Version 10.1 by marissadiazpier on 2023/10/04 16:03
Show last authors
1 == Download the [[WhitePaper PDF.>>attach:WhitePaper PDF.]] ==
2
3 ==== ====
4
5
6 ==== //This online document will be updated as new information is provided by the students.// ====
7
8 == “Young Researchers EBRAINS Workflows White Paper” ==
9
10 Christian Mata, Christian Stephan-Otto, Nalan Karunanayake, Igori Comarovschii, Niccolò Mattiello, 
11 Katia Djerroud, Nathaniel Adibuer, Aziz Ullah Khan, Claudia Bachmann, Alper Yegenoglu, Sandra Díaz, Marissa Díaz.
12
13 September 1, 2023
14
15
16 (% style="text-align: center;" %)
17 ==== Abstract ====
18
19 This white paper presents a collection of scientific workflows crafted by students under the guidance of the Scientific Liaison Unit (SLU) within the EBRAINS infrastructure. This document describes how they translated their research goals into structured workflows using a standardized process. The workflows showcased in this paper span varying levels of maturity, from abstract concepts to well-defined requirements within the scientific process. These young researchers effectively articulated their requirements and harnessed EBRAINS tools to construct scientific workflows. The objective of this white paper is not only to highlight the practical application of SLU-developed tools and methodologies but also to serve as a valuable resource for the EBRAINS community, offering insights into the process of defining and executing scientific goals through modular and traceable workflows.
20
21 (% style="text-align: center;" %)
22 ==== ====
23
24 (% style="text-align: center;" %)
25 ==== 1.Introduction ====
26
27 EBRAINS is the European infrastructure for brain research which consists of an ecosystem of software tools, hardware components and methods capable of interacting with each other. Neuroscience has become an extremely interdisciplinary field which requires tools, methods, procedures, standards and a common language to enable scientists to have a successful exchange of ideas and work together to address complex questions.
28
29 One of the most prominent characteristics of EBRAINS is its modularity and composability. EBRAINS enables users to describe their research journey in the form of workflows which cover different time spans, can be dynamic, recursive and easy to extend, share and reproduce. Tools and services are connected directly or indirectly in such a way that the output of one tool can serve as input to one or many other tools, allowing scientific work to be expressed as workflows. These workflows can be then executed manually or automatically, sequentially or in parallel and locally (in a workstation or local university cluster) or remotely (on dedicated cloud or supercomputing infrastructure) depending on the use case requirements. By doing this, EBRAINS promotes FAIR science and provides the neuroscience community a common space where research questions of different kinds can be addressed in a collaborative and secure manner.
30
31 In this white paper our intention is to illustrate a set of workflows that have been derived from different user groups. These are intended to serve as examples for the community and to help the EBRAINS users identify a process by which it is possible to define requirements to achieve a scientific goal and translate them into a chain of modular tasks which can be easily tracked, reproduced, analyzed and shared.
32
33 In this white paper we cover workflows at different levels of maturity. Some of them are in an abstract stage, where only a general goal or research idea has been identified and the first steps to create a working approach towards it have been taken. Other workflows are more mature and rely on well identified requirements within the scientific journey. These requirements can include: data querying and analysis, model generation, simulation, analysis of experimental and simulated data, visualization, machine learning, interaction with the neurorobotics or the medical informatics platforms, usage of large computing infrastructure, interactive simulations, emulation and simulation on neuromorphic hardware, storage, sharing, secure data transfer, etc.
34
35 The results section first presents workflows from young researchers who have limited or no experience with the EBRAINS ecosystem. The young researchers were guided by the scientific liaison unit of the human brain project and the technical coordination team to describe their requirements and use EBRAINS tools to create both executive and scientific workflows. An executive workflow is the one which describes the interactions between scientists, groups, communities and other stakeholders. A scientific workflow describes the steps to be taken to collect observations, generate hypotheses, design, implement and execute experiments of different kinds, analyze, validate and visualize the results, asses the hypotheses, disseminate and share the results in an interactive manner.
36
37 In a second part of the results we include the EBRAINS showcases, which represent the workflows with higher level of maturity. The EBRAINS showcases are examples of complex research questions which have been addressed using a combination of EBRAINS tools and services and which can be modified, extended or used as a template by end users.
38
39 * topics covered and associated workflows (aim to give the reader a shortcut to workflows that might be interesting to her/him)
40 * (((
41 workflows of different readiness levels:
42
43 1. early-stage workflows: the implementation of the workflow has not started yet or is still in the first phases. Here, mainly Young researchers contributed. If you find a workflow particularly interesting, please do not hesitate to contact the authors.
44 1. middle-stage workflows: workflows that are currently in the implementation process. Here, authors can share already first experiences with the implementation
45 1. late-stage workflows: workflows that have been almost completed. Here, we will also present some showcases of EBRAINS.
46 )))
47 * the general structure how the workflow are represented (include legend for the diagrams)
48 * (((
49 short list of tools and services with name, short description, link to material.[[image:legend_workflows.png||alt="Figure1. Legend Workflows"]]
50 )))
51
52 (% style="text-align:center" %)
53 [[image:Screen Shot 2023-09-11 at 17.57.38.png]]
54
55 (% style="text-align:center" %)
56 [[image:Screen Shot 2023-09-11 at 17.59.04.png||alt="Table 1: Tool overview. Tools are classified according to their field of application: ’Data and knowledge’, ’Atlas and image processing’, ’Community’, ’Data analysis, visualisation and validation’, ’Medical data analysis’, ’Model building and simulation’, ’Brain Inspired Technologies’ and ’High performance computing’. For each tool a short description is given and a link to more information."]]
57
58 (% class="small" %)Table 1: Tool overview. Tools are classified according to their field of application: ’Data and knowledge’, ’Atlas and image processing’, ’Community’, ’Data analysis, visualisation and validation’, Medical data analysis’, ’Model building and simulation’, ’Brain Inspired Technologies’ and ’High performance computing’. For each tool a short description is given and a link to more information.
59
60
61 (% style="text-align: center;" %)
62 ==== 2 Methods ====
63
64 It can be challenging for new EBRAINS users to determine how EBRAINS products can assist them in realizing their scientific projects. In order to help in this process, the Scientific Liaison Unit of EBRAINS (SLU) has developed a standard description of workflows, which allows them to map the scientific project to software and compute resources and plan out its implementation.
65
66 Aside from this support, the HBP also organizes outreach activities targeted at different scientific communities. In this regard, a special focus is placed on young researchers who wish to learn how to use EBRAINS.
67
68 In order to learn how to use EBRAINS, it is helpful to understand how other scientific projects have been translated into EBRAINS workflows. Also in this regard, HBP seeks to support EBRAINS users. Following this idea, we (the authors of this paper) investigated the possibility of writing a white paper that serves as a database of different workflows and making them available to other researchers. In order to collect such usecases, we decided to combine the outreach efforts of the education program and the student ambassadors of the HBP with SLU formalization and technical coordination (TC). A students’ workshop was chosen as the best means of collecting use-cases.
69
70 With the students’ workshop we wanted to accomplish the following goals:
71
72
73 * • Collecting different neuroscientific projects suitable for creating workflows in EBRAINS.
74 * Presenting the different EBRAINS tools and computing resources to the students.
75 * Teaching students how to create their own EBRAINS workflows.
76 * Supporting students in creating their first workflows.
77
78 Accordingly we structured the workflow as follows:
79
80 In the first part we introduced the different EBRAINS tools and services.
81
82 In the second part, we explained the concept of a workflow and how it applies to EBRAINS workflows. In this context, we introduced a workflow generation template in the form of a mural board (see Fig. 2). By using this template, all necessary information on the science side could be collected and translated into a workflow in a graph-like format. Based on this information, this paper’s workflow descriptions were created.
83
84 In the third part we asked students to present their scientific project idea. For this part we asked students to give a presentation prior to the workshop. There was no requirement for the project to have already produced any kind of results or to have used EBRAINS tools. There was only one thing that was important: presenting the scientific idea.
85
86 How to create EBRAINS workflows from these scientific projects was discussed after the presentations. Here, the mural board were of great assistance and helped the students and tutors to streamline. In this last part of the workshop, we reserved extra time for students to ask questions, seek assistance, and/or present and discuss their first Mural boards.
87
88 The workshop idea proved to be successful. During the 1,5 days of the workshop, we had 9 participants, of which 5 agreed to share their workflows in this paper.
89
90 This work was made possible by the following groups sharing their workflows, which were subsequently integrated into this work after a long post-processing period, which included many emails between young researchers and organizers.
91
92 * Research Center Sant Joan de Déu in Spain, Christian Mata and Christian Stephan-Otto (work- flow 2)
93 * Nalan Kraunanayake, a PhD in Biomedical Engineering and Prof. Dr. Stanislav S. Makhanov both working at Thammasat University in Thailand (workflow 3)
94 * Igori Comarovschii, Mickkel Vinding and Prof. Daniel Lundqvist from the Karolinska Institute as well as Pascal Helson for the KTH Royal Institute of Technology in Sweden (workflow 4)
95
96 • PhD student Niccolò Mattiello of Gerardo Biella’s group at the University of Pavia (workflow 5)
97
98 • Katia Djerround, a master student at the Algerian University of Science and Technology Houari Boumediene, Nathaniel Adibuer, a research assistant at the University of Ghana, and Aziz Ullah Khan, a professional engineer.
99
100
101 (% style="text-align:center" %)
102 [[image:Marissa Diaz_2023-03-02_14-43-52.png]]
103
104 (% class="small" %)Figure 2: Mural board presented to the students during the workshop with an invite to fill it out. Each field on the board had a short caption and a short explanation of what to fill in: Team: the team name, the members and and one sentence per each group member to introduce themselves.
105
106
107 Background: Explaining the main trends of the state of the art to address the research question the reasearchers are working on; Problem: Decomposing the main challenge into subproblems; Vision: Where does the research group want to go? What are dreams of the researchers for future researchers working in your field?; Impact: What difference will the researchers make? Who else will benefit from the designed workflow?; Solution elements: Brainstorm the elements which would be key to solve your problem; Challenges:What is missing to achieve the researcher’s vision and can they address this using EBRAINS tools?; Workflow: Associating solution elements to EBRAINS components, and ordering them in a workflow; the construction of diagrams make use of the symbols described in 1.
108
109 In the following chapter, the scientific projects and the corresponding workflows are explained. An overview of the tools used in these workshops is given in the table 1. Tools are categorized here according to their fields of application. A short description of each tool is provided, along with a link to a more detailed explanation.
110
111 (% style="text-align: center;" %)
112 ==== 3. Results ====
113
114 The cases, which we will present in more detail later in this section, are very widely spread in terms of content, data types they ingest, tools they are using, maturity level and background of the corresponding scientists.
115
116 One of the use-cases, for example, focuses on establishing a methodology to facilitate the creation of 6
117
118 brain atlases for a particular target group. Other studies investigate phenomena like contour grouping, oscillation in Parkinson’s Disease, cell and network properties of the entorhinal cortex, and the use of imaging signals, specifically the EEG, to improve brain-computer interfaces.
119
120 In order to help the reader to quickly identify the most relevant use cases, we will briefly sketch each use case in a couple of sentences. This will include a quick overview of the use case content, the data it treats, EBRAINS tools used, maturity level achieved so far and information on the research group.
121
122 subsection 3.2 The researchers at Research Center Sant Joan de Déu in Spain, Christian Mata and Christian Stephan-Otto, desire to create anatomical brain templates of specific human subgroups from images in the BIDS format. Among the tools they are using are the Knowledge Graph, Quick NII, and the Brain Atlas of EBRAINS.
123
124 subsection 3.3 In the third workflow, Nalan Kraunanayake, a PhD in Biomedical Engineering and Prof. Dr. Stanislav S. Makhanov both working at Thammasat University in Thailand, present a project in which contour grouping is investigated. Using robot simulations of visual neuro function, the group explores how local and global stimulus properties affect contour grouping. Here, EBRAINS tools and services such as the NRP, NEST and L2L support this research.
125
126 subsection 3.4 Igori Comarovschii, Mickkel Vinding and Prof. Daniel Lundqvist from the Karolinska Institute as well as Pascal Helson for the KTH Royal Institute of Technology in Sweden present a work- flow that investigates beta oscillations in Parkinson’s Disease and the possibility of DBS treatment. To this end, a multiscale model is constructed based on MEG data collected under different experimental conditions. In particular the effect of DBS should be studied in this model. Thus, the overall goal is to create a workflow that leads from raw MEG data to a prediction of how MEG activity is influenced by DBS. This is accomplished with the aid of BIDS manager, Knowledge Graph, TVB, TVB imaging pipeline, NEST, and L2L. [The maturity level of this project is probably already advanced but this is something we have to ask them]
127
128 subsection 3.5 PhD student Niccolò Mattiello of Gerardo Biella’s group at the University of Pavia is investigating the different functional interactions of heterogeneous neurons in the entorhinal and perirhinal cortex. With this aim in mind he first focuses on single neuron model properties, collecting and analysing the data with the help of Neo, Elephant, Elephant Vis, BluePyEfe and Neuro Tech Mesh and, in a subsequent step investigates how the different experimental results influence the neuron’s activity in a simulated environment using Arbor, Neuron, BluePyEfe, BluePyOpt, L2L, Validation Framework and Hippo Unit. In a second step a larger cortical network is considered. Here data analysis can be carried out via Nutil and VisuAlign, Ilastik and Neuro Tech Mesh. For the creation of a model of a large network and its simulation, Brain Scafold Builder, Arbor, Core Neuron ViSimple, Neuroscheme, BluePyOpt, L2L, and Network Unit are considered.
129
130 Worflow 6:Katia Djerround, a master student at the Algerian University of Science and Technology Houari Boumediene, Nathaniel Adibuer, a research assistant at the University of Ghana, and Aziz Ullah Khan, a professional engineer from X, aim to develop a systematic approach for utilizing EEG data in Brain Computer Interfaces (BCI). A GDPR compliant data storage location combined with a standard data processing procedure should be made possible using the Health data cloud and tools like Frites and Neo. As a second step, the potential of MEG data for steering a robot is examined using the NRP. Finally, the loop is closed by exploring the potential of EEG signals to interact with a robotic application via a TVB simulation embedded in a NRP environment. Currently, this project is in the planning stage.
131
132
133 (% style="text-align: center;" %)
134 **3.1 Workflow 1**
135
136 This workflow will be added and updated on the second version of this withepeper.
137
138 (% style="text-align: center;" %)
139 **3.2 Workflow 2**
140
141 3.2.1 Team
142
143 • Christian Mata - computer vision, PhD
144 • Christian Stephan-Otto - neuroimaging, PhD
145
146 3.2.2 Background
147
148 We want to design and manage a biobank of images for EBRAINS to generate anatomical brain templates representing a specific subpopulation.
149
150 3.2.3 Problem
151
152 There is a lack of standardized brain anatomical templates segregated by ages, gender, disease, etc. to quantify the deviation of different groups of patients in each brain region.
153
154 3.2.4 Vision
155
156 We live in a time in which data is generated at an enormous speed. Researchers see the potential contributions to human knowledge that could be attained from analysing such large, reliable databases. Our approach is to provide researchers and society in general with a tool to better characterize the anatomy of the human brain in all its diversity, ensuring more precise and robust neuroscientific findings.
157
158 3.2.5 Impact
159
160 3.2.6 Solution elements
161
162 One the one hand, we want to create a pipeline that allows to create a biobank of images available to all EBRAINS users. On the other hand, the pipeline should be able to generate anatomical templates representative of the subpopulation the user can choose from the biobank.
163
164 For example, the user can ask for a template for young women aged from 22 to 24, in which all the segmentations of the Multiscale Atlas of the human brain could be projected.
165
166 As mentioned before, being able to create a repository of cases validated by experts and creating a specific database to be able to access specific cases is a very efficient process. One of the main solutions is to be able to collect massive data in a practical and simple way for research projects or post-analysis.
167
168 3.2.7 Challenges
169
170 EBRAINS’ tools offer many solutions and the main challenge is to correctly choose a specific applica- tion. The tool that could be useful for this purpose is the Multiscale Atlas of the Human Brain.
171
172 For this reason, being able to create a repository of cases validated by experts and to create a specific database to be able to access specific cases represents one of the main objectives.
173
174 3.2.8 Workflow
175
176 A possible workflow is depicted in Figure 3 and outlined in the following:
177
178 1. (((
179 Use the ’BIDS manager’ to organize own images, upload those images to the common repository
180
181 and label them according to the subject’s characteristics (age/sex/pathology/. . .).
182 )))
183 1. The user can request a brain template representative of certain profiles (e.g. Rett syndrome patients, all female, between 10 and 12 years old). For this, the ANTs software could be used to create such template, after preparing the images using the alignment and resizing capabilities of the QuickNII tool (https:~/~/ebrains.eu/service/quint). Then, some of the results of the Human Brain Atlas such as segmentations may be transferred to the new template, which in turn is now accessible for other researchers.
184
185 Figure 3: Workflow of the BIDS databank.
186
187 3. Every new template could be part of a live-paper, which would describe the characteristics of the population and the methodology for creating it.
188
189 (% style="text-align: center;" %)
190 **3.3 Workflow 3**
191
192 3.3.1 Team
193
194 • Nalan Karunanayake
195
196 • Prof. Dr. Stanislav S. Makhanov
197
198 Institute: Biomedical Engineering Research Unit, Sirindhorn International Institute of Technology, Thammasat University, Thailand.
199
200 3.3.2 Background
201
202 We investigate how Contour Grouping can be performed using the neurorobotics platform (NRP) tools and associated modules. Furthermore, we develop a biologically plausible model inspired by the human visual system using machine learning, robotics (virtual) and physics-based simulations.
203
204 3.3.3 Problem
205
206 The perception of a shape depends on the organisation of its boundaries from local orientation edges into global continuous contours (contour grouping) [45, 46]. Computer scientists generally assume that grouping is based on local Gestalt principles such as proximity and good continuation [47, 48]. However, there are also reports that the global property of contour closure and surface features are involved in this process [49, 50, 51]. This raises the question: Is contour grouping completely insensitive to the global properties of the stimulus and dependent only on local features, or is it a mixture of interrelated local and global features? The human visual system is an integration of many subprocesses such as attention, eye movement, and prior knowledge. To study the relationship between shape perception and contour grouping, the neurorobotics platform (NRP) is very well suited as it has access to many expert vision modules [52] that can be integrated and tested under different virtual experiment conditions. With the NRP, we can integrate the virtual robot with various visual neuro functions [53] powered by NEST [54] to study brain activities to group contours and understand sensory processing through motor execution.
207
208 3.3.4 Vision
209
210 We want to build a biologically explainable model that can explain the human vision of shape perception. How do we perceive shapes, and on what factors is our perception of shapes based? The true mental model for understanding the environment, colour, depth, shapes, etc. Is learning an essential factor in human visual shape perception? Additionally, we aim to integrate robotics into the artificial vision models to verify the reliability and expand the feature set and strength of the frameworks.
211
212 3.3.5 Impact
213
214 Our approach facilitates implementation, and through collaboration, it is possible to share knowledge among peer researchers. That would make the job easier and more fun. Industrial research institutions can also benefit from this approach. Collaboration, pre-implemented modules, and reproducibility make the EBRAINS workflow very appealing and helpful. The integration of robotics in vision research and expansion of the research space in artificial vision models
215
216 3.3.6 Solution elements
217
218 1. Develop a computational (biologically explainable) experimental vision model using NRP to investigate contour grouping.
219 1. Use the NEST brain model with other modules (Python modules like PyTorch ) and a virtual environment with iCub.
220 1. Use NRPs modules such as the Retina module, the Salience module, and the Laminart model.
221 1. Benefit from the synchronization, scalability and reproducibility of NRPs to make the study flow straightforward.
222 1. Use Gazebo (or Netlogo) with NRP to construct physics-based simulation modules to study brain activities (neural dynamics) of contour grouping and understand sensory processing through motor execution.
223
224 3.3.7 Challenges
225
226 We want to integrate various modules from already implemented experiments and easily add or remove modules on the platform. With the inclusion of a virtual robot unit (iCub), the classic 2D visual research will be extended to 3D to include depth features into the study. Additionally, we would like to integrate the Gazebo simulation platform or the NetLogo multi-agent framework with NEST to investigate the dynamics of the neuronal activities of the contour grouping.
227
228 3.3.8 Workflow
229
230 After successfully implementing the proposed model in NEST on the NRP, we consider running the simulations on the neuromorphic hardware platform (SpiNNaker) using spiking neural networks to make the model more biologically plausible and implement it at the hardware level integrating neuro- morphic sensors and actuators.
231
232 (% style="text-align: center;" %)
233 **3.4. Workflow 4**
234
235 3.4.1 Team
236
237 Figure 4: zebo workflow
238
239 * IgoriComarovschii,AffiliatedtoresearchDepartmentofClinicalNeuroscience,igori.comarovschii@ki.se
240 * MikkelVinding,AffiliatedtoresearchDepartmentofClinicalNeuroscience,NatMEG,mikkel.vinding@ki.se
241 * Pascal Helson, NeuroLogic group KTH Royal Institute of Technology, Stockholm, pashel@kth.se
242 * (((
243 Daniel Lundqvist, Senior research specialist Associate Professor, Head of Neuro division, and coordinator at CIR - Centre for Imaging Research, daniel.lundqvist@ki.se
244
245 3.4.2 Background
246
247 To understand cortical beta burst dynamics alteration it is relevant to consider the beta oscillatory network including the cortex, basal ganglia, and subthalamic nucleus. Indeed beta burst dynamics such as individual bursts’ amplitude and duration as well as burst patterning are altered in Parkinson’s disease (PD) yet normalized by chronic subthalamic deep brain stimulation (cDBS). Modulation and normalization in turn lead to improved motor function at the behavioral level. (Pauls)
248
249 To facilitate a deep understanding of how these neuromodulatory methods and pharmacological treat- ments can shift the brain dynamics from a diseased brain to a healthy state resting-state MEG could be used. MEG beta burst dynamics has the potential not only as the biomarker of PD but also as the biomarker for treatment and disease follow-up.
250
251 Virtual deep brain stimulation (DBS) using The Virtual Brain multiscale co-simulation has the potential for DBS optimization and forecasting of individualized DBS treatment. Multiscale simulation of empirical data-derived cortical beta burst dynamics altered by Parkinson’s disease but normalized by DBS could shed light on mechanisms of PD. The validation of such a model could be the normalization of beta burst activity following virtual DBS.
252
253 3.4.3 Problem
254
255 We explore multiscale PD signatures via brain network reconstruction from empirical data for per- sonalized virtual brain models for adaptive DBS therapy. However, not all categories of the data discussed have the permit to be shared. Furthermore, not all categories of MEG data are accompanied by structural MRI.
256
257 3.4.4 Vision
258
259 Human Brain Project visions and values (https:~/~/community.ebrains.eu/_ideas/-MuqsAgs1AnL0PK2RVJF/ about)
260
261 Link introduces the 3D web application (https:~/~/yngvifrey.github.io/Stockholm/) highlighting HBP values and visions of integration of brain-related data to create digital twins, neuromorphic computing, spiking-based algorithms, services for sensitive data, implementation of data safety by design. Given values and vision provide a basis for brain simulation ecosystem need for multiscale simulation of data derived neuromarkers.
262 )))
263
264 Community development and collaboration through the EBRAINS Co-create AI and Digital Brain Health.
265
266 Support of Sustainable Development Goals
267
268 3.4.5 Impact
269
270 Our approach supports open science by using the EBRAINS FAIR MEG PD Database: ”The Swedish National Facility for Magnetoencephalography (NatMEG) Parkinson’s Disease Dataset” on EBRAINS. We explore the multiscale mechanism of neuromarkers from clinical data with simulation. Finally, we develop clinical analytical tools and improve of therapeutic interventions for Parkinson’s disease.
271
272 3.4.6 Solution elements
273
274 We describe two solutions, the empirical data and the multi-scale simulation. Empirical data
275 The Swedish National Facility for Magnetoencephalography at Karolinska Institute, Stockholm sam- pled the largest existing MEG database on PD. Part of this database-“The Swedish National Facility for Magnetoencephalography Parkinson’s Disease Dataset” is currently under EBRAINS curation. Moreover, together with our collaborators from Aalto University and Helsinki University Hospital, Helsinki, Finland we are updating and merging databases with the DBS PD study.
276
277 Thus we have empirical resting-state MEG data describing brain dynamics consisting of:
278
279 * Chronic subthalamic DBS therapy (cDBS) OFF, Medication OFF (12 hours after the last dopaminergic medication)
280 * cDBS OFF, Medication ON (normal medication)
281 * cDBS ON,Medication OFF (12 hours after the last dopaminergic medication)
282 * cDBS ON, Medication ON.(normal medication)
283 * (((
284 Healthy controls (age-matched)
285
286 Additionally, passive movements task-related brain dynamics will be included:
287
288 The Passive movements study aimed to measure cortical activity related to the processing of propri- oceptive signals affected by PD in comparison to healthy control. The passive movements task was performed with the help of pneumatic artificial muscles (PAM) to induce proprioceptive feedback by brief passive movements (200 ms) of the index finger every 3.5-4.0 seconds (Vinding et.al. 2019).
289
290 Experimental workflows to obtain the data are described in the following articles: • Resting-state MEG [55, 56, 57]
291 • Task-related MEG [58]
292
293 In addition to MEG structural MRI will be included. It should be noted that MRI data exists only in some categories of patients. For example, DBS-related resting-state MEG(for patients and control) is not accompanied by MRI data. In this case, we have only sensor-level data.
294
295 We do not have tractography to construct an individualized connectome as a structural base for the functional oscillatory network. What could be the solution in this case if we want to construct TVB models? Could we use atlas for example human connctome mm1?
296
297 Multiscale simulation
298
299 Individualized brain simulations could be performed by The Virtual Brain (TVB). Data could be integrated using Individual cortical geometry (surface-based simulation), and individual tractography (employing DTI) to estimate structural connectivity where every node in the TVB network represents a brain region and its dynamics are simulated with a mean-field neural mass model (e.g. WilsonCowan, Jansen and Rit). As an option focal stimulation could be introduced. Following numerical integration as an output Local Field Potentials or EEG, MEG, sEEG via forward solution. further, the model could be fitted to the empirical data.
300 )))
301
302 The recent approach extended TVB to multiscale adding fine-scale simulation spiking neurons in the basal ganglia-thalamic-cortical network. The use case is virtual DBS. The validation of such a model could be the normalization of beta burst activity following virtual DBS.[59]
303
304 Beta bursts could be simulated by next-generation neural mass models Wilson-Cowan type. As well as sustained oscillations the model may show bursting when pink noise (1/f) is added. Using excitatory- inhibitory pair and setting the parameter values close to the edge between stationary and oscillatory behavior noise can perturb the system into the oscillatory state and we can observe high amplitude activity bursts [60]
305
306 A recent approach implements inferring excitation/inhibition balance from resting-state MEG PD data ON and OFF medication and healthy controls using 1/f (pink noise) dynamics of aperiodic activity with Welch’s and FOOOF methods. 44 sources were labeled using the human connectome project mm1 atlas and estimated brain-wide distribution of 1/f exponent characteristic with a mean over time. Perspectives are the usage of 1/f exponent dynamics as a biomarker and a combination of data analysis and modeling as E-I balance is important for neuronal network models.(Helson, 2022) (ongoing work)
307
308 Multiscale co-simulation using TVB-NEST/Anarchy could provide knowledge on how electrical stimulus can perturb synchrony of spiking neuronal populations and shift basal ganglia in an asynchronous state and help to fine-tune DBS protocols.
309
310 3.4.7 Challenges
311
312 * BIDS Manager provided by EBRAINS workflows do not support MEG and only on Windows.
313 * BIDS is not yet interoperable with openMINDS.(ongoing work)
314 * There is no standardized electronic data capture that could help with data annotation for BIDS and curation.(ongoing work)
315 * (((
316 Finding appropriate tools for model fitting
317
318 3.4.8 Workflow
319 )))
320
321 * Organization of raw data and metadata into BIDS MEG.
322 * Integration of metadata according to openMINDS to connect it to EBRAINS Knowledge Graph
323 * Integrate data and metadata to be compatible with The Virtual Brain with the help of The Virtual Brain Ontology.
324 * Construction and fitting the multiscale model using for example TVB-NEST and Simulation of resting state beta burst dynamic and task-related beta rebound affected by PD
325 * Model validation For literature see: [55, 60]
326
327 (% style="text-align: center;" %)
328 **3.5. Workflow 5**
329
330 3.5.1 Team
331
332 The team I’m working with at the moment is the Gerardo Biella’s team at the University of Pavia. Among the team’s research projects one of them is studying the properties and the modulation of neurons in the perirhinal cortex. But since in this period they are mainly focussed in other projects and they don’t have funds for a PhD, in the following months I’ll find another lab where I’ll be able to carry on this project.
333
334
335 Figure 5: PD-MEG-Data-Modeling workflow
336
337 The techniques adopted in the laboratory are based on electrophysiological approaches, in particular, patch-clamp recordings performed in mouse brain slices. This is the main and most common technique in order to collect electrophysiological information about the single neurons that constitute the network of the cortical area of our interest. In order to collect also some anatomical data we perform confocal microscopy acquisition of the electrophysiologically recorded cells.
338
339 3.5.2 Background
340
341 The main research question I’m working on is to understand the neural mechanisms underlying the functions carried out by the enthorinal cortex (e.g. spatial navigation, time coding of events) and perirhinal cortex (e.g. multisensory integration, information gating towards hippocampus during memory formation). What I want to understand is how the heterogeneous populations of neurons (mainly from an electrophysiological perspective, but it could be also extended to a molecular point of view) located in those areas orchestrate in order to create structures that can perform the previously mentioned functions. In other words, how each of those neurons with its properties and its connections takes part to the process. In literature are reported a wealth of experimental data describing the electrophysiological behaviour of the different neural populations, their modulation by different neurotransmitters and their anatomical connections. These experimental data are collected thanks to electrophysiological recordings as patch-clamp recordings performed at a single neuron resolution in mouse/rat brain slices or extracellular recordings performed both in vivo and in brain slices. Other experiments commonly conducted are imaging experiments as confocal microscopy acquisitions and calcium imaging microscopy performed both in vivo and in brain slices. At last, some experiments involved in assessing the contribution of a certain neurotransmitter or a specific neural population are performed with behavioural tests. But what is missing is a clear vision on how all those aspects at a single neuron level contribute to the dynamics at a network level. What could help to achieve this wide perspective of the neural dynamics is a neural network that as a scaffold gathers all the experimental evidences at a single neuron level. That could be a key point to make a step forward in understanding how the brain performs those functions and at same point this kind of network could be a useful resource both for the design of new experiments and the development of new paradigms for bio-inspired machine learning and neuromorphic architectures. At the moment different projects aimed in developing neural networks are going on but none of them involves the previously mentioned cortical structures and that’s the reason why I’m interested in design and carry on this workflow. Similar projects carried on by the Human Brain Project can support the feasibility and the potentials of the project I’m proposing, in particular, I’m referring to all the projects focused to model the cerebellum (as the ones conducted by the Egido D’Angelo’s Lab at the University of Pavia). That projects proved to be very useful to make some step forward in better understanding the cerebellar functions and also develop new neurorobotic prototypes.
342
343 3.5.3 Problem
344
345 The projects require to face three main issues:
346
347 1. Gathering anatomical information concerning the distribution of the different neural population and their wiring
348 1. Obtaining computational models of each neuronal type that can be enough reliable to integrate the different processing properties of all its compartments, but at the same time enough light weighted to be implemented in a neural network with a conspicuous number of neurons. As a following step combine them to create a neural network model
349 1. Gathering additional experimental data for the network model validation
350
351 The first point will require the following daily activities:
352
353 * Gathering the already published anatomical/histological data from articles and atlases
354 * (((
355 Performing additional sets of experiments and analysis to refine the data we already possess. (And subsequently sharing them with the community)
356
357 The second point will require the following daily activities:
358 )))
359
360 * Gathering the already published electrophysiological data from articles and databases
361 * Perform additional electrophysiological experiments and analysis to refine the data we already possess
362 * Update the multicompartmental single neuron models already at our disposal with the experimental data we obtained. As final step it will be necessary to create a simplified version of this multicompartmental models
363 * (((
364 Create a neural network that integrates all the anatomical data and the updated single neuron models of the previous points
365
366 The third point will require the following daily activities:
367
368 • Collect and analyse experimental data (electrophysioloogical, behavioural, imaging data) that can be used to validate the network previously built
369
370 3.5.4 Vision
371
372 What I would like to achieve with the future projects I’ll be working on is to develop a model of perirhinal and entorhinal cortex that could let us visualize at a detailed cellular level the neural dynamics that are going on in those area both at rest and during specific tasks. To a far extent what I would like to achieve is to have a clear vision on how these networks of single ”computational” units brings to the emergence of the complex behaviours we experience in our daily lives in order to achieve a complete reductionist description of those phenomena
373
374 3.5.5 Impact
375
376 The main contribution of this project would be the advancement in understanding the mechanisms underlying processes as memory formation/ multisensory integration (and so the representation of the surrounding world)/ spatial navigation and time perception. But along with this theoretical achievement there would be also interesting practical contributions. The network models developed during this project could be used in other fields, for example in neurorobotics or in the development of new information processing paradigms that could be applied in machine learning algorithms or neuromorphic computing architectures. This kind of predictions can be considered feasible achievements since some neurorobotic projects already carried on by the Human Brain Project could benefit from this new type of network. In fact, some of these projects are developing robots able to perform multisensory integration tasks but at the moment they are relying on network based on abstract architectures. A network more similar to the real neural network of the cortical area involved in that task might lead to an optimization of the task performed by the robots
377 )))
378
379 3.5.6 Solution elements
380
381 Some solutions to the challenges and problems we’ll have to face could be:
382
383 Networking, data sharing and collaboration: To achieve the objective of a clear vision of the network dynamics of the previously mentioned cortical areas, the effort of different groups will be required. The contribution of experimental and computational groups is needed and to make it possible networking and collaboration are key points of the project. For this reason, tools as Notebooks, Live Papers and Knowledge Graphs will provide a useful help for data and knowledge sharing. The EBRAINS community as long as it will be active and it will expand/interact with other groups and communities can be a great resource for sharing and develop new collaborations.
384
385 Analysis optimization: something that could help not only to speed up the advancement in this specific project but experimental research in general would be the optimization of the analysis tools. Based on my personal experience the analysis of electrophysiological data still involves time-consuming steps that might be optimized. In order to achieve this goal I think it might be useful relying more in open access tools as the ones offered in the EBRAINS platform. Thanks to those tools and the opportunity to collaborate with the EBRAINS group and the other members of the community it would be possible to develop better performing tools.
386
387 Data similarity/Quantitative approximation: Using data from already better described areas might be a solution to cope with the lack of some experimental data duringthe early stages of the project and preventing the risk of delays in the development of computational models. For example, after appropriate evaluations and metanalysis some electrophysiological data describing hippocampal features shared with the neurons in the area of my interest might be used as preliminary data for the single neuron model development. In this way both the computational flow and the experimental flow can advance at the same time. Once the quantitative data are collected by the experimental flow they will the replace the ones that were used in the early stages.
388
389 Iterative workflow and intermediate deliverables achievement: Since the project can’t be carried out as a single huge project, in order to be sustainable an iterative step workflow design might be the key to achieve the final result. So instead of collecting all the experimental results and input them in the workflow all at once, a good solution might be input the experimental data concerning some specific feature and at the end of the process release an updated version of the neural network implementing some additional feature in comparison to the previous one and then repeat the process. For example, in the first cycle I would like to implement the ”computational” operations performed by the dendrites and the spatial distribution of the different neural populations. So I’ll achieve a first updated network and I could assess its contribution in improving performances in neurorobotics and ML algorithms and in better describing the activity in those cortical areas. Then in a second cycle I could repeat the process in order to implement in the network additional information concerning the synaptic connectivity of the network. On a third cycle I could implement additional data concerning the modulation of certain neurons by a certain neurotransmitter.
390
391 Model optimization: Since using detailed multicompartmental models as building blocks for the creation of a neural network leads to a model too heavy to be simulated in a reasonable amount of time, the development of an intermediate/simplified multicompartmental model can be a useful step in the workflow. For example, replacing detailed dendrites made of various compartment with a single node/compartment that summarize the signal computation exerted by dendrites. Another optimization that might be done is a code optimization, for example writing the code in a manner that could exploit more the computational performances of GPUs (e.g. by using NVIDIA GPU optimized libraries or the GeNN library for GPU enhanced Neural Networks) HPC and Neuromorhpic resourses: Another solution to the computational complexity and weight issues is the opportunity to use resources as Super Computers and Neuromorphic platforms to perform simulations of high number of neurons in a reasonable amount of time.
392
393 3.5.7 Challenges
394
395 The challenges that will be faced during this phase will be:
396
397 * Lack of some specific experimental data. As I mentioned in the background section there is already a large amount of data describing the behaviours and connections of the populations in my regions of interest, but there are still some aspects that require additional quantitative data in order to develop a model that reliably implements those features
398 * This aspect brings two additional challenges that must be faced, time consuming processing during data analysis and the risk of some bottle-necks in the process. The necessity to collect additional experimental data might cause a delay in the progression of the computational models development
399 * Developing a neural network made of detailed neurons require a considerable amount of compu- tational resources.
400 * (((
401 Time and Money. Not as technical as the previous challenges but as important as the other are two issues that every project has to face: the time required to advance across all the steps and achieve the final goal, and the the necessity of fundings to keep the workflow going on
402
403 3.5.8 Workflow
404
405 EBRAINS Tools that could be interesting for you:
406
407 Data and Models: https:~/~/neuromorpho.org/ provides cytological data of the entorhinal cortex Model Catalog: https:~/~/model-catalog.brainsimulation.eu/docs/
408 )))
409
410 Data analysis:-)NeuroFeatureExtract (https:~/~/ebrains-cls-interactive.github.io/) (Electrophysiological data)
411
412 * INutil and VisuAlign (QUINT: Workflow for Quantification and Spatial Analysis of Features in Histological Images From Rodent Brain - HBP Wiki (ebrains.eu) (Histological data)
413 * Ilastik (ilastik - Animal Tracking) (Behavioural data) Model creation and Simulation SNUDDA (https:~/~/github.com/Hjorthmedh/Snudda) BRAINS Scaffold Builder (https:~/~/github.com/dbbs- lab/bsb) Single Cell In SilicoExperiment Tool (BlueNaaS) (https:~/~/blue-naas-bsp-epfl.apps.hbp.eu/#/)
414 * Hippounit (GitHub - KaliLab/hippounit) (optimization of the model parameters)
415 * BluepyOpt (GitHub - BlueBrain/BluePyOpt: Blue Brain Python Optimisation Library) (opti- mization of the model parameters)
416
417 (% style="text-align: center;" %)
418 **3.6. Workflow 6**
419
420 3.6.1 Team
421
422 * Katia Djerround, Bachelor in Biochemistry and Master in Neurobiology.
423 * Nathaniel Adibuer, research and teaching assistant at the University of Ghana, Biomedical En- gineering department
424 * (((
425 Aziz Ullah Khan: professional Engineer, Master’s degree in Electrical Engineering
426
427 3.6.2 Background
428
429 In the recent years, study of the BCI (Brain Computer Interface) technology based on EEG (Elec- troencephalography) has become one of the important parts in the biomedicine engineering. With this technique, people can control equipment or do very basic communication through their brains instead of language or physical actions [61] Present-day BCIs determine the intent of the user from a variety of different electrophysiological signals. These signals include slow cortical potentials, P300 potentials, and mu or beta rhythms recorded from the scalp, and cortical neuronal activity recorded by implanted electrodes [62]. They are translated in real-time into commands that operate a computer display or other device.
430 )))
431 * (((
432 Figure 6: Enthorial cortex simulation model
433 )))
434
435 EEG is not the only non-invasive technique to study the brain fMRI, CT, PET, MEG and FNIRS are commonly used. As for the brain stimulation techniques, we can site the rTMS, CES, tDCS, tRNS and the RINCE. [63]
436
437 3.6.3 Problem
438
439 • Data collection: finding a place in which the sensitive data can be stored in a systematic way • Creating a data processing pipeline that fullfils the following criteria:
440
441 – It can be applied to raw data of different experimental conditions and data formats – Allows to include various analysis techniques (such as AI) in a modular way
442 – Allows to compare the results of different imaging modalities
443
444 • Creating a setup that allows to test, if the obtained knowledge (e.g. a trained AI) can improve BCI. Here probably a simulated brain-machine setting would be the first step to go
445
446 3.6.4 Vision
447
448 We want to contribute to the further development of BCIs by facilitating the research on understanding in how far internal representations can be accessed by brain imaging techniques. For that goal we want to develop an open-source library, in which data of different experiences are sorted according to different categories (such as experiment type, imaging modality, research outcome, further option of
449
450 18
451
452 how to process the data). Of course, also by making use of the library for our own research, we hope that we will identify mechanisms that will not help only BCIs but also brain-to-brain communication and of course also help to understand the brain better.
453
454 3.6.5 Impact
455
456 The differences which HBP will make within our project are:
457
458 * First you permitted us to create a network where every participant is adding up precious information
459 * Secondly using the EBRAINS compound which will help us in working on developing the project.
460 * (((
461 Surely we shouldn’t forget about the unparalleled opportunity to learn and expand our perspectives as scientists.
462
463 Who else will benefit from the designed workflow? All scientists which are interested in:
464 )))
465
466 * Understanding the hidden layers and functioning of the brain
467 * Controlling engines ( robots/devices) by ”the power of the force” = brainwaves
468 * Programming
469 * Analyzing data
470 * Biologist/ doctors which would like to work on regeneration
471 * Teachers and students, can use the data for modelling.
472
473 3.6.6 Solution elements
474
475 * Creating a team with multidisciplinary background
476 * Collaborating with universities, hospitals and laboratories which use these imaging devices for their higher accuracy
477 * Identifying suitable tools for data processing, analysis and simulation
478 * Developing procedures to handle different data formats
479 * (((
480 Using/creating a program which can regulate and validate the data obtained as the library will be open source
481
482 3.6.7 Challenges
483
484 The challenges that we encountered are as follow:
485 )))
486
487 1. An infrastructure to store and process all the data which we will collect from the experiments and the databases from all over the Europe (if possible world) which mean also capable of handling the different countries data regulation. Here the Human health data cloud (https:~/~/www.healthdatacloud.eu/) or future infrastructures would help
488 1. A consortium of experts that can validate the data (if possible this process should be automated. Here the EBRAINS Community space can help to find such experts.
489 1. Appropriate data analysis tools. Here the following tools could help: Frites, Neo, TVB
490
491 3.6.8 Workflow
492
493 The workflow is divided into different steps in order to achieve the following goals:
494
495 ~1. Creating a data base of imaging data either obtained in a BCI task or in tasks that required motor output (starting first with EEG data); Attached to this data base are data analysis tools that allow to analyse the data in a straight forward way.
496
497 19
498
499 1. Developing a virtual robotic environment that allow to test in how far this data can be used to steer a robot
500 1. Substituting the data used in step 2) through simulations. Investigating in how far data properties can be reproduced via simulation. And, investigating which features of the data are crucial for steering the robot
501 1. (((
502 Expanding 1-3) towards new data modalities. Hereby the simulation framework in step 3) can help understanding in how far the different image modalities are needed to fine tune BCI tasks or are redundant and do not add further knowledge.
503
504 Figure 7: BCI–EEG–framwork workflow
505 )))
506
507 4 Discussion
508
509 * summary: in this white paper we wanted to provide an overview of example workflow of different levels of activities which leverage the EBRAINS infrastructure to do innovative reasearch (of course, some more sentences of what we did, workflows covering areas a,b,c
510 * short paragraph about the workflows
511 * focussing on communalites of the workflows (data manipulation, integration of modelling cycles, producing some meaning output that can be shared with community, and other meta patterns
512 * impact of this work: highlighting the importance of young researchers on the codesign of the research infrastructures, we would like to invite more young researchers to integrate this metholo- gies of stating their scientific journeys in which they hopefully make use of EBRAINS tools and shape it, also highlight the template, having a methologolgy in a clear way is a path towards a common language between diciplines and also to open science and to have more standardized solution that are reproducible and more robust
513 * education related to brainsciences would benefit from integrating these type of concept and best practices, making visible to the students to the platforms available in science, students can integrate them in their science
514 * PIs should encourage the usage of platforms, in order to really make the best out of the European Commision to build this infrastructure the research groups should encourage the usage of these methods/tools/standards and even become part of the developer community which also helps for the longterm stability of the infrastructure itself as a community effort
515
516 • shortcommings: in these usecases we have only covered a subset of the capabilities of EBRAInS and the community is invited to explore/ extent these usecases with different tools and services inside and outside of the EBRAINS platform; there are other tools, services and platforms designed to research of the brain (e.g open brain, Patraig Gleeson) we have not made an explicit comparrison with the possibilities the platform offers, however these platforms can also host tools and services from EBRAINS; most of usecases are in the initial state, ideally the science will be implemented later; not all usecases covered in EBRAINS
517
518 z outlook at some point it would be interesting to have the possibility to derive the connection between the tools, not only in a diagram but with automatic links provided by the infrastructure; share the experience of the usecases and use this knowledge to improve the infrastructure; tracking system would be nice also to guarantee a more intense collaboration
519
520 (% style="text-align: center;" %)
521 ==== 5 References ====
522
523 Future work
524
525 1. [1] EBRAINS Knowledge Graph. [Online; accessed 22-November-2022]. url: https:~/~/kg.ebrains. eu/.
526 1. [2] Live Papers. [Online; accessed 22-November-2022]. url: https:~/~/ebrains.eu/service/live- papers/.
527 1. [3] BIDS Manager. [Online; accessed 22-November-2022]. url: https:~/~/github.com/Dynamap/ BIDS_Manager.
528 1. [4] ilastik:overview. [Online; accessed 22-November-2022]. url: https : / / www . ilastik . org / documentation/index.html.
529 1. [5] localizoom: Introduction. [Online; accessed 22-November-2022]. url: https:~/~/localizoom. readthedocs.io/en/latest/introduction.html.
530 1. [6] Multilevel Human Brain Atlas. [Online; accessed 22-November-2022]. url: https:~/~/ebrains. eu/service/human-brain-atlas.
531 1. [7] Nutil - Neuroimaging utilities. [Online; accessed 03-March-2023]. url: https:~/~/www.nitrc. org/projects/nutil/.
532 1. [8] QuickNII - Serial section aligner to volumetric atlases. [Online; accessed 06-March-2023]. url: https:~/~/www.nitrc.org/projects/quicknii/.
533 1. [9] Siibra Explorer User Documentation. [Online; accessed 22-November-2022]. url: https : / / siibra-explorer.readthedocs.io/en/latest/.
534 1. [10] QUINT: Workflow for Quantification and Spatial Analysis of Features in Histological Images From Rodent Brain. [Online; accessed 22-November-2022]. url: https:~/~/wiki.ebrains.eu/ bin/view/Collabs/quint.
535 1. [11] VisuAlign: Software for 2D image registration to 3D atlas. [Online; accessed 06-March-2023]. url: https:~/~/visualign.readthedocs.io/en/latest/.
536 1. [12] The Collaboratory. [Online; accessed 22-November-2022]. url: https:~/~/wiki.ebrains.eu/ bin/view/Collabs/the-collaboratory.
537 1. [13] EBRAINS Community. [Online; accessed 22-November-2022]. url: https:~/~/community.ebrains. eu/feed.
538 1. [14] Elephant - Electrophysiology Analysis Toolkit. [Online; accessed 22-November-2022]. url: https: ~/~/elephant.readthedocs.io/en/latest/index.html.
539 1. [15] Visualization of Elephant - Electrophysiology Analysis Toolkit. [Online; accessed 22-November- 2022]. url: https:~/~/viziphant.readthedocs.io/en/latest/.
540 1. [16] Frites. [Online; accessed 22-November-2022]. url: https:~/~/brainets.github.io/frites/.
541 1. (((
542 [17] Git repository: hippounit. [Online; accessed 22-November-2022]. url: https:~/~/github.com/
543
544 KaliLab/hippounit.
545 )))
546 1. (((
547 [18] HBP Validation Framework - Python Client: Documentation. [Online; accessed 22-November-
548
549 2022]. url: https:~/~/hbp-validation-client.readthedocs.io/en/master/.
550 )))
551 1. [19] neo. [Online; accessed 22-November-2022]. url: https:~/~/neo.readthedocs.io/en/stable/.
552
553 21
554
555 1. [20] NetworkUnit. [Online; accessed 22-November-2022]. url: https:~/~/github.com/INM-6/NetworkUnit.
556 1. (((
557 [21] Welcome to NeuroScheme’s documentation! [Online; accessed 22-November-2022]. url: https:
558
559 ~/~/neuroscheme-documentation.readthedocs.io/en/latest/.
560 )))
561 1. (((
562 [22] Welcome to NeuroTessMesh’s documentation! [Online; accessed 22-November-2022]. url: https:
563
564 ~/~/neurotessmesh-documentation.readthedocs.io/en/latest/.
565 )))
566 1. (((
567 [23] ViSimpl Introduction. [Online; accessed 22-November-2022]. url: https:~/~/viziphant.readthedocs.
568
569 io/en/latest/.
570 )))
571 1. (((
572 [24] The HIP - a platform for state-of-the-art processing and international sharing of HUMAN intrac-
573
574 erebral EEG data. [Online; accessed 22-November-2022]. url: https:~/~/www.humanbrainproject.
575
576 eu/en/medicine/human-intracerebral-eeg-platform/.
577 )))
578 1. (((
579 [25] Medical Informatics Platform. [Online; accessed 22-November-2022]. url: https:~/~/ebrains.
580
581 eu/service/medical-informatics-platform/.
582 )))
583 1. [26] Arbor. [Online; accessed 22-November-2022]. url: https:~/~/docs.arbor-sim.org/en/v0.5/.
584 1. (((
585 [27] The Brain Scaffold Builder. [Online; accessed 22-November-2022]. url: https:~/~/bsb.readthedocs.
586
587 io/en/latest/.
588 )))
589 1. (((
590 [28] BluePyOpt. [Online; accessed 22-November-2022]. url: https:~/~/github.com/BlueBrain/
591
592 BluePyOpt.
593 )))
594 1. (((
595 [29] BluePyMM. [Online; accessed 22-November-2022]. url: https:~/~/bluepymm.readthedocs.io/
596
597 en/latest/index.html.
598 )))
599 1. [30] eFEL. [Online; accessed 22-November-2022]. url: https:~/~/github.com/BlueBrain/eFEL.
600 1. (((
601 [31] BluePyEfe. [Online; accessed 22-November-2022]. url: https:~/~/github.com/BlueBrain/
602
603 BluePyEfe.
604 )))
605 1. (((
606 [32] L2L. [Online; accessed 22-November-2022]. url: https:~/~/meta-optimization.github.io/
607
608 L2L/.
609 )))
610 1. (((
611 [33] Welcome to the NEST simulator documentation! [Online; accessed 22-November-2022]. url:
612
613 https:~/~/nest-simulator.readthedocs.io/en/v3.3/.
614 )))
615 1. (((
616 [34] Neuron:GUIDES AND TUTORIALS. [Online; accessed 22-November-2022]. url: https:~/~/www.
617
618 neuron.yale.edu/neuron/docs.
619 )))
620 1. (((
621 [35] CoreNEURON. [Online; accessed 22-November-2022]. url: https:~/~/github.com/BlueBrain/
622
623 CoreNeuron.
624 )))
625 1. (((
626 [36] THE VIRTUAL BRAIN: THE DOCUMENTATION WEBSITE. [Online; accessed 22-November-
627
628 2022]. url: https:~/~/docs.thevirtualbrain.org/index.html#.
629 )))
630 1. (((
631 [37] TVB image processing pipeline! [Online; accessed 22-November-2022]. url: https:~/~/gitlab.
632
633 ebrains.eu/tvb/fmriprep.
634 )))
635 1. (((
636 [38] BrainScaleS in EBRAINS. [Online; accessed 22-November-2022]. url: https:~/~/wiki.ebrains.
637
638 eu/bin/view/Collabs/neuromorphic/BrainScaleS/.
639 )))
640 1. (((
641 [39] HBP Neurorobotics guide book. [Online; accessed 22-November-2022]. url: https:~/~/neurorobotics.
642
643 net/Documentation/latest/index.html.
644 )))
645 1. (((
646 [40] Software for SpiNNaker. [Online; accessed 22-November-2022]. url: https:~/~/spinnakermanchester.
647
648 github.io/.
649 )))
650 1. (((
651 [41] Fenix- DOCUMENTATION. [Online; accessed 22-November-2022]. url: https:~/~/fenix-ri.
652
653 eu/about-fenix/documentation.
654 )))
655 1. (((
656 [42] Slurm – How–to–use instructions. [Online; accessed 22-November-2022]. url: https:~/~/wiki.
657
658 ebrains . eu / bin / view / Collabs / technical - coordination / EBRAINS % 5C % 20components /
659
660 Supercomputing%5C%20Proxy/?srid=Io86tXrJ.
661 )))
662 1. (((
663 [43] Slurm – How–to–use instructions. [Online; accessed 22-November-2022]. url: https:~/~/github.
664
665 com/HumanBrainProject/coallocation-slurm-plugin/blob/master/README.md.
666 )))
667 1. (((
668 [44] Unicore- DOCUMENTATION. [Online; accessed 22-November-2022]. url: https:~/~/www.unicore.
669
670 eu/documentation/.
671 )))
672 1. (((
673 [45] W. S. Geisler et al. “Edge co-occurrence in natural images predicts contour grouping perfor-
674
675 mance.” In: Vision research 41.6 (2001), pp. 711–724. doi: 10.1016/s0042-6989(00)00277-7.
676 )))
677 1. (((
678 [46] R.F. Hess, W.H. Beaudot, and K.T. Mullen. “Dynamics of contour integration.” In: Vision Re-
679
680 search 41.8 (2001), pp. 1023–1037.
681 )))
682 1. (((
683 [47] Jäkel, F.A. F.and Singh M.and Wichmann, and M.H. Herzog. “An overview of quantitative
684
685 approaches in Gestalt perception”. In: Vision Research 126 (2016), pp. 3–8.
686 )))
687
688 22
689
690 1. [48] S. Han, G.W. Humphreys, and L. Chen. “Uniform connectedness and classical Gestalt principles of perceptual grouping”. In: Perception & psychophysics 61.4 (1999), pp. 661–674.
691 1. [49] I. Kovacs and B Julesz. “A closed curve is much more than an incomplete one: effect of closure in figure-ground segmentation.” In: Proceedings of the National Academy of Sciences 90.16 (1993), pp. 7495–7497.
692 1. [50] T. Kwon et al. “Spatially–global integration of closed, fragmented contours by finding the shortest–path in a log-polar representation”. In: Vision research 126 (2016), pp. 143–163.
693 1. [51] B. Machilsen and J. Wagemans. “Integration of contour and surface information in shape detec- tion”. In: Vision Research 51.1 (2011), pp. 179–186.
694 1. [52] A. Bornet et al. “Running large-scale simulations on the Neurorobotics Platform to understand vision–the case of visual crowding”. In: Frontiers in neurorobotics 13 (2019), p. 33.
695 1. [53] E. Falotico et al. “Connecting artificial brains to robots in a comprehensive simulation framework: the neurorobotics platform”. In: Frontiers in neurorobotics 11 (2017), p. 2.
696 1. [54] M.O. Gewaltig and M. Diesmann. “Nest (neural simulation tool)”. In: Scholarpedia 2.4 (2007), p. 1430.
697 1. [55] KAM Pauls et al. “Cortical beta burst dynamics are altered in Parkinson’s disease but normalized by deep brain stimulation.” In: Neuroimage 257 (2022), p. 119308.
698 1. [56] Mikkel C Vinding et al. “Reduction of spontaneous cortical beta bursts in Parkinsonś disease is linked to symptom severity”. In: Brain Commun. 2.1 (2020), fcaa052.
699 1. [57] Mikkel C Vinding et al. “Changes in non-oscillatory features of the cortical sensorimotor rhythm in Parkinson’s disease across age”. In: (preprint) (2022).
700 1. [58] Mikkel C Vinding et al. “Attenuated beta rebound to proprioceptive afferent feedback in Parkin- sonś disease.” In: Sci Rep. 9.1 (2019), p. 2604.
701 1. [59] Jil M. Meier et al. “Virtual deep brain stimulation: Multiscale co-simulation of a spiking basal ganglia model and a whole-brain mean-field model with The Virtual Brain”. In: Experimental Neurology 354 (2022), p. 114111.
702 1. [60] Áine Byrne et al. “Next–generation neural mass and field modeling”. In: Journal of Neurophysi- ology 123.2 (2020), pp. 726–742.
703 1. [61] Jose del R. Wolpaw Jonathan R.and Millán and Nick F. Ramsey. “Chapter 2 – Brain–computer interfaces: Definitions and principles”. In: Elsevier – Handbook of Clinical Neurology 168 (2020), pp. 15–23.
704 1. [62] Yuanqing Li et al. “An EEG-Based BCI System for 2-D Cursor Control by Combining Mu/Beta Rhythm and P300 Potential”. In: IEEE Transactions on Biomedical Engineering 57.10 (2010), pp. 2495–2505.
705 1. [63] S. Ma T.and Wang et al. “CNN-based classification of fNIRS signals in motor imagery BCI system. Journal of neural engineering”. In: IJournal of neural engineering 18.5 (2021).
706
707 23