Projects
Third-party Funded Projects
In SmallData, we address data analysis and modeling in small data settings, i.e., when there is only little information in a dataset at hand, due to a small number of observations that carry relevant information, relative to the complexity of novel patterns to be uncovered or the level of heterogeneity across observations.
We focus on:
Similarity for pulling in additional data of the same type (Project Area A).
Transfer for transferring additional information to the dataset at hand, such as from data of different type (Project Area B).
Uncertainty for quantifying and reducing uncertainty in particular in similarity and transfer (Project Area C).
This is enabled by a joint methods framework, with a focus on combining knowledge-driven and data-driven modeling.
Funded by DFG
PI: Prof. Dr. Harald Binder
Learn more: https://www.smalldata-initiative.de/
The brain is the most complex information processing system that enables us to feel, act, learn and remember, to process and store information with an efficiency and flexibility that by far surpasses any machine. At the same time, dysfunctions of the brain cause a number of devastating cognitive disorders.
All cognitive functions depend on the cortex, the largest and most powerful region of the mammalian brain. Looking at neurons it comprises ~80% excitatory principal neurons (PNs) but also a smaller but highly diverse class of GABAergic inhibitory interneurons. PNs represent information at the single cell level (‘single neuron code’) or population level (‘population code’). However, their activity is markedly shaped by the GABAergic inhibitory interneurons. Interneurons come with distinct morphologies, molecular and synaptic properties, connectivities and activity profiles and may thereby endow the network with a large number of distinct control points for tuning cortical codes according to dynamically changing computational demands.
Using mouse models for genetic circuit dissection and human tissue we will focus on a number of cortical areas that are vital for higher brain functions. By combining state-of-the-art optical and electrical recordings with pharmaco- and optogenetic perturbation, quantitative behavior, computational modeling, modern tools for high-dimensional data analysis and deep learning, this CRC/TRR will provide multidisciplinary insights on the role of inhibition in controlling the encoding of information in local circuits and complex brain functions and thereby shaping the cortical population code.
IN-CODE is a Transregio (TRR), a Collaborative Research Centre (CRC) located at multiple sites. TRR/CRC-384 is located in Freiburg and Berlin, and collaborations between the two cities are actively promoted. The selected PhD students will enrol in the local graduate schools, SGBM (Spemann graduate school of Biology and Medicine) and ECN (Einstein Center for Neurosciences), respectively, and will have access to the qualification program offered by the entire network.
Funded by DFG
PI: Prof. Dr. Marlena Bartos (Speaker)
Learn more: https://physiology-freiburg.de/de/in-code-is-recruiting/
Can robots learn through observation? ReScaLe is researching new training methods for AI-based robots.
AI-based robots are expected to support numerous tasks in our society, for example by assisting people in everyday life or making production processes more efficient – however, despite rapid advances in research, they are not yet part of our everyday lives. In order to make them more integrable for everyday life, ReScaLe will on the one hand work on the still existing technical challenges in the field of machine learning. On the other hand, the project will also consider social, ethical and legal aspects in order to strengthen trust in these systems.
Innovative machine learning methods will enable ReScaLe robots to learn tasks from humans by demonstrating them to them. To enable robots to efficiently perform the learning task, ReScaLe is developing new approaches to minimize the number of demonstrations required. The research project will introduce novel unsupervised and self-supervised deep learning methods that require only a small amount of annotated data. Further innovative methods will also support deep learning in dealing with uncertainties to further improve data efficiency. ReScaLe will simultaneously pave the way for responsible AI and robotics applications based on human rights, taking an integrated multi-level approach that considers ethical-legal normative requirements in conjunction with risks to core rights and interests, as well as user*oriented design requirements. Specially tailored participatory outreach activities accompany the project to promote community acceptance and enable bidirectional communication with researchers.
Researchers from the fields of computer science, ethics, human-machine interaction, law, mathematics and robotics are participating in ReScaLe. The project strengthens the profile field “Data Analysis and Artificial Intelligence” of the University of Freiburg and will be located in the research building ‘Intelligent Machine-Brain Interfacing Technology’ (IMBIT).
Funded by Carl-Zeiss-Stiftung (Programme CZS Breakthroughs)
PIs: Dr. Noor Awad, Prof. Dr. Joschka Bödecker (Coordinator), Prof. Dr. Frank Hutter, Jun.-Prof. Dr. Philipp Kellmeyer, Prof. Dr. Oliver Müller, Prof. Dr. Thorsten Schmidt, Prof. Dr. Abhinav Valada, Prof. Dr. Silja Vöneky
Learn more: https://uni-freiburg.de/brainlinks-braintools/rescale/
The central theme is the development of state-of-the-art AI algorithms for autonomous driving, with a focus on performance, scalability, and reliability. In contrast to conventional approaches to automated driving, the focus will be on a networked framework in which different modules are integrated and optimized together to improve the robustness of the system. An example is “environmental perception and planning”, where the two tasks are optimized together to avoid overly conservative or overly aggressive behavior depending on the situation, which should lead to more human-like driving behavior of the autonomous vehicle.
The focus shifts from the opaque nature of deep learning models to a “white box” approach. Subcomponents within the system provide intermediate results that can be interpreted by humans, thus promoting trust and simplifying certification processes. In addition, innovative neural architecture search methods will be developed to simplify the design and optimization of network architectures.
In addition to the software prototypes, the expected results include the publication of numerous research papers in prestigious conferences and journals, which should significantly advance the field.
Funded by Bosch
PIs: Prof. Dr. Joschka Boedecker, Prof. Dr. Abhinav Valada (Coordinator), Prof. Dr. Frank Hutter
Lear more: https://www.bosch.com/research/news/collaboration-with-university-of-freiburg/
nxtAIM is a research project that encompasses all essential elements for successful AI development: comprehensive automotive data from industry and academia, powerful computing resources, technical expertise, and innovative foundation models to build upon.
The work of nxtAIM is organized in six sub-projects. Sub-projects 1, 2 and 3 focus on different parts of the chain of effects. Sub-project 1 aims to generate sensor data from the environment model for individual time steps. Sub-project 2 extends this to include the dynamic development of the situation and aims to create generative models for sequences of sensor data.
Generative models not only enable the sampling of sensor data, but also of time series in general. This is used in subproject 3 to generate traffic scenarios in geometrically abstracted environment models and thus improve prediction and planning algorithms.
Generative models are typically based on a latent feature space. Structuring and interpretation of this space is the task of sub-project 4. Sub-project 5 deals with how the new approaches can be implemented in a system that can be executed in the vehicle. Finally, sub-project 6 will evaluate the systems developed in sub-projects 1, 2 and 3 in terms of their degree of realism.
Funded by Federal Ministry for Economic Affairs and Climate Action
PI: Prof. Dr. Thomas Brox
Learn more: https://nxtaim.de/
A paraplegic patient controls a robotic arm with her thoughts. A “locked-in” patient creates letters on a screen using his imagination. Both wear a neuroimplant that reads brain activity via tiny electrodes. These examples show what neuroimplants can already do today.
Dr. Simon Binder’s project is dedicated to the development of novel neuroimplants with biohybrid electrodes. These implants use cultured cells to integrate into the brain tissue and create a living interface between the brain’s neurons and the readout electronics. One focus is the electrode design. It is based on hydrogels that mimic the mechanical properties of brain tissue. This improves biocompatibility and marks a further step towards potentially lifelong implant retention in the body. The biohybrid approach is also expected to provide a high spatio-temporal resolution of the measured brain activity. This is important for complex applications and new insights into the functioning of the brain.
Funded by Carl-Zeiss-Stiftung (Programme CZS Nexus)
PI: Dr. Simon Binder
Learn more: https://uni-freiburg.de/carl-zeiss-stiftung-fordert-nachwuchsgruppe-fur-biohybride-neuroimplantate/
The Imaging Memory and Consolidation Lab (IMaC Lab) studies human memory at the University of Freiburg, Germany. We are part of the Division Neuropsychology at the Institute of Psychology. The lab is headed by Dr. Monika Schönauer, Assistant Professor and Chair of Neuropsychology.
The IMaC Lab wants to gain a better understanding of how our brains form lasting memories. We address this question by studying memory consolidation, the processes occurring after encoding which support long-term memory storage. Our main research interest is how reactivation – via active memory rehearsal in wakefulness or by covert memory processing during sleep – strengthens and stabilizes new memory representations.
Funded by DFG Emmy Noether Research Group
PI: Jun.-Prof. Dr. Monika Schönauer
Learn more: https://imagingconsolidation.wordpress.com/
Machine learning and intelligent systems are the driving forces behind the transformation of artificial intelligence and its many subfields. These technologies enable the development of key innovations that are increasingly applied across a wide range of sectors, leading to profound economic and societal changes. The central question addressed by the Konrad Zuse School of Excellence in Learning and Intelligent Systems (ELIZA) is to explore the potential that learning systems already offer today, and how they should be further developed to meet future challenges.
Funded by DAAD
PIs: Prof. Dr. Thomas Brox, Prof. Dr. Frank Hutter, Prof. Dr. Abhinav Valada
Learn more: https://www.daad.de/de/der-daad/zuse-schools/zuse-school-eliza/
The Move2Treat project, funded by EIC Pathfinder Open, is at the forefront of groundbreaking research aimed at developing revolutionary technologies. This initiative supports the development of minimally invasive, soft multifunctional neural devices leading to improved mapping of brain and spinal cord neural activity. With a competitive success rate of just 6-7%, Move2Treat exemplifies a high-risk, high-gain endeavor, dedicated to realizing technological advancements that will significantly impact Europe and beyond.
Funded by EU
PI: Prof. Dr. Thomas Stieglitz
Learn more: https://move2treat.org
Context-dependent, flexible behavior is crucial for the survival of animals. In mice, the anterolateral motor cortex (ALM) is believed to play a key role in linking contextual information to movement planning. Context-dependent neural responses are observed primarily in cortical layer 2, whereas movement planning first emerges in layer 5. How these two processes are connected, how context- and movement-selective neuronal populations interact, and how context-selective activity influences movement planning remain unclear. Our main goal, therefore, is to gain a better understanding of how contextual information is transformed into movement planning within the ALM of mice.
Funded by DFG
PIs: Dr. Julian Ammer, Prof. Dr. Ilka Diester
Learn more: https://freidok.uni-freiburg.de/proj/10409
A-ROR aims to develop robotic technologies that can perform a wide range of tasks in arable farming. The goal is to enable the use of robots in an economically sustainable way in agriculture and exploit the advantages that come with autonomous systems. To achieve this, we are developing improved hardware technologies, sensors, and supporting algorithms for perception and localization. This is particularly necessary in adverse conditions in the field, such as heavily soaked and thus muddy soil. We are also developing a special safety sensor technology in conjunction with suitable control strategies for increased maneuverability and preventing unintentional failures of the system, while at the same time increasing functional safety.
Funded by Baden-Württemberg Stiftung
PI: Prof. Dr. Abhinav Valada
Learn more: https://rl.uni-freiburg.de/funded-projects
Advanced tissue engineering for organ repair
Shortage of donor organs or tissue grafts significantly limits the clinical outcome of many diseases as well as trauma. Tissue engineering efforts have made significant progress in recent years, but so far cannot recapitulate the full architecture and function of native tissues, except for avascular tissues such as cartilage and skin. Funded by the European Innovation Council, the THOR project aims to develop a new tissue engineering technology that can produce any type of human tissue for organ repair or replacement. The technology will use self-assembling molecules, polymeric fibres and functionalisation to fabricate patient-tailored tissues in fully automated production plants. The generated tissues/organs will be well vascularised and fully functioning, revolutionising the field of transplantation.
Funded by The European Innovation Council (Horizon-EIC-Pathfinder)
PI: Prof. Dr. Andreas Vlachos
Learn more: https://cordis.europa.eu/project/id/101099719
A state-of-the-art ecosystem for neuroscience.
EBRAINS is on a mission to revolutionise how neuroscience is conducted. The digital ecosystem that we provide enables advances in brain research that translate to innovations in neuroscience, healthcare and technology.
EBRAINS provides a digital research infrastructure that accelerates collaborative brain research between leading organizations and researchers across the fields of neuroscience, brain health, and brain-related technologies.
The EBRAINS community exemplifies open science, collaborating on new research projects and infrastructure development, sharing data and knowledge, and working across disciplines, institutions and borders. EBRAINS is a member of the European Open Science Cloud association (EOSC). EBRAINS is also dedicated to promoting Responsible Research and Innovation practices and to helping shape research in ethically sound ways that serve the public interest.
Funded by EU
PI: Prof. Dr. med. Cornelius Weiller (Freiburg Part)
Learn more: https://www.ebrains.eu/
Insufficient sleep is an under-reported epidemic and potentially linked to early signs of neurodegeneration. The EU-funded NAP project will address the issue and seek to break new ground in sleep research. Specifically, it will develop a model that studies individual sleep pathophysiology, merging in vitro modeling, allometric scaling, signal processing, and micromanufacturing. NAP will develop a cyborganoid, set up an experimental procedure to mimic sleep in vitro, exploit allometry and deliver the first tool for Parkinson’s disease early diagnosis. The long-term goal is to lead sleep research and innovation through a predictive medicine twin-on-a-chip. The project has the potential to improve healthcare costs and have a tremendous impact on pharmacological research and the healthcare sector.
Funded by EU
PI: Dr. Patrick Ruther
DiaQNOS aims to revolutionize neurosurgical tumor operations by providing surgeons with precise, real-time answers to three critical questions: Where is functional brain tissue? Where is the tumor? And where is the boundary to healthy tissue?
Using highly sensitive diamond-based quantum sensors, DiaQNOS is developing a Quantum Neuro Analyzer (QNA) — a clinically applicable, imaging endoscope capable of detecting magnetic properties of living brain tissue during surgery. This innovation will boost the safety, precision, and efficiency of brain cancer treatment.
Funded by Federal Ministry of Education and Research (BMBF)
PIs: Prof. Dr. Tonio Ball, Prof. Dr. med. Jürgen Beck, Prof. Dr. Ulrich Hofmann, Prod. Dr. med. Oliver Schnell, Prof. Dr. Andreas Vlachos
Learn more: Verbundprojekt DiaQNOS: Quanten-Neuro-Analysator zur intraoperativen Funktionsdiagnostik und Tumordetektion
Decisions about when to act are dependend on external and internal factors. External factors relate to cues from the environment, while internal factors relate to a diverse set of previous experiences and resulting motivational or physical states. Subjects have to integrate these factors and flexibly adapt their behavior accordingly in order to maximize their success rate. These decisions are partially made in the prefrontal cortex (PFC) and have to be translated into action plans in motor cortical areas. In this project, we focus on prefrontal-thalamic projections with a differentiation between the projections to the mediodorsal and the ventromedial thalamic nucleus (MD and VM). We aim to define whether there is a distinction between the information quality which is sent out from the medial PFC (mPFC) via these two routes and how this relates to the flexible formation of task-dependent and output-dependent ensembles in mPFC. For this we conduct electrophysiological recordings in the prelimbic (PL) portion of the mPFC, and in the MD and VM during a motor preparation/inhibition task in rats. In this task, rats have to respond to an auditory signal with a lever release. Via a reinforcement learning approach we identify the rules according to which the rats act and which factors relate to neuronal responses and pathways. In order to elucidate circuit specific effects, we selectively manipulate the pathways between the two structures by either blocking the PL-MD or PL-VM pathway. The project is crucial for FOR 5159 Resolving Prefrotal Flexibility as it investigates how the MD and VM pathways influence prefrontal flexibility.
Funded by DFG
PI: Prof. Dr. Ilka Diester
Learn more: https://www.for5159.de/
Microglia play a crucial role in modulating synaptic plasti city, acting as regulators of synaptic change and stability in the central nervous system. However, significant gaps in our understanding of synaptic plasticity-associated microglia persist, particularly regarding their transcriptional profiles, signaling pathways, dynamics, and their potential overlap with disease-associated microglia states. This knowledge gap is most evident in the adult human cortex, where insights into microglial contributions to synaptic plasticity remain limited. Our project aims to investigate microglial interactions with pyramidal neurons in both physiological and pathological conditions in mouse and human cortical tissue. We focus on the amyloid-β (Aβ) peptide and its impact on the spine apparatus organelle to unravel the signals and neuronal targets through which microglia affect synaptic plasticity. We aim to investigate and modulate microglial synaptic functions across neurosurgical specimens (Aim 1 ), APP-transgenic mice, and Aβ seeding models (Aim 2 ). By integrating a spatially resolved framework that combines advanced electrophysiology, microscopy, and transcriptomic analyses (Aim 3), our objective is to delineate the transition of microglial functions from facilitating to impairing synaptic plasticity. Collaborating within the CRC/TRR, our goal is to achieve a substantial molecular understanding of synaptic plasticityassociated microglia and their link to Alzheimer’s disease (AD)-associated microglia states.
Funded by DFG (CRC/TRR 167 NeuroMac)
PI: Prof. Dr. Andreas Vlachos
Learn more: Impact of amyloid-β on microglia-mediated synaptic plasticity in the rodent and human cortex : SFB/TRR 167
DEEPER project (Deep Brain Photonic Tools for Cell-Type Specific Targeting of Neural Diseases) clusters technological, neuroscientific and clinical experts together with innovative start-ups and leading companies with the aim of developing photonic tools for imaging and manipulating of the neuronal activity in deep brain regions. The long-term vision of the project is to exploit photonics for meeting medical and research needs in revealing the molecular and cellular dysfunctions underlying the pathogenesis of neurological diseases. These goals will result in less invasive and more effective treatments of dramatic social impacting pathologies, such as Alzheimer’s disease, depression, schizophrenia.
Funded by EU
PI: Dr. Patrick Ruther
Learn more: Deeper Project – Home
Transcranial Magnetic Stimulation (TMS) is a non-invasive brain stimulation technique used in neurology and psychiatry for diagnosis, therapy, and research. By using electromagnetic induction, it can activate targeted brain areas through the skull of conscious individuals. When applied repeatedly (repetitive TMS or rTMS), it can induce lasting changes in brain activity, making it a promising tool to study and influence neuronal networks.
Our research investigates how rTMS affects the balance between excitatory and inhibitory neural circuits, with a focus on its ability to reduce inhibition (“disinhibition”). This process may support the brain’s ability to form specific and lasting changes in synaptic connections, which could explain how external stimulation leads to targeted improvements in brain function. Using animal models, we study these mechanisms to better understand how rTMS can be harnessed for both clinical treatments and translational neuroscience research.
Funded by DFG / CRC-TRR384
PI: Prof. Dr. Andreas Vlachos
Learn more: Assessment and modulation of cortical inhibition using transcranial magnetic stimulation
Responding to environmental challenges through movement is crucial for animal life and requires translating movement plans into muscle activity. While the thalamus is increasingly recognized for its role in motor control, models of the motor system often overlook the thalamic reticular nucleus (TRN). However, the TRN appears to play important roles in sensorimotor signal processing. Evidence from pathology, function, behavior, and anatomy shows that the TRN is involved in movement disorders, epilepsy, motor planning, reward anticipation, attention, and action selection. Despite this, knowledge about the TRN’s role in motor control and diseases like Parkinson’s remains limited.
This project aims to clarify the functional connections between the TRN, as a potential center for inhibitory control, and the motor signaling within the thalamocortical-basal ganglia system. Although inhibitory input from the globus pallidus to TRN neurons has been demonstrated, it has mainly been studied in the context of attention and has not yet been characterized electrophysiologically. It is therefore timely to investigate whether the TRN provides a mechanism for state-dependent initiation of motor programs, whether it drives inhibitory control within the thalamocortical-basal ganglia system, whether it modulates cortical motor activity depending on state, and whether it contributes to switching between motor planning and execution.
Our hypotheses will be tested using: 1) whole-cell patch clamp combined with functional two-photon calcium imaging and two-photon optogenetic manipulation in TRN and cortical neurons; 2) functional calcium imaging and optogenetic control of TRN and cortical neurons in freely moving mice performing delayed reach tasks; and 3) behavioral experiments analyzing inhibitory control conditions and activity patterns during proactive delay.
Additionally, the project will examine whether post-inhibitory burst activity in TRN neurons represents a fundamental mechanism of inhibitory control. The results are expected to enhance understanding of motor control and inhibitory signaling networks in the motor system, and potentially identify the TRN and its mechanisms as targets for future therapeutic interventions.
Funded by DFG CO 1334/4
PIs: PD Dr. Philippe Coulon, Prof. Dr. Ilka Diester
Learn more: DFG – GEPRIS – Inhibitorische Kontrolle des motorischen Systems durch den Nucleus reticularis thalami
Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive brain stimulation technique used to induce neuronal plasticity in healthy individuals and patients. Designing effective and reproducible rTMS protocols poses a major challenge in the field as the underlying biomechanisms of long-term effects remain elusive. Current clinical protocol designs are often based on studies reporting rTMS-induced long-term potentiation or depression of synaptic transmission. Herein, we employed computational modeling to explore the effects of rTMS on long-term structural plasticity and changes in network connectivity. We simulated a recurrent neuronal network with homeostatic structural plasticity among excitatory neurons, and demonstrated that this mechanism was sensitive to specific parameters of the stimulation protocol (i.e., frequency, intensity, and duration of stimulation). Particularly, the feedback-inhibition initiated by network stimulation influenced the net stimulation outcome and hindered the rTMS-induced structural reorganization, highlighting the role of inhibitory networks. These findings suggest a novel mechanism for the lasting effects of rTMS, i.e., rTMS-induced homeostatic structural plasticity, and highlight the importance of network inhibition in careful protocol design, standardization, and optimization of stimulation.
Funded by BMBF CRCNS / NIH
PI: Prof. Dr. Andreas Vlachos
Flexible, context-dependent behavior is vital for animal survival. In mice, the anterolateral motor cortex (ALM) is thought to play a key role in linking contextual information with movement planning. Context-dependent neural responses are mainly observed in cortical layer 2, while movement planning initially emerges in layer 5. How these processes are connected, how context- and movement-selective neural populations interact, and how context-selective activity influences movement planning remain unclear.
Our main goal is to better understand how contextual information is transformed into movement planning within the ALM of mice. To study the neural processes underlying context-dependent movement selection in the ALM, we designed a bidirectional motor task in a virtual environment where the rewarded response to an auditory stimulus depends on the spatial context.
Our first aim is to characterize the activity of neural populations involved in processing context or movement planning in the ALM using two-photon microscopy. The second aim is to demonstrate the ALM’s role in this behavior through optogenetic manipulations. The third aim is to investigate the network mechanisms underlying context-dependent movement selection by selectively manipulating functionally defined neuronal subpopulations using holographic optogenetic stimulation while simultaneously recording responses with two-photon microscopy.
This project will provide new insights into the network mechanisms and neural processes in the ALM that underlie context-dependent action selection.
Funded by DFG DI 1908/16
PI: Prof. Dr. Ilka Diester
Learn more: DFG – GEPRIS – Neuronale Mechanismen kontextabhängiger Handlungsauswahl in der sekundär-motorischen Rinde
Sleep and emotional functions are closely connected. Sleep supports both emotional memory consolidation and regulation, which are essential for survival by enabling the retrieval of emotional memories and appropriate responses to threats or rewards. However, the mechanisms underlying emotional processing during sleep remain unclear. Reactivation of waking experiences during sleep supports memory and may also underlie emotional consolidation and regulation.
Rapid-eye-movement (REM) sleep is a promising candidate for supporting emotional functions, benefiting emotional memory consolidation, aiding emotional regulation, and often being disrupted in emotional disorders. While memory reactivation dynamics have been studied extensively in non-REM sleep, little is known about information processing during REM sleep and its role in emotion.
This project aims to determine REM sleep’s role in emotional memory consolidation and regulation. First, we hypothesize that the valence of prior experiences is processed during REM sleep to support emotional memory consolidation. Using novel machine learning methods, we will detect and quantify reactivation of positive or negative experiences in core emotional memory regions (Aim 1).
Second, we propose that emotional regulation is linked to homeostatic neural processes during REM sleep, which may co-occur or compete with consolidation. We will record and perturb activity in emotional brain regions during REM sleep to assess effects on behavioral and physiological emotional responses (Aim 2).
We will combine human EEG and fMRI recordings with in vivo electrophysiology and optogenetics in animal models during emotional tasks and sleep. Since disrupted sleep and memory dysfunctions contribute to psychiatric disorders such as PTSD, our findings will advance understanding of emotional memory and regulation in sleep, informing new treatments targeting sleep.
Funded by ANR-DFG NLE Grant
PI: Jun.-Prof. Dr. Monika Schönauer
Funded by DFG STI 185/14-2
PI: Prof. Dr. Thomas Stieglitz
Learn more:
Funded by DFG / SFB
PI: Prof. Dr. Ilka Diester, Dr. Patrick Ruther
Learn more: Datensatz #357581 – GSI Repository
In this Horizon 2020 project, researchers from seven European organizations examine how the vision of visually impaired people can be restored using electrical stimulation of the brain. The aim of the project is to develop a neuroprosthesis with thousands of electrodes driven by adaptive machine learning algorithms for a new brain-computer interfacing technology.
Funded by EU
PI: Dr. Patrick Ruther
Learn more: https://www.neuraviper.eu/overview
The prefrontal cortex plays a key role in representing cognitive parameters, including task-relevant sensory cues, attention, decision-making, and task rules. However, little is known about how GABAergic inhibitory interneurons (INs) contribute to the formation and stabilization of context and rule representations in active cell assemblies.
We aim to address this fundamental question in the medial prefrontal cortex (mPFC) of the mouse, the brain’s cognitive hub, for several reasons:
- Pyramidal cells (PCs), the majority of cortical neurons, are tightly regulated by GABAergic inhibitory INs.
- Fast-spiking neurons (likely GABAergic INs) in the PFC show context-selective activity, while regular-spiking cells (likely glutamatergic PCs) play a key role in rule representation.
- Parvalbumin (PV) and Somatostatin (SOM) expressing INs (PVIs, SOMIs) have distinct roles in working memory and spatial representation.
- INs receive direct input from deep CA1 pyramidal cells, a highly active population involved in spatial coding.
Building on this, we will investigate how PVIs and SOMIs influence the activity of individual neurons and cell assemblies in the mPFC during context and rule representation. We will use single-photon calcium imaging of prefrontal neuron populations in freely moving mice, combined with optogenetic manipulation of PVIs and SOMIs.
Our goal is to elucidate the circuit mechanisms underlying context and rule representation governed by interneurons in the mPFC. This project will provide new insights into the specific roles of IN types in supporting abstract representations in the prefrontal cortex.
Funded by DFG DI 1908/12-1
PI: Prof. Dr. Ilka Diester
Movement preparation and inhibition is a crucial skill-set for successfully interacting with the environment. As the executive hub, the medial prefrontal cortex (mPFC) plays a major role in learning, memorizing and executing this skill. Structural and synaptic plasticity in pyramidal cells have been shown to be a key component of learning, executing and recall of learned skills in the mPFC (Frankland et al., 2004; Restivo et al., 2009; Euston et al., 2012). However, how GABAergic inhibitory interneurons (INs) may contribute to the fine-tuning of movement preparation in the mPFC and how inhibitory signaling is changed during skill learning remains far of being understood. Here we hypothesize that changes in the efficacy of excitatory glutamatergic inputs targeting INs, similar to observations in the primary motor cortex (M1, TP6) and the hippocampus (TP1, TP7) may play a role in controlling movement preparation. We further propose that among the various IN types, parvalbumin (PV)-expressing, fast-spiking, perisomatic-inhibitory INs (PVIs) are particularly relevant in this skill (Kvitsiani et al., 2013, Pinto & Dan, 2015, Lagler et al., 2016). Previously, we have established a behavioural movement protocol combined with in vivo optogenetics and electrophysiology to investigate the neuronal underpinnings of movement control in rodents and found that specific subsections of the PFC differentially contribute to this behaviour (Hardung et al., 2017a; Hardung et al., 2017b). Here, we will focus on the role of PVIs in learning and executing the behavioural task. We aim (1) to define discharge activity of mPFC-PVIs during motor control and learning and (2) to identify the impact of mPFC-PVI neuronal activity on movement initiation and inhibition using a combination of in vivo optogenetics and electrophysiology. This will be paralleled by comparative experiments in M1 and the hippocampus by our partners (TP6 Poulet, TP7 Csicsvari). (3) We will further determine the effect of PVI activity on local neural network oscillations and – in collaboration with TP6 (Poulet) – on the putative communication with downstream motor cortex activity. (4) Finally, by applying ex vivo investigations in collaboration with Bartos (TP1), we will measure amplitudes of spontaneous and evoked excitatory signals in PVIs to examine whether PVIs inputs underwent plastic changes. Data will be compared between trained and untrained mice. With these four parallel approaches, the proposed project will provide detailed information on the impact of PVI-specific synaptic plasticity and PVI activity on cognitive movement control and learning and thereby bridge the gap between synaptic plasticity, neuronal activity, and behaviour.
Funded by DFG / SFB TP08
PI: Prof. Dr. Ilka Diester
Despite its importance in both industrial and service robotics, mobile manipulation remains a significant challenge as it requires a seamless integration of end-effector trajectory generation with navigation skills as well as reasoning over long-horizons. Existing methods struggle to control the large configuration space and to navigate dynamic and unknown environments. As a result, mobile manipulation is commonly reduced to sequential base navigation followed by static arm manipulation at the goal location. This simplification is restrictive as many tasks such as door opening require the joint use of the arm and base and is inefficient as it dismisses simultaneous movement and requires frequent repositioning.
Funded by Toyota
PI: Prof. Dr. Abhinav Valada
Learn more: N2M2: Neural Navigation for Mobile Manipulation
There is an urgent need for innovative treatments to enhance the recovery of lower limb motor function after stroke as the regaining of ambulatory ability is crucial for independence. In this regard, Brain Computer Interfaces (BCIs) have a unique position as they enable a direct re-connection between the brain and body, holding great potential to improve motor function recovery after stroke. However, current BCI solutions hardly manage to address the complexity and inter-subject variability that characterizes motor function loss and regaining after stroke. The ambition of the REWIRING project is to deploy a beyond state of art, comprehensive BCI system, the Brain-Body Computer Interface (B-BCI), able to reestablish a functional connection between the brain and the periphery to eventually improve recovery of walking after stroke (B-BCI control module). The novelty of the proposed system lays in the re-connection that will be operated according to internal neural models of (residual) motor skills grounded on the understanding of reciprocal influence between action and perception. As such, these models will be generated thanks to a multimodal monitoring of individual motor performance parameters that are operated by the B-BCI under different locomotor-related movement conditions (B-BCI monitoring module). The B-BCI will ultimately enable for an unprecedented, authentically personalized, evidence-based BCI-based reconnection between the brain and periphery.
Funded by EU / ERA-Net-Neuron
PI: Prof. Dr. Natalie Mrachacz-Kersting
Learn more: PROJECT: JTC2024 – Brain-Body: REWIRING – ERA-NET NEURON
Funded by Else Kröner-Fresenius Stiftung
PIs: Prof. Dr. Ilka Diester, Prof. Dr. Thomas Stieglitz
Learn more: Licht als Medizin: Neuer Zellschalter bringt Hoffnung für Seh-, Hör- und Herzerkrankungen
Today, around 50 million people worldwide live with dementia, and nearly 10 million new cases are diagnosed each year. Addressing the challenges of aging societies is therefore of critical importance for countries such as Japan, France, and Germany.
The aim of this project is to harness the potential of artificial intelligence (AI) approaches to promote healthy aging. To this end, we will investigate machine-learning–driven biomarkers to assess cognitive interventions and support personalized therapies. We plan to develop novel, dedicated machine learning (ML) methods tailored to the specific signal types that can be recorded from the human brain. Our methods will be made publicly available in an open-source reference software package focusing on unsupervised learning, data augmentation, domain adaptation, and interpretable ML models.
Our scientific objectives are threefold: (1) to optimize the decoding of information about brain function, (2) to identify biomarkers that indicate the risk of cognitive decline and various forms of dementia, and (3) to employ these improved methods to guide AI-driven cognitive training. These research efforts will be accompanied by a strong emphasis on the ethical and societal dimensions of AI in the context of aging, as well as participatory, transnational outreach activities designed to foster dialogue between the scientific community and the wider public.
This project brings together complementary expertise from three international research groups: the team of M. Otake-Matsuura in Japan, with longstanding experience in developing interventional technologies to enhance cognitive health in older adults and advanced EEG measurement and data analysis methods; the group of A. Gramfort at Inria, France, with deep expertise in statistical machine learning algorithms for EEG data analysis and in the development and maintenance of internationally recognized open-source software (Scikit-Learn for general ML and MNE for EEG processing); and the team of T. Ball in Freiburg, Germany, with extensive experience in translational neurotechnology and applied research using artificial intelligence.
Funded by DFG
PIs: Prof. Dr. Tonio Ball, Jun.-Prof. Dr. Philipp Kellmeyer, M.Phil.
Funded by Hector Research Career Development Award
PI: Jun. Prof. Dr. Monika Schönauer
Sleep is crucial for brain and cardiovascular health. Aging alters sleep patterns, affecting physical and cognitive well-being. Our research, using functional and structural MRI, explores the causal relationship between sleep, body, and brain functions. The scientists investigate how sleep disturbances impact brain fluid dynamics, revealing connections to neurodegenerative disorders. They further examine heart-brain interplay during sleep, assessing hemodynamic changes and potential links to elevated blood pressure. Finally, investigating age-dependent activity in the locus coeruleus during sleep, a key brainstem region, sheds light on physiological regulation and cognition. The project will thus provide insights for healthy aging and potential strategies against neurodegenerative disorders.
Funded by Else Kröner-Fresenius Stiftung First and Second Application / Project funding
PI: Dr. Deniz Kumral
Learn more: The impact of (age-related) sleep disturbances on body, brain, and memory | Else Kröner-Fresenius-Stiftung
The initiation and inhibition of movements is one of the most fundamental but yet sophisticated behavioral phenomena in humans and animals. The PFC as the hub of executive control plays an instrumental role in this context. The exact functional roles of distinct PFC subareas as well as their projections to subcortical areas remains elusive. In recent studies in rats, we observed that inactivation of medial PFC subareas strongly influence proactive behavior (according to internal cues) with orthogonal effects of infralimbic and prelimbic cortex, while inhibition of orbitofrontal cortex subareas significantly impaired reactive movements (in response to external cues). The project aims at elucidating the role of outgoing projections of the PFC subsections to their subcortical targets. For this, we will combine in vivo extracellular recordings with optogenetic manipulation of the projections, connectivity analysis and modeling.
Funded by euSNN / Grant Agreement
PIs: Prof. Dr. Ilka Diester
Learn more: 5 Optogenetic dissection of cortico-subcortical interactions during movement control in rodents | European School of Network Neuroscience
Threat (or fear) memory is critical for survival, and a leading model for dissecting how sensory input is linked to adaptive behavior by learning. Moreover, dysfunctions of threat perception such as excessive fear or chronic anxiety can lead to human anxiety disorders and depression, highly prevalent conditions causing enormous personal and societal cost. In consequence, decades of work have investigated the neuronal mechanisms of auditory threat memory. However, the lion’s share of this research has so far used very simple and reductionist auditory conditioned stimuli (CSs), for which perception is not a challenge. In addition, this work was strongly focused on subcortical pathways and mechanisms. We recently discovered that the auditory cortex plays a critical role for the acquisition and expression of threat memory when complex, naturalistic CSs are used. In contrast, conditioning to simple auditory CSs does not require auditory cortex function. These results provide new entry points on the question how threat memories are formed in the brain, raising the possibility that corticofugal feed-back projections emanating from the auditory cortex may play a key role for threat memory under naturalistic conditions. To test this hypothesis, we aim to use a combination of general and pathway-specific optogenetic inactivation, high-density in vivo extracellular recordings and discriminative auditory threat conditioning in mice. We propose to focus on two key candidate output pathways of the auditory cortex. First, the projection to the lateral amygdala has long been implicated in memory, but how this pathway affects CS encoding in the amygdala, and how threat conditioning changes this influence is not understood, in particular for the complex, perceptually challenging CSs we investigate. Second, the higher-order auditory thalamus also has a key role for threat memory, but was so far largely investigated as a source of feed-forward information transmitted to the lateral amygdala and the auditory cortex. However, the higher-order thalamic nuclei also receive strong feed-back projections from the auditory cortex, which may therefore play a key role for complex CS threat memory. In addition to dissecting the brain-wide mechanisms and coding principles of naturalistic threat memory, we aim to link our work to the overwhelming majority of studies which have used simple pure tone CSs by performing analogous experiments in a side-by-side fashion. We expect that our results will reveal so far little understood mechanisms by which the auditory cortex contributes to the processing of behaviorally relevant stimuli and to memory in subcortical brain regions. In addition to thus expanding our understanding of the brain-wide circuitry underlying associative threat memory, we expect that the use of naturalistic rather than simple CSs will furthermore generate insights with greater relevance for human experiences leading to adaptive or maladaptive outcomes.
Funded by DFG priority program 2411
PIs: Prof. Dr. Johannes Letzkus
Learn more: DFG – GEPRIS – Corticofugal mechanisms of naturalistic threat memory
Funded by BMBF
PIs: Prof. Dr. Thomas Stieglitz
Learn more:
Funded by BW Invest
PIs: Dr. Laura Comella, Prof. Dr. Peter Woias
Learn more:
Neural activity of the hippocampus and memory-dependent behavior are critically influenced by the function of synaptic glutamate receptors of the AMPA type. We will investigate the function of AMPA receptors in mossy cells of the dentate gyrus. These occupy a central position within the anatomical loops of the hippocampus. We will specifically alter the temporal dynamics of synaptic excitation by manipulating auxiliary proteins of the AMPA receptor. The effects of these manipulations will be examined using in vivo electrophysiological recordings and behavioral tests. In addition, the insights gained will be compared with predictions derived from computational models. In particular, we will test the prediction that slow EPSC kinetics support phase precession. In a further step, we will investigate the consequences of altered phase precession on dentate gyrus–dependent memory tasks, such as conditioned place preference and object location in mice. We will also refine existing single-cell and circuit models of phase precession to bring them into alignment with the obtained datasets.
PI: Prof. Dr. Christian Leibold
Learn more: https://gepris.dfg.de/gepris/projekt/545616893
Funded by Carl-Zeiss-Stiftung Wildcard
PI: Prof. Dr. Thomas Stieglitz
Learn more: https://freidok.uni-freiburg.de/proj/10206
Most mammalian species including mice have dichromatic cone vision. Trichromatic color vision has presumably emerged in evolution by first integrating an additional opsin with deviating spectral sensitivity into the retina. Then the downstream networks were adjusted to process new and hitherto unseen inputs. But which exact steps were actually necessary to integrate a whole new information channel into the existing circuits of visual information processing? An answer to this question is of immediate concern for understanding functional adaptations of the nervous system during evolution. Here, we will study this question by characterizing transgenic mice with trichromatic vision, and analyze the physiology of color vision in animals, side-by-side with a neuronal network model of primary visual cortex exposed to suitable chromatic input from an artificial retina. Our joint experimental-theoretical approach will allow us to generate new insight how trichromatic color vision might have emerged in evolution from dichromatic progenitors. Further, we hope that this particular example will also allow us to draw more general conclusions about the evolution of sensory perception and computation.
Funded by DFG / SPP2205
PI: Prof. Dr. Stefan Rotter
Learn more: Plug and play integration of a new sensory channel in evolution – Georg-August-Universität Göttingen
Funded by Toyota Motor Europe
PI: Prof. Dr. Abhinav Valada
Learn more:
Funded by Eliteprogramm für Postdoktorandinnen und Postdoktoranden, Baden-Württemberg Stiftung / Project funding
PI: Dr. Deniz Kumral
Learn more:
Funded by Honda
PI: Prof. Dr. Abhinav Valada
Learn more:
In radiology and other medical fields, informed consent often rely on paper-based forms, which can overwhelm patients with complex terminology. These forms are also resource-intensive. The KIPA project addresses these challenges by developing an AI-assisted patient information system to streamline the consent process, improve patient understanding, and reduce healthcare workload. The KIPA system uses natural language processing (NLP) to provide real-time, accessible explanations, answer questions, and support informed consent. KIPA follows an ‘ethics-by-design’ approach, integrating user feedback to align with patient and clinician needs. Interviews and usability testing identified requirements, such as simplified language and support for varying digital literacy. The study presented here explores the participatory co-creation of the KIPA system, focusing on improving informed consent in radiology through a multi-method qualitative approach. Preliminary results suggest that KIPA improves patient engagement and reduces insecurities by providing proactive guidance and tailored information. Future work will extend testing to other stakeholders and assess the impact of the system on clinical workflow.
Funded by BMBF
PI: Jun.-Prof. Dr. Philipp Kellmeyer, M.Phil.
Among epileptic patients 1/3 are drug-resistant and wait for strategies to improve life. Performing real-life seizure risk assessment & prediction will improve life quality. Until now, risk assessment & prediction have been based on low quality data limiting performance. Recently, devices for ambulatory ultra-long-term (ULT) monitoring have become available, providing data with superior quality.
We aim to empower the alliance between partners with complementary epilepsy expertise: the German (GE) partner is a reference clinical center; the Portuguese (PT) partner has an extensive experience in seizure prediction by machine-learning (ML).
The main outcome is to capture the interest of young researchers (YRs). The scientific goal is to make improvements by using ULT data supported by ML.
PT YRs will have access to practical clinical knowledge. GE YRs will profit from the PT experience in signal processing and ML.
Funded by DAAD
PI: PD Dr. Matthias Dümpelmann
Epilepsy is one of the most common neurological disorders, characterized by recurrent seizures as the main symptom. These seizures are usually caused by excessive abnormal electrical activity in the brain, which can lead to muscle stiffening, jerking movements of the limbs, and complete or partial loss of consciousness. In many cases, seizures severely affect patients’ quality of life and may result in accident-related injuries or even death.
Long-term risk prediction and short-term warning of impending seizures can give patients valuable time to assume safer body positions or take medication for seizure control in a targeted manner. In addition, the severity of the disease and the effectiveness of different therapies can be assessed to facilitate the selection of therapeutic options. An automated and personalized implant for monitoring epileptic activity outside the hospital setting could improve patients’ quality of life and reduce the burden on the healthcare system.
Existing systems for seizure detection, many of which are characterized by high energy consumption, allow only limited interventions. This is due to their reliance on standard electronics and the use of cloud-based, complex AI models, which, in this data-intensive form, lack adaptability to changing seizure dynamics and therefore prove unsuitable for continuous use in personalized medicine.
With this project, we aim to transform the lives of epilepsy patients through the use of state-of-the-art embedded neuromorphic hardware. The proposed NESPreS system is designed to enable both long-term prediction of seizure risk and short-term warnings of seizure onset. Its design philosophy is centered on the long-term application of spiking neural networks implemented using neuromorphic technology. The resulting robust, continuously deployed real-time seizure prediction provided by this intelligent, wearable system will represent a quantum leap in personalized healthcare and biomedical instrumentation.
The principal novelty of the proposal lies in the joint development of hardware and software components for a system that will be tested and validated in clinical studies. NESPreS aims to significantly reduce the physical, psychological, and social burden of epilepsy through a new technology that offers patients continuous and reliable monitoring and support. Furthermore, by employing integrated mixed-signal brain-on-chip hardware, the overall project framework will serve as a foundation for developing a new generation of permanently available edge-computing sensors for a wide range of biomedical signal processing platforms.
Funded by DFG
PI: Prof. Dr. Andreas Schulze-Bonhage
Learn more: https://gepris.dfg.de/gepris/projekt/537857806
Hippocampal pyramidal neurons can encode the spatial location of an animal through localized firing patterns, or place cells. Studies on humans and animals have shown a critical role for the hippocampus in spatial and episodic memory. It is largely thought that long-term spatial memory arises from ensembles of cells that retain their spatial coding properties over time periods relevant to long-term memory. Recently, however, a surprising level of instability has been demonstrated in the hippocampus. In fact, hippocampal spatial representations have both dynamic and stable facets. In rodents, there is day-to-day turnover in the set of hippocampal cells representing environments, which might help distinguish the representations of the different visits. However, individual CA1 neural place fields can exhibit long-term stability. Overall, the mechanisms supporting turnover versus stability of hippocampal representations are completely unknown.Ensembles of neurons undergoing coordinated activity-dependent plasticity not only represent experience but are also functional for learning and memory recall, thus they are largely believed to be cellular memory engrams. The activity of engram neurons has a key role in systems memory consolidation. As hippocampal representations are postulated to support memory formation and consolidation, engram neurons should also be important for the dynamics of these representations. Still, this issue is largely unexplored mostly owing to technical difficulties. In this project we propose to combine the complementary expertise of our laboratories to investigate the systems mechanisms that lead to turnover and specificity of representations in hippocampal CA1. Specifically, we will investigate how activity of CA3 engram neurons affects turnover, specificity and directionality of CA1 spatial representations. To achieve these aims we will take advantage of our laboratories’ cutting-edge technologies and analysis frameworks. We will combine Wide Field Head Mounted microscopes – to record neuronal activity in the CA1 of mice as they explore different environments – with optogenetics – to control the activity of CA3 engram neurons – and used advanced analysis tools to investigate the effects of forced re-activation of CA3 engram neurons on activity patters of CA1 neurons.
Funded by DFG
PI: Prof. Dr. Christian Leibold
Learn more: DFG – GEPRIS – How does activity of CA3 engram neurons affect CA1 spatial codes?
The prefrontal cortex (PFC) is thought to be engaged in diverse cognitive tasks such as working memory, decision making, response inhibition, category learning, time estimation, and many more. There is consensus that prefrontal cortex neurons exhibit mixed selectivity, i.e., a single neuron’s activity generally contributes to several of such tasks. It is therefore likely that such distributed coding could be based on the dynamical organization of prefrontal ensembles across tasks. To test this core hypothesis 1 of the Research Unit across species and developmental stages, we propose to develop and extend required data analytical tools that allow both to identify recurring population patterns (ensembles) and to relate them to task parameters. For the ensemble detection, we follow established approaches from our own lab and recent developments in the literature. For connecting patterns to behavioral parameters, we adopt recent developments (so called adversarial attacks) from deep learning research, allowing us to identify classification boundaries in high dimensional feature spaces. The methods developed and validated in this project will then be applied in collaboration with the experimental laboratories of this Research Unit a) to identify neuronal ensembles that are most informative about certain task, b) to explore to which extent these ensembles are already intrinsically present as pre-structured activity patterns before engagement in tasks, c) to explore how the representations change across development and species, and d) to study, in line with the core hypotheses 2 and 3, which neurons contribute to the ensembles and, eventually, whether those ensembles are input or output defined.
Funded by DFG
PI: Prof. Dr. Christian Leibold
Mesial temporal lobe epilepsy (MTLE) is the most frequent form of drug-resistant epilepsy in adults, in which seizures typically originate from temporal lobe structures such as the hippocampus or entorhinal cortex. Surgical resection of the epileptic focus currently represents the only curative approach for pharmacoresistant patients. However, in cases with multiple seizure foci or high risk for resection-related complications, surgery is impossible, demonstrating an urgent need for new therapeutic avenues. One promising approach to alleviate intractable seizures is deep brain stimulation (DBS) at high frequencies (100-200 Hz). Commonly, high-frequency stimulation (HFS) is applied either continuously, discontinuously, or on-demand. However, HFS has low seizure-suppressive efficacy in MTLE patients with hippocampal sclerosis (HS), presumably due to extensive neuronal loss and glial scarring. Low-frequency stimulation (LFS) represents an alternative approach, which was applied in small cohort studies including pharmacoresistant patients with HS.In the proposed project, we build on our previous preclinical findings that identified 1 Hz LFS of entorhinal afferents as a promising approach for seizure suppression. Therefore, our aim is to optimize target structures and stimulation parameters for DBS and assess potential cognitive side effects. To this end, we will use an established mouse model for MTLE with HS, which closely reflects the human pathology. Firstly, we will apply optogenetic tools to identify the most suitable neuronal population for seizure-suppressive LFS in the entorhinal-hippocampal circuit. In a second step, we will translate our findings to electrical stimulation. We will compare the three commonly used HFS configurations (continuous, discontinuous, and on-demand) to LFS with the final aim to establish long-lasting seizure control. Finally, we will determine the influence of LFS on cognitive performance with standardized behavioral tests for mobility and anxiety, as well as learning and memory. Taken together, our project will largely extend the current knowledge regarding DBS for the treatment of intractable epilepsy and therefore, could introduce LFS as a promising approach for seizure control in MTLE.
Funded by DFG
PI: Dr. Ute Häussler, since 10/2024; Prof. Dr. Carola A. Haas, until 9/2024
Learn more: https://gepris.dfg.de/gepris/projekt/511199316
Funded by DFG AT 168/9
PI: Dr. Çağlar Ataman, Prof. Dr. Ilka Diester
Learn more:
Decisions about when to act are depend on external and internal factors. External factors relate to cues from the environment, while internal factors relate to a diverse set of previous experiences and resulting motivational or physical states. Subjects have to integrate these factors and flexibly adapt their behavior accordingly in order to maximize their success rate. These decisions are partially made in the prefrontal cortex (PFC) and have to be translated into action plans in motor cortical areas. In this project, we focus on prefrontal-thalamic projections with a differentiation between the projections to the mediodorsal and the ventromedial thalamic nucleus (MD and VM). We aim to define whether there is a distinction between the information quality which is sent out from the medial PFC (mPFC) via these two routes and how this relates to the flexible formation of task-dependent and output-dependent ensembles in mPFC. For this we conduct electrophysiological recordings in the prelimbic (PL) portion of the mPFC, and in the MD and VM during a motor preparation/inhibition task in rats. In this task, rats have to respond to an auditory signal with a lever release. Via a reinforcement learning approach we identify the rules according to which the rats act and which factors relate to neuronal responses and pathways. In order to elucidate circuit specific effects, we selectively manipulate the pathways between the two structures by either blocking the PL-MD or PL-VM pathway. The project is crucial for FOR 5159 as it investigates how the MD and VM pathways influence prefrontal flexibility.
Funded by DFG DI 1908/11-1 FOR5159 TP06
PI: Prof. Dr. Ilka Diester
Learn more: https://www.for5159.de/tp6/
The prefrontal cortex is thought to be engaged in diverse cognitive tasks such as working memory, decision making, response inhibition, category learning, time estimation, and many more. There is consensus that prefrontal cortex neurons exhibit mixed selectivity, i.e., a single neuron’s activity generally contributes in several of such tasks. It is therefore likely that such distributed coding could be based on the dynamical organization of prefrontal ensembles across tasks. To test this key hypothesis of the Research Unit across species and developmental stages, we propose to develop and extend required data analytical tools that allow both to identify recurring population patterns (ensembles) and to relate them to task parameters. Specifically, we intend to connect patterns to behavioral parameters by adopting recent developments (so called adversarial attacks) from deep learning research, which allow to identify classification boundaries in high dimensional feature spaces, i.e., the neuronal ensembles that are most informative about the behavior. The methods developed and validated in this project will be applied in collaboration with the experimental laboratories of this Research Unit to a) identify neuronal ensembles that are most informative about certain task b) explore to which extent these ensembles are already present as intrinsic patterns before engagement in tasks, c) how the representations change across development and species, and d) which neurons contribute to the ensembles.
Funded by DFG / FOR TP09
PI: Prof. Dr. Christian Leibold
Learn more: FOR5159 – TP9
Morphogenic Impact of Excitatory-Inhibitory Interaction on the Structure of Neuronal Networks The architecture of the mammalian neocortex and other brain regions is not exclusively based on the execution of genetic programs but strongly relies on the homeostatic regulation of neuronal activity guiding the structural and functional differentiation of neurons and networks. The maturation of inhibition during the activity-dependent wiring process implies complex developmental interactions between excitatory and inhibitory neurons that are poorly understood so far. This project will utilize generic networks in vitro and in silico to address the following question herein:
Is the activity-dependent interaction of neuronal migration, neurite outgrowth and maturation of inhibition influencing the spatial distribution of excitatory and inhibitory neurons and their inter-connectivity?
Is the activity-dependent structural differentiation of neurons contributing to aspects of excitation and inhibition balance within local circuits or at the network level?
Is maturing inhibition promoting modular network architectures that are considered beneficial for stable activity dynamics and neuronal computation?
Close collaboration between an experimental and a theoretical PhD study is aimed at dissecting how network activity and the morphological respectively functional development of inhibition shape a network’s structure in a homeostatic loop. Versatile experimental techniques will be used to investigate and manipulate the development of inhibition and network architecture in vitro. In turn, computational modelling will allow to generate specific hypotheses and expectations for different developmental scenarios that can be tested and constrained experimentally.
Conjointly, both approaches will work towards a general mechanistic understanding of the role of maturing inhibition in the activity-dependent structural and functional development of neuronal networks.
Funded by DFG funding for material resources
PI: Dr. Samora Okujeni
Funded by EU / State of Baden-Wuerttemberg
PI: Dr. Vera Dinkelacker, Prof. Dr. Andreas Schulze-Bonhage
DREIMS is a research and innovation project that will refine a novel experimental treatment to repair the spinal cord.
Activities for 2023 to 2026 include testing the technology in animals, with clinical studies on human patients expected for 2027 and beyond.
No recruitment of persons afflicted with spinal cord injury or other neurological diseases will be carried out by any of the participant organisations in the period 2023-2026.
Nevertheless, DREIMS will mature the technology close to human application. Our endeavour is increasing as much as possible the benefit / risk balance for patients before testing the technology in human clinical trials.
Funded by EU
PI: Prof. Dr. Thomas Stieglitz
Learn more: https://dreims-project.eu/
The long-term stability of implantable and flexible thin-film electrodes is a decisive prerequisite for successful applications in neuroscience research, clinical diagnostics, and rehabilitation. Although metal electrodes have already been used very successfully as neural interfaces, delamination and crack formation have been observed with increasingly thinner layers during prolonged electrical stimulation. The systematic investigation of failure mechanisms is the aim of the “NeuroVibes” project, with a particular focus on electro-mechanical coupling in thin-film materials as the presumed main cause. For the first time, deformation vibrations of platinum thin-film electrodes induced by electrical stimulation will be measured with high spatial and temporal resolution using digital holographic microscopy. Furthermore, studies will be conducted on chemical mass transport near the electrode during electrochemical charge transfer, and innovative integrated pH sensors will be developed to detect corrosive exchange reactions and chemical imbalances. The experimental studies will be complemented by computer models that simulate the mechanical deformation behavior of membranes under, for example, cyclic mechanical loading. The resulting model will serve as a guide for stimulation parameters and the development of more stable electrode geometries. In this way, the crucial requirement for durability in neuroprosthetics can be met, and, moreover, the results can significantly contribute to the stability of all electrolytic electrode configurations, such as fuel cells in vehicles, capacitors, or batteries.
Funded by DFG funding for material resources
PI: Prof. Dr. Thomas Stieglitz
Learn more: https://gepris.dfg.de/gepris/projekt/461626908
Demand for smart system technologies has grown in most fields, both in the private and the public sectors. The industrial sector has also experienced a surge in smart technology applications as seen with smart industrial control systems. Unfortunately, due to the limited and locally embedded nature of computational resources on industrial control systems and the need for reliable algorithms with verifiable and interpretable behaviour, which currently are not present, smart industrial control systems cannot reach the level of optimisation seen in other fields. The EU-funded ELO-X project aims to combat these problems by assembling a team of PhD students and partner organisations that will research and develop solutions and methodologies for overcoming these difficulties.
Funded by European Commission Horizon 2020 program
PI: Prof. Dr. Joschka Bödecker
Learn more: https://elo-x.eu/
To date, patients with complete quadriplegia have no solution to retrieve lost hand functions. AI-HAND aims at the restoration of hand movements in patients through selective neural electrical stimulation. The project will implement cutting-edge findings in electrophysiology through radically innovative, safe and software-free implant technologies, that will lead to a generic powerful new generation of AIMD (Active Implanted Medical Device). The demonstration of the clinical relevance of this approach will be achieved through a first-in-man proof of concept .
Multiphasic stimulus waveforms, multiple synchronized currents sources for 3D current shaping over a multi contact neural cuff electrode and complex interleaved stimulation are the strong breakthrough innovations proposed, that will allow to answer unmet needs in a wide range of medical applications.
Funded by EU
PI: Prof. Dr. Thomas Stieglitz
Learn more: https://www.aihand.eu/
Funded by Mertelsmann Foundation 4
PI: Prof. Dr. Joschka Bödecker
Learn more:
Funded by Bosch1
PI: Prof. Dr. Joschka Bödecker
Learn more:
Motivation
In surgical and internal medicine hospital departments, up to 50 percent of patients experience acute but reversible cognitive impairments, known as delirium. These are often accompanied by complications and deteriorations in health. However, delirium can be avoided when it is recognized in good time and appropriate preventive measures are taken.
Objectives and Approach
Together with nursing professionals, the researchers of the KIDELIR project are developing a hybrid AI support system for detecting delirium risk. The aim is to create a practical decision-making aid for nursing staff that facilitates the timely and individualized implementation of prevention and treatment measures in daily care. The underlying data is drawn from various sources, merged to enable the system to provide targeted recommendations through pattern recognition.
Innovation and Perspectives
The system will enable AI-assisted risk assessment, helping to reduce the risk of delirium through data-driven decision support and, at the same time, lessen the high level of burden for nursing staff.
Funded by BMBF
PI: Prof. Dr. Joschka Bödecker
Learn more: https://www.interaktive-technologien.de/projekte/kidelir
Funded by BMBF
PI: Prof. Dr. Joschka Bödecker
Learn more:
Start in 2024
PI: Christian Leibold
In a previous paper (Leibold, 2020) I suggested that prewired hippocampal sequences can be used as a reservoir to build a map of a completely unknown environment with only very few landmark inputs and that this map can be successfully used for RL-based navigation. This proof of principle paper, however, was still restricted to a grid world without realistic sensory input.
In a previous BrainLinks-BrainTools-project with Valada and Welschehold we intended to bring this algorithm onto robotic platforms. Despite having two suitable potential PhD applicants, we were not able to fill the PhD position, and I used half of the funds to continue this project in my lab supported with further own funding. So far, we were able to generate a proof-of-principle implementation on the Webots platform. An agent solely receives 2-d image data of the robot camera from which it constructs a sparse map of the environment in a sequence generating network. Single units in this network exhibit hippocampal place cell like activity (including phase precession). The sparse sequence activity is used as input to a Deep RL Actor-Critic network. We show that as a result of this self-supervised learning the Value networks learns the geometry of space and expresses task phase dependent tuning (similar as PFC). This work is submitted as an abstract to Cosyne. In order to make this work publishable for a robotics community and to make the outcome usable for other researchers, we intend to transfer the system to Godot, add self-motion inputs (thereby creating an internal model of self-action in the world) and run it on standard virtual navigation benchmarks.
PIs: Ilka Diester, Harald Binder, Mona Garvert, Marlene Bartos
Whether humans, mice, and AI share fundamental principles in their internal world model design is unknown. To address this question in humans and mice, Mona Garvert and Ilka Diester designed a project within the CRC initiative BrAInWorlds in which humans and mice solve a sequence detection task. In this task stimulus dimensions (color, texture, and shape) change with fast, medium and slow frequencies, respectively. Subjects get trained and must report in the test phase which stimulus is most likely to occur next based on the learned sequences and learned frequencies of changes. Due to the different cognitive capacities and sensory preferences of the two species, the complexities of the tasks differ: Humans will get markedly more complex underlying task structures than mice and also more dimensions per stimulus. While this practical approach will ensure the feasibility of the behavior per species, it might raise a problem of comparability. Thus, the first aim of the project is the evaluation of statistical comparability of the two tasks. For this statistical evaluation, Binder will train transformer models on the same two tasks to check for the required intrinsic dimensionality and curvature to solve the task. Additionally, the transformer will check the intrinsic dimensionality and curvature of the neuronal data obtained within the project. To collect this pilot data, a mouse will be trained in a first version of the planned task (Diester, Bartos). This will provide us insights into whether the chosen stimuli and dimension are appropriate to let a mouse realize that there is a logical sequence in the data structure. Diester will record from the frontal cortex, which is important for action planning whereas Bartos from the hippocampus a key structure for spatial and contextual memory formation. Garvert will pilot behavior in humans to assess whether participants can learn such complex relational structures and provide the data to Harald Binder. This paves the way for designing fMRI experiments enabling us to investigate the representation of these structures in the human brain. By combining recordings from frontal, hippocampal and brain-wide structures, we will be enable to address the question whether internal world models are specific to brain regions.
PIs: Ilka Diester, Harald Binder
A fundamental question within the BrAInWorld initiative is whether the brain uses multiple internal world models which are associated to specific brain areas. By comparing previously acquired recordings from prefrontal, premotor, motor and primary sensory cortex of rats who conducted the same task, we will address this question. i.e. whether internal world models are specific to cortical regions. The electrophysiological single unit data were required from rats which performed a simple lever pressing task which we previously modelled via reinforcement learning tools with Joschka Bödecker. Here, we aim to analyze the data set via manifold analyses applying dimensionality reduction techniques, i.e. a transformer approach. The transformer approach will have to be adapted to the multidimensional dataset but if successful will provide an estimate of the intrinsic dimensionality and curvature of the manifolds. This will provide a first glimpse into commonalities and differences of the underlying internal world models which are represented in the different brain areas.
PIs: Abhinav Valada, Tim Welschehold
Internal world models provide robots with the ability to predict the outcomes of their actions [1], enabling them to learn efficiently by minimizing reliance on trial-and-error in the real world. This predictive capability is crucial for scaling skill learning to long-horizon tasks, where success often requires the ability to not only refine existing skills to adapt to changing requirements [2, 3, 4] but also acquire entirely new skills when needed. By leveraging world models, robots can simulate and plan within virtual environments [5, 6], enhancing their ability to tackle complex and dynamic challenges with greater autonomy and adaptability. This project focuses on leveraging internal world models to advance robotic autonomy through two primary aims:
Aim 1: Skill Refinement within a World Model: Optimize pre-existing robotic skills by simulating diverse contexts and iteratively improving performance without extensive physical trials. Although prior works [2, 4, 7] have optimized pre-trained robot skills efficiently, they still rely heavily on resource-intensive physical interactions. To address this, we will learn a robust world model from offline teleoperated play data, providing a virtual environment for simulating scenarios and testing skills. We will then refine existing skills by optimizing their policies within the world model, enabling iterative and cost-effective improvement.
Aim 2: Learning Novel Skills via a World Model: Identify and acquire new skills by integrating Vision Language Models (VLMs) to detect skill gaps and world models to simulate learning and validate the results. This approach bridges the semantic reasoning of VLMs [8] with the physical grounding of world models, enabling robots to autonomously expand their skill repertoire for dynamic environments. Specifically, we will generate goal visualizations from textual inquiries using VLMs [9], providing the robot with clear visual targets for novel skills. The robot will then use its world model to simulate and practice actions that achieve these goal visualizations, allowing it to learn new skills efficiently in a virtual environment before transferring them to the real world.
PI: Monika Schönauer
The dynamics of the real world often follow stochastic processes that can be learned through observation. Based on these observations, we construct an internal world model to predict the outcomes of our decisions and thus select optimal, goal-directed actions. Humans and animals continuously integrate new experiences and external multimodal information with past knowledge, internal states, and future projections to refine their internal world models. Reinforcement learning models allow us to model multifactorial contributions to behavioral decision making and thus to assess the internal world model of individuals while they solve complex tasks. In collaboration with Prof. Joschka Bödecker and Hao Zhu, we have demonstrated this in a visual hierarchical category learning task (Kleespies*, Paulus*, et al., preprint expected in early 2025). For future use in cross-species approaches, we already implemented a standardized data structure and analysis pipeline (Diester, Schneider, Bödecker, Zhu, Schönauer, Paulus).
The current project idea originates from a joint Carl Zeiss Wildcard application by Prof. Ilka Diester, Prof. Joschka Bödecker, and myself that builds on this past collaboration: we want to demonstrate that humans are able to learn a complex multimodal rule that determines which of two realistic coffee shops would be the better choice to enter (“coffee shop decision task”). The task structure follows our visual hierarchical category learning task, but translates it into a multi-dimensional, more naturalistic scenario. Participants will have to learn and integrate values across multimodal dimensions of the coffeeshops, such as the background music, architectural style, and theme of the place (e.g., cozy vs. minimalistic). While multimodal feature binding in humans is well known, it remains unclear whether humans are able to integrate feature values across multimodal dimensions during information integration learning. Moreover, it is unknown whether the same computational dynamics control visual vs. multimodal information integration processes. Our ability to navigate complex multimodal environments, however, makes this highly likely. We will develop the naturalistic coffeeshop decision making task and pilot its feasibility in 50+ human participants. Further, we will test whether it can be modeled using reinforcement learning methods (in collaboration with Bödecker and Zhu). We aim to publish a behavioral and cognitive modeling paper on hierarchical category learning in a life-like scenario (timeline 12-18 months). Dr. Anika Löwe is an ideal candidate to conduct this work, based on her strong prior experience in computational modeling and the cognitive neuroscience of decision making.
Relevance: The highly accessible coffeeshop task can serve as an excellent pilot project demonstrating the feasibility of assessing internal world models using reinforcement learning techniques, and can be used as an example and demonstrator for the upcoming reviews of the Collaborative Research Center BrAInWorlds. In this consortium, Ilka Diester is planning to study multimodal multidimensional decision making in a virtual environment in rodents, such that a naturalistic and multimodal decision task in humans would provide multiple clear connecting points for cross-species comparisons.
PI: Thomas Stieglitz
Cutting edge technologies are competing for the “best” ECoG array towards functional Brain-Machine Interfaces. However, the importance of understanding adjacent processes with potentially significant long-term effects on such implants is often forgotten. Investigation of explanted neural devices after chronic implantation are rarely done. The mechanical and chemical factors during explantation and sample preparation which could compromise electrode array integrity, potentially leading to incorrect conclusions about structural changes not only within host tissue but also at the implant level are not yet considered. These issues raise critical concerns about the long-term functionality and safety of recording/stimulating devices. Our proposed study aims to deliver a comprehensive analysis of ECoG electrode integrity across “life-time”, starting from pre-implantation stage until the final check-up of the post-experimental electrode integrity. The results will contribute to a full-picture of potential mechanisms occurring at the electrode level from different perspectives, addressing degradation mechanisms with cellular origin, aging processes and oxidative species involvement, together with mechanical impact associated with post-implantation sample handling. In addition, we conduct an in-vitro study of the metal-electrode polymer substrate and the reactions over time in contact with the simulated harsh conditions of a body environment as means to evaluate potential failure mechanisms of the system. Analysis and evaluation of whole brain-ECoG implant system will bring new insights into the in-vivo and in-vitro influence on device functionality by using SEM imaging after establishing cross sections by focused ion beam (FIB-SEM) cutting and analyzing the change of thin-film metal structures with means of mass spectroscopy (ToF-SIMS) to find causes of metal embrittlement. The comprehensive assembly of methods holds significant potential for advancing the understanding of failure mechanisms, to deliver propositions towards improvement of longevity, and emphasize the critical role of sample preparation protocols.
PI: Julian Ammer
To interrogate the neural circuits that implement the animals’ internal representations of the world, methods are needed to manipulate both the internal and the external world with high precision. The OptoRoboRat setup at the IMBIT is combining holographic optogenetics (for precise ‘internal’ neuronal manipulations) with a virtual reality setup (for precise manipulation of external cues). As this setup is central to a number of projects and will be increasingly busy, we plan to set up three additional virtual reality setups. We envision that these will be used threefold:
- As training setups in which we can train animals in parallel before moving them to the OptoRoboRat setup for recording. This will free up time from the OptoRoboRat setup and allow us to run experiments in parallel more efficiently.
- For miniscope recordings in which we can test how animals that were trained in a freely moving task generalize their internal task model to the same task in a virtual reality under head fixation.
- As augmented reality setup: In the future we could easily adapt the current VR to an augmented reality setup in which we could add fast and precise control over visual cues to an existing freely moving task.
These setups will be set up in the same way as the existing virtual reality setup at the OptoRoboRat which was custom built and programmed with open source software. Therefore, the blueprint for these setups already exists and we have a full parts list available for immediate start of ordering in the case of funding.
PIs: Deniz Kumral, Andreas Schulze-Bonhage
Cognitive maps represent space in humans and animals. Yet, it is still not clear whether these maps encode space in uniform manner, or alternatively by segmenting it into multiple subspaces. The latter process is known as schematization, which does not just enable the segmentation of space, but potentially can offer a valuable generalization mechanism, drawing on an internal world model (IWM) of how the spatial world is structured. This allows navigation in known and unknown spatial environments utilizing similar cognitive maps. It is unknown whether this schematization depends purely on spatial information or whether it also incorporates semantic maps. In this project, we will develop a naturalistic spatial navigation task to investigate the interaction of spatial and semantic segmentation of a cognitive map.
To enhance ecological validity, the project will build a detailed virtual reality model of Freiburg city, allowing for the study of spatial IWMs under naturalistic conditions. To be applicable to both healthy and partially memory-impaired participants (patients: BrAInWorlds proposal P14 Schulze-Bonhage/Vlachos), elderly: P15 Kumral/Schönauer), an adjustable task that can be tailored to individual patients’ and older participants’ spatial capacities will be designed by the research groups. The adaptable nature of this task will have significant benefits, because it will 1) enable behaviorally relevant single-neuron recordings and brain imaging in humans, 2) allow for the investigation of real-city navigation behavior in patients with material-specific memory deficits and aging individuals. 3) facilitate comparisons between spatial IWMs of memory-impaired individuals and the general population. In the long run, this joint project can reveal the daily navigational challenges faced by memory-impaired individuals as well as the compensatory strategies they employ during navigation. We can thus identify potential avenues for developing navigation support interventions.
Aim: The main aim of this project is to design and pilot the virtual Freiburg navigation task. Kumral and Schulze-Bonhage will work closely together on this project, in collaboration with Monika Schönauer. The main task of Kumral will be the task programming and task design. Jointly, Kumral and Schulze-Bonhage will then make adjustments to the task based on pilot testing in different populations (healthy young adults, patients) and experimental demands. Finally, Schulze-Bonhage will run behavioral tests on the virtual navigation task in a patient population with simultaneous intracranial recordings to test feasibility in the patient experimental setting.
Relevance: By combining behavioral measures and multi-scale neural recordings in a naturalistic virtual environment, the planned SFB BrAInWorlds projects of the PIs (P15 Kumral/Schönauer, P14 Schulze-Bonhage/Vlachos) will significantly advance our understanding of semantic representations in cognitive maps and their pivotal role in spatial navigation. The pilot project we propose here will pave the way to the success of these SFB projects, which will employ the naturalistic Freiburg virtual reality task. We will not only be able to assess the behavioral and navigational strategies of human subjects in a naturalistic environment, but also to extract the neural representations of spatial and semantic maps and determine their interaction. Neural measurements using fMRI allow to extract spatial IWMs at a macroscale, which we will complement with single neuronal and local field potential representations of space (e.g. place cells) and semantics (e.g. concept cells) in patients undergoing intracranial recordings in brain regions of interest (particularly temporal and parietal). The planned multiscale neural measurements will provide a comprehensive understanding of schematization mechanisms of spatial cognitive maps from a single neuron scale to population and regional dynamics.
PIs: Cristian Pasluosta, Thomas Stieglitz
Sensory feedback and motor actions are naturally uncertain and thus humans rely on prior and current sensorimotor information to perform motor tasks. Previous physical interactions with our surroundings shape our sensorimotor memory. New incoming sensorimotor information provides evidence of how likely a next event is to happen. The brain integrates this prior and current information to define our perceptions and actions over motor tasks. Further, this integration weights prior and current sensorimotor information depending on their degree of uncertainty, following a Bayesian probabilistic framework. Disruption of lower limb proprioceptive and cutaneous sensations after limb loss increases sensorimotor uncertainty, which may lead to a higher reliance on prior sensorimotor information to control gait patterns. This research will investigate for the first time how lower limb amputees deal with sensorimotor uncertainties during gait using a Bayesian framework. Participants will wear a virtual reality (VR) headset while walking on an omnidirectional treadmill. The VR headset will project marks on the floor indicating where to place the foot during the next step. Visual feedback of the foot position will only be provided for a short period of time halfway from toe off to heal strike. This visual feedback will be shifted randomly by a known amount, thus representing the prior distribution of foot placement. Further, participants will walk under three different levels of vision blur. Foot and trunk position will be recorded using the VR trackers placed on each foot and pelvis. The different levels of sensory uncertainties will be used to estimate how much amputees rely on previous sensorimotor memory during these sensory challenging tasks. This will be performed to compare whether, and if so, how do they learn the imposed prior distribution of target foot placements.
Start in 2023
- Learning Multisensory Integration for Neural Circuits Modeling (HippoSLAM); Leibold, Valada
- Combining Bidirectional Optogenetic fMRI of PFC subsections with Behavioral Studies (opto-fMRI with behavior); Diester, Zaitev
- Development of a neural scanning probe for treatment of depression by deep brain stimulation (DB-Scan); Stieglitz, Coenen
- Brain-Body axis at your wrist – quantifying effects of brain stimulation with wearable technology (Bra-Bo-Stim); Schulze-Bonhage, Vlachos
- Investigations on ultrasound for energy supply and data transmission in neural implants (TalkativeImplants); Stieglitz, Rupitsch
- Monitoring of electrode degradation during current-controlled stimulation (MEDCOS); Weltin, Kieninger
- Causal role of beta bursts in inhibitory control of movement (Beta bursts and inhibitory control); Diester, Ball
- Translational VR paradigms for neuroscience: From rodents to humans (TransVR); Ball, Diester
- Learning Mobile Manipulation for Rearrangement from Scene Graphs (MoMaSceneGraphs); Valada, Brox, Welschehold
- Neurotechnological Human-Robot Interaction: From offline to online (NeuroHRI); Bödecker, Valada, Ball
- Engaging Science and Technology: Part 1 – Robotics (EnSciTec); Livanec, Kellmeyer
- How does the past influence what we learn in the future? Investigating the role of prior knowledge on learning in new situations; Schönauer, Mehring
- Holographic mapping of brain state – dependent functional connectivity in neural ensembles (State dependence ensemble mapping); Ammer, Leibold