Abstract During the last few years, rich-club (RC) organization has been studied as a possible brain-connectivity organization model for large-scale brain networks. At the same time, empirical and simulated data of neurophysiological models have demonstrated the significant role of intra-frequency and inter-frequency coupling among distinct brain areas. The current study investigates further the importance of these couplings using recordings of resting-state magnetoencephalographic activity obtained from 30 mild traumatic brain injury (mTBI) subjects and 50 healthy controls. Intra-frequency and inter-frequency coupling modes are incorporated in a single graph to detect group differences within individual rich-club subnetworks (type I networks) and networks connecting RC nodes with the rest of the nodes (type II networks). Our results show a higher probability of inter-frequency coupling for (δ−γ 1 ), (δ−γ 2 ), (θ−β), (θ−γ 2 ), (α−γ 2 ), (γ 1 −γ 2 ) and intra-frequency coupling for (γ 1 −γ 1 ) and (δ−δ) for both type I and type II networks in the mTBI group. Additionally, mTBI and control subjects can be correctly classified with high accuracy (98.6%), whereas a general linear regression model can effectively predict the subject group using the ratio of type I and type II coupling in the (δ, θ), (δ, β), (δ, γ 1 ), and (δ, γ 2 ) frequency pairs. These findings support the presence of an RC organization simultaneously with dominant frequency interactions within a single functional graph. Our results demonstrate a hyperactivation of intrinsic RC networks in mTBI subjects compared to controls, which can be seen as a plausible compensatory mechanism for alternative frequency-dependent routes of information flow in mTBI subjects.
Transcranial direct current stimulation (tDCS) has been used to noninvasively reduce epileptic activity in focal epilepsy. In this proof-of-principle N-of-1 trial in a patient with refractory focal epilepsy in left frontal lobe, we propose distributed constrained maximum intensity (D-CMI) for individually targeted and optimized multi-channel (mc-) tDCS [[1]Khan A. Antonakakis M. Suntrup-Krüger S. Lencer R. Nitsche M.A. Paulus W. Gross J. Wolters C.H. Can individually targeted and optimized multi-channel tDCS outperform standard bipolar tDCS in stimulating the primary somatosensory cortex?.Brain Stimul. 2023; 16: 1-16Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar] to reduce epileptic activity. Combined electro- and magnetoencephalography (EMEG) source analysis in a realistic calibrated head model [[2]Antonakakis M. Schrader S. Aydin U. Khan A. Gross J. Zervakis M. et al.Inter-subject variability of skull conductivity and thickness in calibrated realistic head models.Neuroimage. 2020; 223117353PubMed Google Scholar] with modeled skull-burr holes (Fig. 1b) defines location and orientation of the target epileptogenic source. Converging evidence for this determination is achieved by retrospective identification of a cortical malformation in magnetic resonance imaging (MRI) and by successful EMEG-guided invasive EEG (iEEG). Our first goal was to contribute to the identification of the epileptogenic zone by means of EMEG source analysis of sub-averaged interictal spikes. Following the sub-averaging strategy [[3]Aydin Ü. Vorwerk J. Dümpelmann M. Küpper P. Kugel H. Heers M. et al.Combined EEG/MEG can outperform single modality EEG or MEG source reconstruction in presurgical epilepsy diagnosis.PLoS One. 2015; 10e0118753https://doi.org/10.1371/journal.pone.0118753Crossref Scopus (71) Google Scholar], we performed EMEG source reconstructions at the spike onset (Fig. 1a). We then co-registered the source reconstructed EMEG centroid to the patient's 3D-FLAIR MRI (Fig. 1c). It was then found that the cortex was thickened and the grey-white matter boundary was slightly blurred, suggesting a cortical malformation, which had not been detected beforehand (Fig. 1c, area in red circle). Due to the converging evidence of seizure semiology, EMEG source reconstruction at spike onset and the retrospectively suspected cortical malformation in the 3D-FLAIR MRI, video-EEG with invasive electrodes (iEEG) was performed in 2021 (Fig. 1d). iEEG implantation was guided by the EMEG centroid result at spike onset. iEEG then confirmed that the seizure onset zone is located in the area of the suspected cortical malformation. Second, we computed an individually D-CMI optimized 8-channel mc-tDCS montage [[1]Khan A. Antonakakis M. Suntrup-Krüger S. Lencer R. Nitsche M.A. Paulus W. Gross J. Wolters C.H. Can individually targeted and optimized multi-channel tDCS outperform standard bipolar tDCS in stimulating the primary somatosensory cortex?.Brain Stimul. 2023; 16: 1-16Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar] targeting the centroid of the EMEG reconstructed onset of the interictal discharges (Fig. 1e). Our finite element method (FEM) based field simulation showed that it leads to high current intensity (D-CMI: 0.3 A/m2/Active-sham: 0.04 A/m2) and directionality (D-CMI: 0.21 A/m2/Active-sham: 0.03 A/m2) in the target area and relatively low current intensity in the non-target regions (D-CMI: 0.09 A/m2/Active-sham: 0.01 A/m2). Third, we performed a double-blind sham-controlled EEG/mc-tDCS/EEG experiment for our patient with the goal to non-invasively reduce the epileptic activity. Our hypothesis was that the real stimulation using the D-CMI montage inhibits the irritative zone and thus reduces the frequency of IEDs in EEG's measured after stimulation when compared to those measured before, while sham stimulation does not. We stimulated the patient twice per day for 20 minutes each, with a pause of 20 min in between, because 2 × 20-min daily stimulation was found to be superior to the protocol using 20-min daily stimulation only [[4]Yang D. Wang Q. Xu C. Fang F. Fan J. Li L. Du Q. et al.Transcranial direct current stimulation reduces seizure frequency in patients with refractory focal epilepsy: a randomized, double-blind, sham-controlled, and three-arm parallel multicenter study.Brain Stimul. 2020; 13: 109-116https://doi.org/10.1016/j.brs.2019.09.006Abstract Full Text Full Text PDF PubMed Scopus (64) Google Scholar]. Due to the circadian rhythm of IEDs [[5]Kaufmann E. Hordt M. Lauseker M. Palm U. Noachtar S. Acute effects of spaced cathodal transcranial direct current stimulation in drug resistant focal epilepsies.Clin Neurophysiol. 2021; 132: 1444-1451https://doi.org/10.1016/j.clinph.2021.03.048Crossref PubMed Scopus (15) Google Scholar], we always stimulated between 11 and 12 a.m. Each stimulation condition was applied for 5 days in a single week, with a washout period of 5 weeks between the two stimulation weeks to avoid any interference or carry over effect [[6]Woods A.J. Antal A. Bikson M. Boggio P.S. Brunoni A.R. Celnik P. et al.A technical guide to tDCS, and related non-invasive brain stimulation tools.Clin Neurophysiol. 2016; 127: 1031-1048https://doi.org/10.1016/j.clinph.2015.11.012Crossref PubMed Scopus (902) Google Scholar]. One-hour blocks of EEG were measured directly before and after the stimulation block every day. Before the first stimulation in each stimulation week, 2 hours of EEG were measured for better baseline IED detection, while in the analysis, the marked IEDs in the two-hour block were directly averaged, unless otherwise stated. The goal of reducing epileptic activity non-invasively was successfully achieved when comparing the IEDs marked by the three epileptologists in the EEGs before and after stimulation. We find a statistically highly significant (p < 0.0001) reduction of IEDs in the D-CMI stimulation week, while this is not the case for Active-sham (p > 0.05) (Fig. 1f). On average across the three epileptologists, the percentage reduction of IEDs between Pre and Post stimulation for the five days of stimulation per condition are always higher for D-CMI (between +37% and +81%) (mean ± SD: 58 ± 19%) than for Sham (between −23% and +36%) (mean ± SD: 16 ± 26%), where in the latter it also varies between reduction and increase (Fig. 1f, in blue). When compared to [[5]Kaufmann E. Hordt M. Lauseker M. Palm U. Noachtar S. Acute effects of spaced cathodal transcranial direct current stimulation in drug resistant focal epilepsies.Clin Neurophysiol. 2021; 132: 1444-1451https://doi.org/10.1016/j.clinph.2021.03.048Crossref PubMed Scopus (15) Google Scholar], who reported reductions of IEDs of mean 30% with SD of ±21% (p = 0.001), we thus achieved a high effect size with our individualized therapy. The success of the optimization is dependent on the targeting accuracy with regard to both target location and target orientation as well as on the idiosyncrasies of the individual head volume conductor model [[1]Khan A. Antonakakis M. Suntrup-Krüger S. Lencer R. Nitsche M.A. Paulus W. Gross J. Wolters C.H. Can individually targeted and optimized multi-channel tDCS outperform standard bipolar tDCS in stimulating the primary somatosensory cortex?.Brain Stimul. 2023; 16: 1-16Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar]. Here, we targeted by combined EEG/MEG (EMEG) source analysis using the different sensitivity profiles and complementary information contained in both modalities [[7]Dassios G. Fokas A.S. Hadjiloizi D. On the complementarity of electroencephalography and magnetoencephalography.Inverse Probl. 2007; 23: 2541-2549https://doi.org/10.1088/0266-5611/23/6/016Crossref Scopus (41) Google Scholar,[8]Ebersole J.S. Wagner M. Relative yield of MEG and EEG spikes in simultaneous recordings.Journal of clinical neurophysiology official publication of the American Electroencephalographic Society. 2018; 35: 443-453https://doi.org/10.1097/WNP.0000000000000512Crossref PubMed Scopus (14) Google Scholar]. For EEG, MEG and mc-tDCS electromagnetic field modeling, we used FEM in a realistic anisotropic six-compartment head model calibrated for individual skull-conductivity [[1]Khan A. Antonakakis M. Suntrup-Krüger S. Lencer R. Nitsche M.A. Paulus W. Gross J. Wolters C.H. Can individually targeted and optimized multi-channel tDCS outperform standard bipolar tDCS in stimulating the primary somatosensory cortex?.Brain Stimul. 2023; 16: 1-16Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar,[2]Antonakakis M. Schrader S. Aydin U. Khan A. Gross J. Zervakis M. et al.Inter-subject variability of skull conductivity and thickness in calibrated realistic head models.Neuroimage. 2020; 223117353PubMed Google Scholar] and taking into account cranial burr holes from a previous invasive EEG [[9]Lau S. Güllmar D. Flemming L. Grayden D.B. Cook M.J. Wolters C.H. Haueisen J. Skull defects in finite element head models for source reconstruction from magnetoencephalography signals.Front Neurosci. 2016; 10: 1-15PubMed Google Scholar,[10]Datta A. Bikson M. Fregni F. Transcranial direct current stimulation in patients with skull defects and skull plates: high-resolution computational FEM study of factors altering cortical current flow.Neuroimage. 2010; 52: 1268-1278https://doi.org/10.1016/j.neuroimage.2010.04.252Crossref PubMed Scopus (162) Google Scholar] (Fig. 1b). To the best of our knowledge, this level of personalization in targeting, optimization and head modeling for non-invasive diagnosis and therapy of focal epilepsy has not yet been proposed and used before. The proposed procedure was well-tolerated by our patient and parameterized a now registered group clinical trial (Study registration number: DRKS00029384). A detailed description of our study can be found in the supplementary material. Fully anonymized data that support the findings of this study are available on request by the first author. The original data are not publicly available due to privacy and ethical restrictions. Marios Antonakakis: Data curation, Formal analysis, Funding acquisition, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing, Investigation. Fabian Kaiser: Data curation, Formal analysis, Investigation, Validation, Writing – review & editing. Stefan Rampp: Conceptualization, Funding acquisition, Supervision, Writing – review & editing, Investigation. Stjepana Kovac: Validation, Writing – review & editing, Investigation. Heinz Wiendl: Resources. Walter Stummer: Resources, Investigation. Joachim Gross: Resources, Writing – review & editing. Christoph Kellinghaus: Conceptualization, Writing – review & editing, Investigation. Maryam Khaleghi-Ghadiri: Writing – review & editing, Investigation. Gabriel Möddel: Conceptualization, Funding acquisition, Investigation, Validation, Writing – review & editing. Carsten H. Wolters: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing – original draft. None This work was supported by the Bundesministerium für Gesundheit (BMG) as project ZMI1-2521FSB006, under the frame of ERA PerMed as project ERAPERMED2020-227 PerEpi, by the German Research Foundation (DFG) through projects WO1425/7–1 and RA2062/1-1, by EU project ChildBrain (Marie Curie innovative training network, grant no. 641652), by DAAD project 57663920 and by the Onassis Scholarship Foundation. We acknowledge support from the Open Access Publication Fund of the University of Muenster. In addition, we thank Andreas Wollbrink and Christian Glatz for technical assistance and Luca-J. Bombardelli, Hildegard Deitermann, Juliana Gericks, Gabriele Kemper, Ute Trompeter, Pia Wenge and Karin Wilken for their help with the EEG/MEG/MRI and tDCS data collection. The following is the Supplementary data to this article. Download .docx (6.18 MB) Help with docx files Multimedia component 1
The provided dataset consists of two high-quality realistic head models and combined EEG/MEG data which can be used for state-of-the-art methods in brain research, such as modern finite element methods (FEM) to compute the EEG/MEG forward problems using the software toolbox DUNEuro (http://duneuro.org). A combined EEG/MEG dataset from a somatosensory experiment is provided (sep_sef.zip): Somatosensory evoked potentials (SEP) and fields (SEF) were elicited by stimulating the median nerve at the wrist of the right arm with monophasic square-wave electrical pulses with 0.5 ms duration. A random stimulus onset asynchrony between 350 and 450 ms was used and the strength was adjusted to invoke a clear movement of the thumb. The duration of the experiment was 10 minutes for a measurement of 1200 trials and data was acquired with a sampling rate of 1200 Hz and online low pass filtered at 300 Hz. An artifact reduction was achieved by reversing the polarity of the stimulation during the second half of the measurement. A 74-channel EEG (EASYCAP GmbH, Herrsching, Germany), for which the electrode positions were digitized using a Polhemus device (FASTRAK, Polhemus Incorporated, Colchester, Vermont, U.S.A.), and a whole-head MEG with 275 axial gradiometers and 29 reference coils (OMEGA2005, VSM MedTech Ltd., Canada) were used in the measurement. Ethics Statement: One healthy subject (49 years, male) participated in this study. The subject had no history of psychiatric or neurological disorders and had given written informed consent before the experiment. All procedures had been approved by the ethics committee of the University of Erlangen, Faculty of Medicine on 10.05.2011 (Ref. No. 4453). Additionally, two different advanced realistic head models are supplied, which both use a six-compartment segmentation from T1/T2-MRI of the test subject. They differentiate between scalp, skull compacta, skull spongiosa, cerebrospinal fluid (CSF) and gray and white matter tissue. One head model is a tetrahedral volumetric mesh (realistic_tet_mesh_6c.msh), while the other provides the geometric information by level-sets for each tissue boundary (realistic_levelsets_6c.zip). A detailed description of the construction of the tetrahedral mesh can be found here (subsection 2.3), the main steps are presented in the following. First, the MR images were co-registered and resampled so that the voxels of the anatomical data are cubic. Furthermore, the images were cut sufficiently below the skull of the participant. Subsequently, the segmentation of the T1w and T2w was performed in order to create six volumetric masks representing the six tissue compartments. The brain compartment was segmented via the FreeSurfer software. The remaining preprocessing and creation of the volumetric masks was entirely performed via routines available in FieldTrip. In particular, the scalp and skull segmentations were done via the spm12 software, embedded in FieldTrip. Once the masks were assembled, a volumetric tetrahedral mesh was created using the CGAL software embedded in iso2mesh, resulting in 885,214 nodes and 5,335,615 tetrahedrons. The mesh is provided in gmsh format, including information about the node positions, elements defined by their node indices, and labels for each element indicating the tissue compartment. For the construction of the unfitted head model, a six-compartment voxel segmentation was constructed based on the T1- and T2-weighted MR images, distinguishing between skin, skull compacta and spongiosa, CSF, gray and white matter using SPM12 via Fieldtrip, FSL and internal MATLAB routines. Surfaces were extracted from this voxel segmentation to distinguish between the different tissue compartments. In order to smooth the surfaces while sustaining the available information from the segmentation, we applied an anti-aliasing algorithm created for binary voxel images presented in (Whitaker, 2000). The resulting smoothed surfaces are represented as discrete level-set functions, i.e., by \(N^3\)-dimensional arrays (\(N\)=257), the value on each node indicates the signed distance to the respective surface.
Breast cancer is the most common cancer in women worldwide. It is the most frequently diagnosed cancer among women in 140 countries out of 184 reporting countries. Lesions of breast cancer are abnormal areas in the breast tissues. Various types of breast cancer lesions include (1) microcalcifications, (2) masses, (3) architectural distortion, and (4) bilateral asymmetry. Microcalcification can be classified as benign, malignant, and benign without a callback. In the present manuscript, we propose an automatic pipeline for the detection of various categories of microcalcification. We performed deep learning using convolution neural networks (CNNs) for the automatic detection and classification of all three categories of microcalcification. CNN was applied using four different optimizers (ADAM, ADAGrad, ADADelta, and RMSProp). The input images of a size of 299 × 299 × 3, with fully connected RELU and SoftMax output activation functions, were utilized in this study. The feature map was obtained using the pretrained InceptionResNetV2 model. The performance evaluation of our classification scheme was tested on a curated breast imaging subset of the DDSM mammogram dataset (CBIS–DDSM), and the results were expressed in terms of sensitivity, specificity, accuracy, and area under the curve (AUC). Our proposed classification scheme outperforms the ability of previously used deep learning approaches and classical machine learning schemes.
During sleep., breathing-related sleep disorders (BSD) are very probable to cause distortions on human health and even be life-threatening. Among the different types of BSD., apnea accounts for one of the most common. Many detection algorithms have been proposed for spotting and classifying apneas, using one feature or being designed for binary classification. Also, many proposed clinical setups for respiratory data acquisition are invasive, making the application to patients a non-trial task. In this study, we aim to propose an easy-to-apply and patient-friendly clinical setup with a BSD detection that utilizes a multi-feature classification scheme for binary (apnea, healthy), as well as multiple classes (healthy, central, mixed, and obstructive apneas and hypopneas). Our clinical setup includes a high-resolution microphone attached to the bed at a very close distance to the patient. Our multi-feature approach contains spectral, statistical, and symbolic-based characteristics of respiratory signals of five patients admitted for a first BSD diagnosis and assesses the performance of different classification algorithms iteratively. The results show a high classification performance ( $>$ 98% for binary and $>$ 84% for multi-class classification) for either classification scheme. A robust classification scheme is thus proposed, utilizing the entire content of the recorded respiratory signal. Such a classification scheme leads to a promising result towards the design of portable devices with multi-features for real-time detection of BSD.
Introduction One of the primary motivations for studying the human brain is to comprehend how external sensory input is processed and ultimately perceived by the brain. A good understanding of these processes can promote the identification of biomarkers for the diagnosis of various neurological disorders; it can also provide ways of evaluating therapeutic techniques. In this work, we seek the minimal requirements for identifying key stages of activity in the brain elicited by median nerve stimulation. Methods We have used a priori knowledge and applied a simple, linear, spatial filter on the electroencephalography and magnetoencephalography signals to identify the early responses in the thalamus and cortex evoked by short electrical stimulation of the median nerve at the wrist. The spatial filter is defined first from the average EEG and MEG signals and then refined using consistency selection rules across ST. The refined spatial filter is then applied to extract the timecourses of each ST in each targeted generator. These ST timecourses are studied through clustering to quantify the ST variability. The nature of ST connectivity between thalamic and cortical generators is then studied within each identified cluster using linear and non-linear algorithms with time delays to extract linked and directional activities. A novel combination of linear and non-linear methods provides in addition discrimination of influences as excitatory or inhibitory. Results Our method identifies two key aspects of the evoked response. Firstly, the early onset of activity in the thalamus and the somatosensory cortex, known as the P14 and P20 in EEG and the second M20 for MEG. Secondly, good estimates are obtained for the early timecourse of activity from these two areas. The results confirm the existence of variability in ST brain activations and reveal distinct and novel patterns of connectivity in different clusters. Discussion It has been demonstrated that we can extract new insights into stimulus processing without the use of computationally costly source reconstruction techniques which require assumptions and detailed modeling of the brain. Our methodology, thanks to its simplicity and minimal computational requirements, has the potential for real-time applications such as in neurofeedback systems and brain-computer interfaces.
Functional brain connectivity networks exhibit "small-world" characteristics and some of these networks follow a "rich-club" organization, whereby a few nodes of high connectivity (hubs) tend to connect more densely among themselves than to nodes of lower connectivity. Current study followed an "attack strategy" to compare the rich-club and small-world network organization models using Magnetoencephalo¬graphic (MEG) recordings from mild traumatic brain injury (mTBI) patients and neurologically intact controls to identify the topology that describes the underlying intrinsic brain network organization. We hypothesized that the reduction in global efficiency caused by an attack targeting a model's hubs would reveal the "true" underlying topological organization. Connectivity networks were estimated using mutual information as the basis for cross-frequency coupling. Our results revealed a prominent rich-club network organization for both groups. In particular, mTBI patients demonstrated hyper-synchronization among rich-club hubs compared to controls in the δ band and the δ-γ1, θ-γ1, and β-γ2 frequency pairs. Moreover, rich-club hubs in mTBI patients were overrepresented in right frontal brain areas, from θ to γ1 frequencies, and underrepresented in left occipital regions in the δ-β, δ-γ1, θ-β, and β-γ2 frequency pairs. These findings indicate that the rich-club organization of resting-state MEG, considering its role in information integration and its vulnerability to various disorders like mTBI, may have a significant predictive value in the development of reliable biomarkers to help the validation of the recovery from mTBI. Furthermore, the proposed approach might be used as a validation tool to assess patient recovery.
'/%3e%3cpath fill='%23d57800' d='m727.132 103.967-2.355.938-.923.368-2.854-.157-2.673 1.283-4.74 3.239-.209.15-1.767.461-1.641-.708-.932.424-.321 1.92-.887 1.379-1.092.767-3.783.387-3.19 1.645-4.15-1.111-1.928.755-4.1 3.692-2.095.992-.732-.042-2.778-.143-1.173.384-1.752 1.693-3.315.258-1.033.715-1.618 3.26-1.753 1.808-1 .212-1.087-.4-.817.144-.37 2.086-.793.874-2.219.764-1.773 1.477-1.556.8-2.084-.052h-.479l-2.076 1.146-3.9.342-1.624 1.549-.358.331-.758.458-1.24.74-.3.184-1.5-.267-1.75.693-.708-1.089-1.211.727-1.6.069-2-.889-1.522-.678-1.113.161-.387 1.94-.039.206-1.17 1.557-2.184 1.323-.3.387-2.485 3.208-4.722 4.966-3.993 2.028-4.139 1.259-2.9 2.38-7.6 3.767-11.995 5.973-2.453.786-3.52.562-6.278 2.312-5.539 1.671-.308.094-1.1.332-7.81 2.363-3.635-.384-2.153.782-5.605-.622-3.9.077-2.467.513-4.637 2.224-7.827 3.762-2.586 2.366-4.027 2.1-4.715 1.493.011-1.981-.028-.084-1.636.517-1.181.373-3.825.59h-2.128l-1.321-.384-.2.069-7.818 2.569-8.719.842-4.363 1.4-3.26-.052-2.038.606-4 .489-1.428-.381-.322-.088-11.75.51-5.372-.607-2.564.583-4.3-1.446-6.269-.727-1.465-.439-3.3-.978-1.491.859-1.116.119-2.667-.967-.96-.024-2.2.8-1.267-.325-1.177-.938-2.761-.411-1.682-1.414-9.4 1.246-2.551-.75-8.554-2.492-1.361.069-1.7 1.271-2.548.929-2.212.575-3.012.1-3.5-.918-3.606-1.748-1.333-.317-2.974.3-.91.082-5.466-2.932-7.042-4.4-4.778-2.3-1.794-.291-.241.853.89 2.8.3 2.774-.091 2.563-.049 1.587.446 1.425 1.725 1.912.719 1.92.526 5.284-.006 5.362-.839 8.551-.283 1.3-1 4.43-.914 4.069-3.546 10.833-.974 1.343-2.391 1.631-5.419 3.688-3.99 2.3-1.263.523-3.239.245-2.018-.1-2.513-1.361-2.467-.635-3.232-2.324-3.6-.744-3.941-2.278-1.026-1.233-2.146-.2-2.976-.893v-.005l-1.093-.327-.755-.22-3.8-.1-3.54-1.634-1.864-.45-2.547-.129-2.666 1.239-1.351.629-1.707-.678-1.235.077-1.4 2.115-.172.323-.964.51-1.2-.017-.939-.025-1.016.444-1.368.593-1.208.527-.59.255-.01-.01-.363.073-.645.112-.8.153-1.41-.824-.75-.442-1.257-.256-.677.489-.133 2.3-.618 1.378-1.961 1.734-1.993 1.755-1.389 2.047-3.676 8.39-2.3 3.385-.715.789-2.418 2.66-2.136 1.657-4.922 3.815-4.659 1.741-4.013.82-1.951.008-3.7-.492-3.323-1.02-3.65-2.282-4-3.217-5.794-5.724-.82-.493-.171-.114-1.825-1.167-1.354.035-.53.874-.29 1.263-.245 1.12-.664 6.367.049.369.349 2.622 4.4 6.15 1.358 3.267.223.373.908 1.525.691 1.163.069.118 1.5 4.488-1.057 2.534.534 1.7-1.336 1.037-.245 1.465 3.951 4.953.862 2.185-.963 3.029-1.536 1.671-.429.457.216 1.28 1.818 1.635 3.089 2.75 1.647 4.086 1.253 1.1 1.456-.31.81.618 1.228-.021 1.208 1.208.782.379 1.236.616.977 1.272.216 3.911 1.822 4.86.028 2.635.01.374 1.134 1.948.356 1.538-.6 4.069 1.445 1.306 1.567-.32 1.012.16 1.626 1.354 1.969 3 1.926-.282 1.222.589 5.9 5.334 1.339.676.074.042.963.483 1.2 1.208 1.856-1.232.206-.021 2.018-.178.886.6 1.7 1.149 2.234-.029 4.844 1.485 2.119.543 4.261 2.648 1.811 1.12 1.072.871 1.424 1.161 2.845 1.224 2.44.584 1.322.314 1 .384.025-.062h.014l-.024.063 5.284 2.023 2.862.668 2.648.975 1.927.7 1.414.023 2-1.682 1.937-.112 1.609.4 1.57-.263 2.485-1.531.485-.719 1.762-.839 5.944-.738 1.737.564 4.46-1.94 2.988.831 2.645-.938 5.779 1.284 1.536 1.081 1.717 1.965.21.009 1.912.037-.8 1.331 2.523 2.647 2.279 3.445.154.455 1.657 4.864 1.232 1.723.862 2.405.042 1.568-1.34 1.022-.22.583-.076.216.272.655.506-.276.963-.533 1.172-.182 2.066.22 1.364.143 2.154-1.069 1.427-.71 2.293.987 2.52-.028 1.058.471 3 1.358 1.529.144.565-.371.36-.856-.185-1L397 377.23l-2.531-2.938-1.036-1.51-.789-2.186-.244-2.796-.174-1.885.314-2.477.68-.884.5-1.968h.009l.129-.515 1.406-1.866 3.724-2.393 4-3.582 3.162-2.085 3.26-1.412.087-.4.363.146 8.292-2.819.137-.016 3.692-.643 24.92 1.755.924-.279v-.006l.83-2.33.447-.432.3-.282 2.74-1.206 1.281-.183 3.556.878 1.456.359 2.332-1.307 1.654.028 3.417-1.866 2.111.053.8-.321 4.112-2.955 3.961-1.135 1.03-.548.363-.2 4.06-2.144 2.579-1.981 2.226-1.183 2.438-.629 6.38-.449 1.283-2.138 2.744-.366 1.657-1.87 1.9-.771 1.359-1.958 1.139-1.641 1.591-.921 5.1-.32 6.045.749.8-.485 1.366-4.875 1.43-.915 3.851-5.612.013-2.045v-1.708l.633-2.432-.43-4.391.51-4.446 2.394-5.65 2.109-2.279 3.4-2.3 1.729-.753 2.449-.412v.001l.36-.061.624-.107 8.433-.116c.117-.087 3.067-.042 3.067-.042l2.812-.038 7.116 1.264.506.088 2.4.779 2.774 2.157 2.938 2.869.551.535 1.9.8.575.251 1.571-.251 2.237-1.165 1.688-1.427 2.3-1.257.056-.08 1.295-1.948.039-.04.028.007v-.011h-.022l.511-.775 4.447-1.733 5.23-.3.588-.318.327-.175 1.382-1.623 1.295-.01 3.135 1.134 2.2-.438 1.75.7 1.36-.161 2.565-.3 2.754 1.937 1.605.2 5.658 3.31.294.048.152.028.324.054.671-.2 1-.289.132-.039.126.175.643.907.882.126 1.044-1.393-.415-.632-.157-.23-2.021-.411-1.815-3.1 1.728-1.861-2.711-3.326-.45-.551-.153-.234-.937-1.423-4.746-7.235-6.182-4.935-2.165-1.732-.014-.01-.758-.6-3.3-3.358-2.358-3.13-.309-.674-.487-1.062-1.413-3.078-2.462-1.9-2.006-2.191-.1-.129-4.14-5.626-.68-.943-.978-.565-2.066-.014-.2-.2-.116-.121.07-.074.991-.985 1.105-.278.476-1.089-2.108-6.1-.014-.213-.157-1.95 1.718-8.855.22-1.079 2.928-5.958 1.814-1.486 1.98-3.892 1.7-2.488 1.625-1.582.284-.168 2.934-1.776 2.465-.347 2.348-.331 4.1 1.264 3.9-.076.761-.081 1.77-.2 2.911-.917 1.374-.848.713-1.093 1.487-4.987.478-1.587.917-1.724 5.306-6.108 4.011-3.71 9.019-6.613 4.415-2.639 2.254-1.347 20.126-8.946 5.23-5.3 2.725-2.758 4.821-3.235 5.641-2.311 4.854-3.842 1.06-1.334 1.336-4.3 1.113-.269 1-2.174.286-.618 3.961-2.792.448-.223 15.388-7.71 2.236.307 1.2-1.793 4.477-.751.824-.14 1.1-.551.939-1.6v-.316l.129-4.273.956-1.155.521-2.816.491-.65.513-.676 1.268-.891-.22-.531.032.031z'/%3e%3c/svg%3e)
'/%3e%3cuse xlink:href='%23a' transform='rotate(144 450 300)'/%3e%3cuse xlink:href='%23a' transform='rotate(216 450 300)'/%3e%3cuse xlink:href='%23a' transform='rotate(288 450 300)'/%3e%3c/svg%3e)
'%3e%3cpath fill='%23012169' d='M0 0v30h60V0z'/%3e%3cpath stroke='%23fff' stroke-width='6' d='m0 0 60 30m0-30L0 30'/%3e%3cpath stroke='%23C8102E' stroke-width='4' d='m0 0 60 30m0-30L0 30' clip-path='url(%23b)'/%3e%3cpath stroke='%23fff' stroke-width='10' d='M30 0v30M0 15h60'/%3e%3cpath stroke='%23C8102E' stroke-width='6' d='M30 0v30M0 15h60'/%3e%3c/g%3e%3c/svg%3e)
