count = 4), SF 81 (count = 3), and SF 99 (count = 3); (b) MSE group differences. The x-axis represents the SFs of MSE ranging from 51 to 100. The y-axis represents the AAL brain regions, which were categorized into different RSNs (indicated by thick lines and text in different colors) for functional attribution. The color bar represents between-group t-values after 5000 permutation tests, with the most positive t-values (PDMs > CONs) shown in bright yellow and the most negative t-values (PDMs < CONs) shown in bright blue; (c) Brain regions that revealed significant MSE group differences were grouped into five SF-intervals for demonstration: SFs 51~60, 61–70, 71~80, 81~90 and 91~100. Brain regions where MSE measures in PDMs were higher or lower than those in CONs are represented in red or blue spheres, respectively. Size of sphere represents the count of scale factors that showed significant group differences. PDM, primary dysmenorrhea patients; CON, healthy female controls; L, left hemisphere; R, right hemisphere; LIMBIC, limbic network; DMN, default mode network; SAN, salience network; SMN, sensorimotor network; ECN, executive control network; AMYG, amygdala; ANG, angular gyrus; ITG, inferior temporal gyrus; CAL, calcarine; ROL, rolandic operculum; IPL, inferior parietal lobule; LING, lingual gyrus; IFGoperc, opercularis of inferior frontal gyrus; SOG, superior occipital gyrus; CUN, cuneus; PHG, parahippocampal gyrus; FFG, fusiform gyrus; MTG, middle temporal gyrus; MOG, middle occipital gyrus; ACG, anterior cingulate gyrus; THA, thalamus; HIP, hippocampus; PCUN, precuenus. Full article ">Figure 3
Hemispheric asymmetry of multiscale sample entropy (asymMSE) group differences from resting-state MEG brain signals during pain-free state. Significant group differences were found across all scale factor-intervals with the most differences seen between SFs 31~60. (a) Count of each SF that showed significant group differences (permutation testing with 5000 iterations, p < 0.01, two-tailed); (b) Group differences in hemispheric asymmetry of MSE (asymMSE). The x-axis represents the scale factors (SFs) of asymMSE ranging from 1 to 100; the y-axis represents the AAL brain regions which were categorized into different RSNs (indicated by thick lines and text in different colors) for functional attribution; the color bar represents between-group t-values after 5000 permutation tests (α = 0.01) as listed in <a href="#app1-entropy-19-00680" class="html-app">Table S2</a>, with the most positive t-values (PDMs > CONs) shown in bright yellow and the most negative t-values (PDMs < CONs) shown in bright blue; (c) Altered hemispheric asymmetry of MSE in PDMs. Brain regions that revealed significant asymMSE group differences were grouped into six SF-intervals for demonstration (<a href="#app1-entropy-19-00680" class="html-app">Table S2</a>): SFs 1~30, 31~40, 41~50, 51~60, 61–80 and 81~100. CON group exhibited mostly left-lateralization. Compared to CONs, grey, red, and blue spheres indicate the brain regions with loss of hemispheric asymmetry, right-lateralization, and left-lateralization in PDMs, respectively. Size of sphere represents count of scale factors that showed significant group differences. PDM, primary dysmenorrhea patients; CON, healthy female controls; LIMBIC, limbic network; DMN, default mode network; SAN, salience network; SMN, sensorimotor network; ECN, executive control network; L, left hemisphere; R, right hemisphere; MFG, middle frontal gyrus; ORBmid, orbital MFG; ROL, rolandic operculum; PUT, putamen; HES, Heschl gyrus; SFG, superior frontal gyrus; ORBsupmed, medial orbital SFG; SFGmed, medial SFG; ACG, anterior cingulate gyrus; CAU, caudate; IFGoperc, opercularis of inferior frontal gyrus; AMYG, amygdala; PHG, parahippocampal g.; SFGdor, dorsolateral SFG; HIP, hippocampus; THA, thalamus. Full article ">Figure 4
Group differences of hemispheric asymmetry of spectral and other complexity features from resting-state MEG brain signals during pain-free state. Significant group differences were only disclosed in hemispheric asymmetry of high gamma relative band power, median frequency, and spectral edge frequency, and were left-lateralized mainly in the frontal regions in PDMs. (a) Group differences in hemispheric asymmetry. The x-axis of the matrix represents the hemispheric asymmetry of spectral and other brain complexity features; the y-axis represents the AAL brain regions which were categorized into different RSNs (indicated by thick lines and text in different colors) for functional attribution; the color bar represents between-group t-values after 5000 permutation tests (α = 0.01), with the most positive t-values (PDMs > CONs) shown in bright yellow and the most negative t-values (PDMs < CONs) shown in bright blue; (b) Hemispheric asymmetry alterations in PDMs were displayed (<a href="#app1-entropy-19-00680" class="html-app">Table S4</a>). Blue spheres depict the brain regions with enhanced left lateralization in PDMs compared to CONs. PDM, primary dysmenorrhea patients; CON, healthy female controls; L, left hemisphere; R, right hemisphere; g., gyrus; IFG, inferior frontal gyrus; MFG, middle frontal gyrus; ORBmid, orbital MFG; ORBinf, orbital IFG, orbital; LING, lingual gyrus; RP, relative band power; δ, <math display="inline"> <semantics> <mrow> <mi>RP</mi> <mrow> <mo>(</mo> <mi>δ</mi> <mo>)</mo> </mrow> </mrow> </semantics> </math>; θ, <math display="inline"> <semantics> <mrow> <mi>RP</mi> <mrow> <mo>(</mo> <mi>θ</mi> <mo>)</mo> </mrow> <mo>;</mo> </mrow> </semantics> </math> α, <math display="inline"> <semantics> <mrow> <mi>RP</mi> <mrow> <mo>(</mo> <mi>α</mi> <mo>)</mo> </mrow> <mo>;</mo> </mrow> </semantics> </math> β, <math display="inline"> <semantics> <mrow> <mi>RP</mi> <mrow> <mo>(</mo> <mi>β</mi> <mo>)</mo> </mrow> <mo>;</mo> </mrow> </semantics> </math> Lγ, <math display="inline"> <semantics> <mrow> <mi>RP</mi> <mrow> <mo>(</mo> <mrow> <mi>l</mi> <mi>o</mi> <mi>w</mi> <mo> </mo> <mi>γ</mi> </mrow> <mo>)</mo> </mrow> <mo>;</mo> </mrow> </semantics> </math> Hγ, <math display="inline"> <semantics> <mrow> <mi>RP</mi> <mrow> <mo>(</mo> <mrow> <mi>h</mi> <mi>i</mi> <mi>g</mi> <mi>h</mi> <mo> </mo> <mi>γ</mi> </mrow> <mo>)</mo> </mrow> <mo>;</mo> </mrow> </semantics> </math> MF, median frequency; SEF, Spectral edge frequency; SSE, Shannon sample entropy; LZC, Lempel-Ziv complexity. Full article ">Figure 5
Spearman correlation between multiscale sample entropy and pain experiences or psychological characteristics in primary dysmenorrhea patients (PDMs). The x-axis of each subfigure is the value of MSE or hemispheric asymmetry of MSE (asymMSE) in each PDMs; the y-axis is the pain and/or psychological scores. Spearman rho and p-values are listed for each correlation. (a) Brain complexity in the right hippocampus positively correlated with depression score; (b) Brain complexity in the left thalamus positively correlated with depression scores; (c) Brain complexity in the left Rolandic operculum negatively correlated with physical component score of quality of life; (d) Asymmetry of brain complexity in the parahippocampal gyrus was positively correlated with anxiety score. MSE, multiscale sample entropy; asymMSE, hemispheric asymmetry of MSE; BDI, Beck depression inventory; BAI, Beck anxiety inventory; PCS, pain catastrophizing scale; MPQ, McGill pain questionnaire; PRI, MPQ pain rating index; PPI, MPQ present pain index. Full article ">

4601 KiB

Open AccessArticle

Automated Detection of Paroxysmal Atrial Fibrillation Using an Information-Based Similarity Approach

by Xingran Cui, Emily Chang, Wen-Hung Yang, Bernard C. Jiang, Albert C. Yang and Chung-Kang Peng

Entropy 2017, 19(12), 677; https://doi.org/10.3390/e19120677 - 10 Dec 2017

Cited by 27 | Viewed by 6025

Abstract

Atrial fibrillation (AF) is an abnormal rhythm of the heart, which can increase heart-related complications. Paroxysmal AF episodes occur intermittently with varying duration. Human-based diagnosis of paroxysmal AF with a longer-term electrocardiogram recording is time-consuming. Here we present a fully automated ensemble model [...] Read more.

Atrial fibrillation (AF) is an abnormal rhythm of the heart, which can increase heart-related complications. Paroxysmal AF episodes occur intermittently with varying duration. Human-based diagnosis of paroxysmal AF with a longer-term electrocardiogram recording is time-consuming. Here we present a fully automated ensemble model for AF episode detection based on RR-interval time series, applying a novel approach of information-based similarity analysis and ensemble scheme. By mapping RR-interval time series to binary symbolic sequences and comparing the rank-frequency patterns of m-bit words, the dissimilarity between AF and normal sinus rhythms (NSR) were quantified. To achieve high detection specificity and sensitivity, and low variance, a weighted variation of bagging with multiple AF and NSR templates was applied. By performing dissimilarity comparisons between unknown RR-interval time series and multiple templates, paroxysmal AF episodes were detected. Based on our results, optimal AF detection parameters are symbolic word length m = 9 and observation window n = 150, achieving 97.04% sensitivity, 97.96% specificity, and 97.78% overall accuracy. Sensitivity, specificity, and overall accuracy vary little despite changes in m and n parameters. This study provides quantitative information to enhance the categorization of AF and normal cardiac rhythms. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Figure 1
Representative inter-beat (RR) interval time series derived from an electrocardiographic recording of a patient with paroxysmal atrial fibrillation. The dark circles represent consecutive RR intervals and the solid line indicates the presence/absence of AF episodes as reported in the annotations of PhysioNet database. During an episode of atrial fibrillation, the line is set to “AF”; otherwise it is set to “Non-AF”, which means a rhythm that is not atrial fibrillation. Full article ">Figure 2
Schematic illustration of the mapping procedure for 6-bit words (m = 6). Full article ">Figure 3
Representative inter-beat time series for a healthy subject (a) and an AF patient (b); (c) Rank order comparison of the time series for two healthy subjects; (d) Rank order comparison of the time series for two AF patients; (e) Rank order comparison of the time series in (a,b). The results in (c–e) are for the case m = 6. Full article ">Figure 4
(a) Graphic illustration of detection results for a testing data (record 04908, m = 9, observation window n = 150, Δ = 0.1); (b) enlarged AF segment derived from (a); (c) Enlarged signal segment of neither AF nor normal beats to compare with (b). Full article ">Figure 5
Graphic illustration of detection results for a testing data (record 04043, m = 9, n = 150, Δ = 0.05). Full article ">Figure 6
Computation time according to m and n (the computation times are the average values of 100 trials). Full article ">

3539 KiB

Open AccessArticle

Association between Multiscale Entropy Characteristics of Heart Rate Variability and Ischemic Stroke Risk in Patients with Permanent Atrial Fibrillation

by Ryo Matsuoka, Kohzoh Yoshino, Eiichi Watanabe and Ken Kiyono

Entropy 2017, 19(12), 672; https://doi.org/10.3390/e19120672 - 7 Dec 2017

Cited by 3 | Viewed by 4947

Abstract

Multiscale entropy (MSE) profiles of heart rate variability (HRV) in patients with atrial fibrillation (AFib) provides clinically useful information for ischemic stroke risk assessment, suggesting that the complex properties characterized by MSE profiles are associated with ischemic stroke risk. However, the meaning of [...] Read more.

Multiscale entropy (MSE) profiles of heart rate variability (HRV) in patients with atrial fibrillation (AFib) provides clinically useful information for ischemic stroke risk assessment, suggesting that the complex properties characterized by MSE profiles are associated with ischemic stroke risk. However, the meaning of HRV complexity in patients with AFib has not been clearly interpreted, and the physical and mathematical understanding of the relation between HRV dynamics and the ischemic stroke risk is not well established. To gain a deeper insight into HRV dynamics in patients with AFib, and to improve ischemic stroke risk assessment using HRV analysis, we study the HRV characteristics related to MSE profiles, such as the long-range correlation and probability density function. In this study, we analyze the HRV time series of 173 patients with permanent AFib. Our results show that, although HRV time series in patients with AFib exhibit long-range correlation (1/f fluctuations)—as observed in healthy subjects—in a range longer than 90 s, these autocorrelation properties have no significant predictive power for ischemic stroke occurrence. Further, the probability density function structure of the coarse-grained times series at scales greater than 2 s is dominantly associated with ischemic stroke risk. This observation could provide valuable information for improving ischemic stroke risk assessment using HRV analysis. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

2152 KiB

Open AccessArticle

Assessment of Heart Rate Variability during an Endurance Mountain Trail Race by Multi-Scale Entropy Analysis

by Montserrat Vallverdú, Aroa Ruiz-Muñoz, Emma Roca, Pere Caminal, Ferran A. Rodríguez, Alfredo Irurtia and Alexandre Perera

Entropy 2017, 19(12), 658; https://doi.org/10.3390/e19120658 - 1 Dec 2017

Cited by 7 | Viewed by 6042

Abstract

The aim of the study was to analyze heart rate variability (HRV) response to high-intensity exercise during a 35-km mountain trail race and to ascertain whether fitness level could influence autonomic nervous system (ANS) modulation. Time-domain, frequency-domain, and multi-scale entropy (MSE) [...] Read more.

The aim of the study was to analyze heart rate variability (HRV) response to high-intensity exercise during a 35-km mountain trail race and to ascertain whether fitness level could influence autonomic nervous system (ANS) modulation. Time-domain, frequency-domain, and multi-scale entropy (MSE) indexes were calculated for eleven mountain-trail runners who completed the race. Many changes were observed, mostly related to exercise load and fatigue. These changes were characterized by increased mean values and standard deviations of the normal-to-normal intervals associated with sympathetic activity, and by decreased differences between successive intervals related to parasympathetic activity. Normalized low frequency (LF) power suggested that ANS modulation varied greatly during the race and between individuals. Normalized high frequency (HF) power, associated with parasympathetic activity, varied considerably over the race, and tended to decrease at the final stages, whereas changes in the LF/HF ratio corresponded to intervals with varying exercise load. MSE indexes, related to system complexity, indicated the existence of many interactions between the heart and its neurological control mechanism. The time-domain, frequency-domain, and MSE indexes were also able to discriminate faster from slower runners, mainly in the more difficult and in the final stages of the race. These findings suggest the use of HRV analysis to study cardiac function mechanisms in endurance sports. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Figure 1
Elevation profile of the route. The 36 segments of 5-min duration selected from the fastest runner are displayed. Full article ">Figure 2
Time series of selected performance variables in different segments of the race for the fastest runner: (a) <math display="inline"> <semantics> <mrow> <mi>T</mi> <mi>S</mi> </mrow> </semantics> </math>, terrain slope; (b), <math display="inline"> <semantics> <mrow> <mi>E</mi> <mi>L</mi> </mrow> </semantics> </math>, exercise load; (c) FI, fatigue interval; and (d) <math display="inline"> <semantics> <mrow> <mi>R</mi> <mi>S</mi> </mrow> </semantics> </math>, running speed. Full article ">Figure 3
Boxplots of meanNN index (a); sdNN index (b) and LF/HF ratio (c) for the 36 RR-segments analysed over the course considering the whole group of runners. Those segments showing inter-segment statistical differences (p-value < 0.05) are marked with *. Full article ">Figure 4
Cross-tabulated p-values chart (as-log10(p-values)) comparing the meanNN indexes from all RR-segments of the race. Cells containing significant p-values are coloured dark red (p-value < 0.005), medium red (0.005 ≤ p-value < 0.01), or light red (0.01 ≤ p-value ≤ 0.05). Full article ">Figure 5
Changes in the SampEn values during the race for the whole group of runners: (a) Mean value and standard error of MSE indexes as a function of the scale τ derived from the thirty-six RR-segments during the entire race; (b) Number of MSE indexes which significant differed among the thirty-six RR-segments. Full article ">Figure 6
Boxplots of SampEn for the 36 RR-segments analysed over the course considering all the runners at scales τ = 1 (a), τ = 6 (b) and τ = 14 (c). Segments showing inter-segmental statistical differences (p-value < 0.05) are marked with *. Full article ">Figure 6 Cont.
Boxplots of SampEn for the 36 RR-segments analysed over the course considering all the runners at scales τ = 1 (a), τ = 6 (b) and τ = 14 (c). Segments showing inter-segmental statistical differences (p-value < 0.05) are marked with *. Full article ">Figure 7
MSE values and their averaged values (dashed line) at each τ for the two groups of runners in a specific segment: s9 in (a); s17 in (b); s25 in (c); s31 in (d). Grey circles correspond to the faster group of runners (FR) while black circles correspond to the slower group of runners (SR). Full article ">

2018 KiB

Open AccessFeature PaperArticle

by Theerasak Chanwimalueang and Danilo P. Mandic

Entropy 2017, 19(12), 652; https://doi.org/10.3390/e19120652 - 30 Nov 2017

Cited by 40 | Viewed by 9679

Abstract

The nonparametric Sample Entropy (SE) estimator has become a standard for the quantification of structural complexity of nonstationary time series, even in critical cases of unfavorable noise levels. The SE has proven very successful for signals that exhibit a certain degree of the [...] Read more.

The nonparametric Sample Entropy (SE) estimator has become a standard for the quantification of structural complexity of nonstationary time series, even in critical cases of unfavorable noise levels. The SE has proven very successful for signals that exhibit a certain degree of the underlying structure, but do not obey standard probability distributions, a typical case in real-world scenarios such as with physiological signals. However, the SE estimates structural complexity based on uncertainty rather than on (self) correlation, so that, for reliable estimation, the SE requires long data segments, is sensitive to spikes and erratic peaks in data, and owing to its amplitude dependence it exhibits lack of precision for signals with long-term correlations. To this end, we propose a class of new entropy estimators based on the similarity of embedding vectors, evaluated through the angular distance, the Shannon entropy and the coarse-grained scale. Analysis of the effects of embedding dimension, sample size and tolerance shows that the so introduced Cosine Similarity Entropy (CSE) and the enhanced Multiscale Cosine Similarity Entropy (MCSE) are amplitude-independent and therefore superior to the SE when applied to short time series. Unlike the SE, the CSE is shown to yield valid entropy values over a broad range of embedding dimensions. By evaluating the CSE and the MCSE over a variety of benchmark synthetic signals as well as for real-world data (heart rate variability of three different cardiovascular pathologies), the proposed algorithms are demonstrated to be able to quantify degrees of structural complexity in the context of self-correlation over small to large temporal scales, thus offering physically meaningful interpretations and rigor in the understanding the intrinsic properties of the structural complexity of a system, such as the number of its degrees of freedom. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

2659 KiB

Open AccessArticle

Parametric PET Image Reconstruction via Regional Spatial Bases and Pharmacokinetic Time Activity Model

by Naoki Kawamura, Tatsuya Yokota, Hidekata Hontani, Muneyuki Sakata and Yuichi Kimura

Entropy 2017, 19(11), 629; https://doi.org/10.3390/e19110629 - 22 Nov 2017

Cited by 2 | Viewed by 4410

Abstract

It is known that the process of reconstruction of a Positron Emission Tomography (PET) image from sinogram data is very sensitive to measurement noises; it is still an important research topic to reconstruct PET images with high signal-to-noise ratios. In this paper, we [...] Read more.

It is known that the process of reconstruction of a Positron Emission Tomography (PET) image from sinogram data is very sensitive to measurement noises; it is still an important research topic to reconstruct PET images with high signal-to-noise ratios. In this paper, we propose a new reconstruction method for a temporal series of PET images from a temporal series of sinogram data. In the proposed method, PET images are reconstructed by minimizing the Kullback–Leibler divergence between the observed sinogram data and sinogram data derived from a parametric model of PET images. The contributions of the proposition include the following: (1) regions of targets in images are explicitly expressed using a set of spatial bases in order to ignore the noises in the background; (2) a parametric time activity model of PET images is explicitly introduced as a constraint; and (3) an algorithm for solving the optimization problem is clearly described. To demonstrate the advantages of the proposed method, quantitative evaluations are performed using both synthetic and clinical data of human brains. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

1209 KiB

Open AccessArticle

Re-Evaluating Electromyogram–Force Relation in Healthy Biceps Brachii Muscles Using Complexity Measures

by Xiaofei Zhu, Xu Zhang, Xiao Tang, Xiaoping Gao and Xiang Chen

Entropy 2017, 19(11), 624; https://doi.org/10.3390/e19110624 - 19 Nov 2017

Cited by 12 | Viewed by 6240

Abstract

The objective of this study is to re-evaluate the relation between surface electromyogram (EMG) and muscle contraction torque in biceps brachii (BB) muscles of healthy subjects using two different complexity measures. Ten healthy subjects were recruited and asked to complete a series of [...] Read more.

The objective of this study is to re-evaluate the relation between surface electromyogram (EMG) and muscle contraction torque in biceps brachii (BB) muscles of healthy subjects using two different complexity measures. Ten healthy subjects were recruited and asked to complete a series of elbow flexion tasks following different isometric muscle contraction levels ranging from 10% to 80% of maximum voluntary contraction (MVC) with each increment of 10%. Meanwhile, both the elbow flexion torque and surface EMG data from the muscle were recorded. The root mean square (RMS), sample entropy (SampEn) and fuzzy entropy (FuzzyEn) of corresponding EMG data were analyzed for each contraction level, and the relation between EMG and muscle torque was accordingly quantified. The experimental results showed a nonlinear relation between the traditional RMS amplitude of EMG and the muscle torque. By contrast, the FuzzyEn of EMG exhibited an improved linear correlation with the muscle torque than the RMS amplitude of EMG, which indicates its great value in estimating BB muscle strength in a simple and straightforward manner. In addition, the SampEn of EMG was found to be insensitive to the varying muscle torques, almost presenting a flat trend with the increment of muscle force. Such a character of the SampEn implied its potential application as a promising surface EMG biomarker for examining neuromuscular changes while overcoming interference from muscle strength. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

2071 KiB

Open AccessArticle

Multiscale Sample Entropy of Cardiovascular Signals: Does the Choice between Fixed- or Varying-Tolerance among Scales Influence Its Evaluation and Interpretation?

by Paolo Castiglioni, Paolo Coruzzi, Matteo Bini, Gianfranco Parati and Andrea Faini

Entropy 2017, 19(11), 590; https://doi.org/10.3390/e19110590 - 4 Nov 2017

Cited by 22 | Viewed by 5131

Abstract

Multiscale entropy (MSE) quantifies the cardiovascular complexity evaluating Sample Entropy (SampEn) on coarse-grained series at increasing scales ?. Two approaches exist, one using a fixed tolerance r at all scales (MSE_FT), the other a varying tolerance r(? [...] Read more.

Multiscale entropy (MSE) quantifies the cardiovascular complexity evaluating Sample Entropy (SampEn) on coarse-grained series at increasing scales ?. Two approaches exist, one using a fixed tolerance r at all scales (MSE_FT), the other a varying tolerance r(?) adjusted following the standard-deviation changes after coarse graining (MSE_VT). The aim of this study is to clarify how the choice between MSE_FT and MSE_VT influences quantification and interpretation of cardiovascular MSE, and whether it affects some signals more than others. To achieve this aim, we considered 2-h long beat-by-beat recordings of inter-beat intervals and of systolic and diastolic blood pressures in male (N = 42) and female (N = 42) healthy volunteers. We compared MSE estimated with fixed and varying tolerances, and evaluated whether the choice between MSE_FT and MSE_VT estimators influence quantification and interpretation of sex-related differences. We found substantial discrepancies between MSE_FT and MSE_VT results, related to the degree of correlation among samples and more important for heart rate than for blood pressure; moreover the choice between MSE_FT and MSE_VT may influence the interpretation of gender differences for MSE of heart rate. We conclude that studies on cardiovascular complexity should carefully choose between fixed- or varying-tolerance estimators, particularly when evaluating MSE of heart rate. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

808 KiB

Open AccessArticle

Challenging Recently Published Parameter Sets for Entropy Measures in Risk Prediction for End-Stage Renal Disease Patients

by Stefan Hagmair, Martin Bachler, Matthias C. Braunisch, Georg Lorenz, Christoph Schmaderer, Anna-Lena Hasenau, Lukas Von Stülpnagel, Axel Bauer, Kostantinos D. Rizas, Siegfried Wassertheurer and Christopher C. Mayer

Entropy 2017, 19(11), 582; https://doi.org/10.3390/e19110582 - 31 Oct 2017

Cited by 2 | Viewed by 4529

Abstract

Heart rate variability (HRV) analysis is a non-invasive tool for assessing cardiac health. Entropy measures quantify the chaotic properties of HRV, but they are sensitive to the choice of their required parameters. Previous studies therefore have performed parameter optimization, targeting solely their particular [...] Read more.

Heart rate variability (HRV) analysis is a non-invasive tool for assessing cardiac health. Entropy measures quantify the chaotic properties of HRV, but they are sensitive to the choice of their required parameters. Previous studies therefore have performed parameter optimization, targeting solely their particular patient cohort. In contrast, this work aimed to challenge entropy measures with recently published parameter sets, without time-consuming optimization, for risk prediction in end-stage renal disease patients. Approximate entropy, sample entropy, fuzzy entropy, fuzzy measure entropy, and corrected approximate entropy were examined. In total, 265 hemodialysis patients from the ISAR (rISk strAtification in end-stage Renal disease) study were analyzed. Throughout a median follow-up time of 43 months, 70 patients died. Fuzzy entropy and corrected approximate entropy (CApEn) provided significant hazard ratios, which remained significant after adjustment for clinical risk factors from literature if an entropy maximizing threshold parameter was chosen. Revealing results were seen in the subgroup of patients with heart disease (HD) when setting the radius to a multiple of the data’s standard deviation (

r = 0.2 \cdot ?

); all entropies, except CApEn, predicted mortality significantly and remained significant after adjustment. Therefore, these two parameter settings seem to reflect different cardiac properties. This work shows the potential of entropy measures for cardiovascular risk stratification in cohorts the parameters were not optimized for, and it provides additional insights into the parameter choice. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

856 KiB

Open AccessFeature PaperArticle

Entropy of Entropy: Measurement of Dynamical Complexity for Biological Systems

by Chang Francis Hsu, Sung-Yang Wei, Han-Ping Huang, Long Hsu, Sien Chi and Chung-Kang Peng

Entropy 2017, 19(10), 550; https://doi.org/10.3390/e19100550 - 18 Oct 2017

Cited by 35 | Viewed by 7947

Abstract

Healthy systems exhibit complex dynamics on the changing of information embedded in physiologic signals on multiple time scales that can be quantified by employing multiscale entropy (MSE) analysis. Here, we propose a measure of complexity, called entropy of entropy (EoE) analysis. The analysis [...] Read more.

Healthy systems exhibit complex dynamics on the changing of information embedded in physiologic signals on multiple time scales that can be quantified by employing multiscale entropy (MSE) analysis. Here, we propose a measure of complexity, called entropy of entropy (EoE) analysis. The analysis combines the features of MSE and an alternate measure of information, called superinformation, useful for DNA sequences. In this work, we apply the hybrid analysis to the cardiac interbeat interval time series. We find that the EoE value is significantly higher for the healthy than the pathologic groups. Particularly, short time series of 70 heart beats is sufficient for EoE analysis with an accuracy of 81% and longer series of 500 beats results in an accuracy of 90%. In addition, the EoE versus Shannon entropy plot of heart rate time series exhibits an inverted U relationship with the maximal EoE value appearing in the middle of extreme order and disorder. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

2554 KiB

Open AccessArticle

A Permutation Disalignment Index-Based Complex Network Approach to Evaluate Longitudinal Changes in Brain-Electrical Connectivity

by Nadia Mammone, Simona De Salvo, Cosimo Ieracitano, Silvia Marino, Angela Marra, Francesco Corallo and Francesco C. Morabito

Entropy 2017, 19(10), 548; https://doi.org/10.3390/e19100548 - 17 Oct 2017

Cited by 19 | Viewed by 6005

Abstract

In the study of neurological disorders, Electroencephalographic (EEG) signal processing can provide valuable information because abnormalities in the interaction between neuron circuits may reflect on macroscopic abnormalities in the electrical potentials that can be detected on the scalp. A Mild Cognitive Impairment (MCI) [...] Read more.

In the study of neurological disorders, Electroencephalographic (EEG) signal processing can provide valuable information because abnormalities in the interaction between neuron circuits may reflect on macroscopic abnormalities in the electrical potentials that can be detected on the scalp. A Mild Cognitive Impairment (MCI) condition, when caused by a disorder degenerating into dementia, affects the brain connectivity. Motivated by the promising results achieved through the recently developed descriptor of coupling strength between EEG signals, the Permutation Disalignment Index (PDI), the present paper introduces a novel PDI-based complex network model to evaluate the longitudinal variations in brain-electrical connectivity. A group of 33 amnestic MCI subjects was enrolled and followed-up with over four months. The results were compared to MoCA (Montreal Cognitive Assessment) tests, which scores the cognitive abilities of the patient. A significant negative correlation could be observed between MoCA variation and the characteristic path length (

?

) variation (

r = - 0.56, p = 0.0006

), whereas a significant positive correlation could be observed between MoCA variation and the variation of clustering coefficient (CC,

r = 0.58, p = 0.0004

), global efficiency (GE,

r = 0.57, p = 0.0005

) and small worldness (SW,

r = 0.57, p = 0.0005

). Cognitive decline thus seems to reflect an underlying cortical “disconnection” phenomenon: worsened subjects indeed showed an increased

?

and decreased CC, GE and SW. The PDI-based connectivity model, proposed in the present work, could be a novel tool for the objective quantification of longitudinal brain-electrical connectivity changes in MCI subjects. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

889 KiB

Open AccessArticle

Automatic Epileptic Seizure Detection in EEG Signals Using Multi-Domain Feature Extraction and Nonlinear Analysis

by Lina Wang, Weining Xue, Yang Li, Meilin Luo, Jie Huang, Weigang Cui and Chao Huang

Entropy 2017, 19(6), 222; https://doi.org/10.3390/e19060222 - 27 May 2017

Cited by 225 | Viewed by 15466

Abstract

Epileptic seizure detection is commonly implemented by expert clinicians with visual observation of electroencephalography (EEG) signals, which tends to be time consuming and sensitive to bias. The epileptic detection in most previous research suffers from low power and unsuitability for processing large datasets. [...] Read more.

Epileptic seizure detection is commonly implemented by expert clinicians with visual observation of electroencephalography (EEG) signals, which tends to be time consuming and sensitive to bias. The epileptic detection in most previous research suffers from low power and unsuitability for processing large datasets. Therefore, a computerized epileptic seizure detection method is highly required to eradicate the aforementioned problems, expedite epilepsy research and aid medical professionals. In this work, we propose an automatic epilepsy diagnosis framework based on the combination of multi-domain feature extraction and nonlinear analysis of EEG signals. Firstly, EEG signals are pre-processed by using the wavelet threshold method to remove the artifacts. We then extract representative features in the time domain, frequency domain, time-frequency domain and nonlinear analysis features based on the information theory. These features are further extracted in five frequency sub-bands based on the clinical interest, and the dimension of the original feature space is then reduced by using both a principal component analysis and an analysis of variance. Furthermore, the optimal combination of the extracted features is identified and evaluated via different classifiers for the epileptic seizure detection of EEG signals. Finally, the performance of the proposed method is investigated by using a public EEG database at the University Hospital Bonn, Germany. Experimental results demonstrate that the proposed epileptic seizure detection method can achieve a high average accuracy of 99.25%, indicating a powerful method in the detection and classification of epileptic seizures. The proposed seizure detection scheme is thus hoped to eliminate the burden of expert clinicians when they are processing a large number of data by visual observation and to speed-up the epilepsy diagnosis. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

3987 KiB

Open AccessArticle

Investigation of the Intra- and Inter-Limb Muscle Coordination of Hands-and-Knees Crawling in Human Adults by Means of Muscle Synergy Analysis

by Xiang Chen, Xiaocong Niu, De Wu, Yi Yu and Xu Zhang

Entropy 2017, 19(5), 229; https://doi.org/10.3390/e19050229 - 17 May 2017

Cited by 24 | Viewed by 8173

Abstract

To investigate the intra- and inter-limb muscle coordination mechanism of human hands-and-knees crawling by means of muscle synergy analysis, surface electromyographic (sEMG) signals of 20 human adults were collected bilaterally from 32 limb related muscles during crawling with hands and knees at different [...] Read more.

To investigate the intra- and inter-limb muscle coordination mechanism of human hands-and-knees crawling by means of muscle synergy analysis, surface electromyographic (sEMG) signals of 20 human adults were collected bilaterally from 32 limb related muscles during crawling with hands and knees at different speeds. The nonnegative matrix factorization (NMF) algorithm was exerted on each limb to extract muscle synergies. The results showed that intra-limb coordination was relatively stable during human hands-and-knees crawling. Two synergies, one relating to the stance phase and the other relating to the swing phase, could be extracted from each limb during a crawling cycle. Synergy structures during different speeds kept good consistency, but the recruitment levels, durations, and phases of muscle synergies were adjusted to adapt the change of crawling speed. Furthermore, the ipsilateral phase lag (IPL) value which was used to depict the inter-limb coordination changed with crawling speed for most subjects, and subjects using the no-limb-pairing mode at low speed tended to adopt the trot-like mode or pace-like mode at high speed. The research results could be well explained by the two-level central pattern generator (CPG) model consisting of a half-center rhythm generator (RG) and a pattern formation (PF) circuit. This study sheds light on the underlying control mechanism of human crawling. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

2200 KiB

Open AccessArticle

Muscle Fatigue Analysis of the Deltoid during Three Head-Related Static Isometric Contraction Tasks

by Wenxiang Cui, Xiang Chen, Shuai Cao and Xu Zhang

Entropy 2017, 19(5), 221; https://doi.org/10.3390/e19050221 - 11 May 2017

Cited by 7 | Viewed by 7338

Abstract

This study aimed to investigate the fatiguing characteristics of muscle-tendon units (MTUs) within skeletal muscles during static isometric contraction tasks. The deltoid was selected as the target muscle and three head-related static isometric contraction tasks were designed to activate three heads of the [...] Read more.

This study aimed to investigate the fatiguing characteristics of muscle-tendon units (MTUs) within skeletal muscles during static isometric contraction tasks. The deltoid was selected as the target muscle and three head-related static isometric contraction tasks were designed to activate three heads of the deltoid in different modes. Nine male subjects participated in this study. Surface electromyography (SEMG) signals were collected synchronously from the three heads of the deltoid. The performances of five SEMG parameters, including root mean square (RMS), mean power frequency (MPF), the first coefficient of autoregressive model (ARC1), sample entropy (SE) and Higuchi’s fractal dimension (HFD), in quantification of fatigue, were evaluated in terms of sensitivity to variability ratio (SVR) and consistency firstly. Then, the HFD parameter was selected as the fatigue index for further muscle fatigue analysis. The experimental results demonstrated that the three deltoid heads presented different activation modes during three head-related fatiguing contractions. The fatiguing characteristics of the three heads were found to be task-dependent, and the heads kept in a relatively high activation level were more prone to fatigue. In addition, the differences in fatiguing rate between heads increased with the increase in load. The findings of this study can be helpful in better understanding the underlying neuromuscular control strategies of the central nervous system (CNS). Based on the results of this study, the CNS was thought to control the contraction of the deltoid by taking the three heads as functional units, but a certain synergy among heads might also exist to accomplish a contraction task. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Figure 1
Three head-related static isometric contraction tasks: (a) Task 1; (b) Task 2; (c) Task 3. Full article ">Figure 2
Placement of EMG sensors on the deltoid. Here “Ch” denotes the channel of EMG sensor. Full article ">Figure 3
MRMS of the three deltoid heads during three head-related tasks with different loads. The boxes and whiskers show the mean and standard deviation respectively for all nine subjects. Full article ">Figure 4
SVR indexes of five SEMG parameters for the anterior head (a), the lateral head (b) and the posterior head (c) during three isometric contraction tasks with different loads. The boxes and whiskers show the mean and standard deviation values for all nine subjects. Full article ">Figure 5
Proportional distribution of two classes, corresponding to the positive or negative characteristics of linear regression slopes for five SEMG parameters. (a) In the anterior head; (b) In the lateral head; (c) In the posterior head. Full article ">Figure 6
The linear fit of HFD against time (number of epochs) in the three heads of the deltoid under nine conditions for one subject. (a) Task1 and load = 1 kg; (b) Task1 and load = 2 kg; (c) Task1 and load = 3 kg; (d) Task2 and load = 1 kg; (e) Task2 and load = 2 kg; (f) Task2 and load = 3 kg; (g) Task3 and load = 1 kg; (h) Task3 and load = 2 kg; (i) Task3 and load = 3 kg. Full article ">Figure 7
Boxplot of linear regression slopes of HFD for all subjects under different conditions. Box plots show the results for all subjects, the middle line in each box plot represents the median value, and the whisker indicates the range. The bottom and top limits of each box reveal the interquartile range and black plus signs denote outliers. Full article ">

Review

Jump to: Research

29 pages, 1005 KiB

Open AccessFeature PaperEditor’s ChoiceReview

Information Theoretic Approaches for Motor-Imagery BCI Systems: Review and Experimental Comparison

by Rubén Martín-Clemente, Javier Olias, Deepa Beeta Thiyam, Andrzej Cichocki and Sergio Cruces

Entropy 2018, 20(1), 7; https://doi.org/10.3390/e20010007 - 2 Jan 2018

Cited by 28 | Viewed by 7311

Abstract

Brain computer interfaces (BCIs) have been attracting a great interest in recent years. The common spatial patterns (CSP) technique is a well-established approach to the spatial filtering of the electroencephalogram (EEG) data in BCI applications. Even though CSP was originally proposed from a [...] Read more.

Brain computer interfaces (BCIs) have been attracting a great interest in recent years. The common spatial patterns (CSP) technique is a well-established approach to the spatial filtering of the electroencephalogram (EEG) data in BCI applications. Even though CSP was originally proposed from a heuristic viewpoint, it can be also built on very strong foundations using information theory. This paper reviews the relationship between CSP and several information-theoretic approaches, including the Kullback–Leibler divergence, the Beta divergence and the Alpha-Beta log-det (AB-LD)divergence. We also revise other approaches based on the idea of selecting those features that are maximally informative about the class labels. The performance of all the methods will be also compared via experiments. Full article

(This article belongs to the Special Issue Information Theory Applied to Physiological Signals)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Information Theory Applied to Physiological Signals

Share This Special Issue

Special Issue Editors

Special Issue Information

Keywords

Benefits of Publishing in a Special Issue

Related Special Issue

Published Papers (19 papers)

Research

Review

Further Information

Guidelines

MDPI Initiatives

Follow MDPI