Computational models of dynamics in brain networks underlying action selection
Lead Research Organisation:
University of Oxford
Department Name: UNLISTED
Abstract
In Parkinson’s disease, neurons in certain parts of the brain produce abnormal activity. For example, their activity tends to oscillate, which causes the tremor of patients’ hands. One common treatment for the disease involves implanting electrodes in the affected brain regions and providing electric stimulation. Recently a new generation of such deep brain stimulators has been developed, which include multiple contacts that can measure brain activity and provide stimulation according to the measured signals. However, to take advantage of this technology, it needs to be understood what patterns of activity are produced during action selection in the healthy brain, because restoring such patterns should be a goal of the stimulation. Furthermore, we need to understand how to stimulate with multiple contacts to achieve desired neural dynamics. The overall aim of the programme is to provide mathematical description of the dynamics of brain networks underlying action selection and to understand how these dynamics can be modified by treatments for disorders affecting the system. This research is important, because it will contribute to development of a new generation of brain stimulators that will more effectively ameliorate symptoms of Parkinson’s disease and produce fewer side-effects.
Technical Summary
Recent advances in brain computer interfaces open new possibilities of normalizing pathological neural activity underlying symptoms of Parkinson’s disease. For example, patients are now implanted with closed-loop DBS systems including multiple recording and stimulation contacts, allowing the independent control of multiple neural populations. However, to take advantage of this technology, it needs to be understood what patterns of activity are produced during action selection in the healthy brain, because restoring such patterns should be a goal of closed-loop DBS systems. Furthermore, we need to understand how to stimulate with multiple contacts to achieve desired neural dynamics. Such insights are currently missing, so there is a need to develop a theory providing them. The overall aim of the programme is to provide mathematical description of the dynamics of brain networks underlying action selection and to understand how these dynamics can be modified by treatments for disorders affecting the system. The programme has three specific goals that focus on the three neural signals are particularly distorted in Parkinson’s disease. The first goal is to develop a theory of dopamine function in learning and action planning. Understanding its function is important because Parkinson’s disease is caused primarily by the dysfunction and death of neurons releasing dopamine, and medications increasing dopamine level are the most common treatment for Parkinson’s disease and many psychiatric conditions. The second goal is to describe the dynamics of beta oscillations during action planning. These oscillations are thought to be related with the symptoms of Parkinson’s disease, because in Parkinson’s disease the duration of intervals with high beta oscillations is longer when patients are off medications and their movement difficulties are more pronounced. The third goal is to identify control policy supressing tremor for closed-loop DBS with multiple contacts. To achieve these goals, the computational models will be developed based on data gathered in experimental neuroscience and neurology groups within our MRC Unit, and the models will inform development and refinement of interventions, through a collaboration with the neural engineering group.
Publications
Song Y.
(2020)
Can the brain do backpropagation? - Exact implementation of backpropagation in predictive coding networks
in Advances in Neural Information Processing Systems
Song Y
(2020)
Can the Brain Do Backpropagation? -Exact Implementation of Backpropagation in Predictive Coding Networks.
in Advances in neural information processing systems
Salvatori T
(2022)
Learning on Arbitrary Graph Topologies via Predictive Coding.
in Advances in neural information processing systems
Salvatori T
(2021)
Associative Memories via Predictive Coding.
in Advances in neural information processing systems
Salvatori T.
(2021)
Associative Memories via Predictive Coding
in Advances in Neural Information Processing Systems
Van Swieten MMH
(2023)
Gambling on an empty stomach: Hunger modulates preferences for learned but not described risks.
in Brain and behavior
Van Swieten MMH
(2021)
Hunger improves reinforcement-driven but not planned action.
in Cognitive, affective & behavioral neuroscience
Bogacz R
(2020)
Dopamine role in learning and action inference.
in eLife
Zamora M
(2021)
Case Report: Embedding "Digital Chronotherapy" Into Medical Devices-A Canine Validation for Controlling Status Epilepticus Through Multi-Scale Rhythmic Brain Stimulation.
in Frontiers in neuroscience
Millidge B.
(2022)
Predictive Coding: Towards a Future of Deep Learning beyond Backpropagation?
in IJCAI International Joint Conference on Artificial Intelligence
Duchet B
(2020)
Phase-dependence of response curves to deep brain stimulation and their relationship: from essential tremor patient data to a Wilson-Cowan model.
in Journal of mathematical neuroscience
Calder-Travis J
(2023)
Expressions for Bayesian confidence of drift diffusion observers in fluctuating stimuli tasks
in Journal of Mathematical Psychology
Duchet B
(2021)
Optimizing deep brain stimulation based on isostable amplitude in essential tremor patient models.
in Journal of neural engineering
Duchet B
(2023)
How to entrain a selected neuronal rhythm but not others: open-loop dithered brain stimulation for selective entrainment
in Journal of Neural Engineering
Averna A
(2023)
Spectral Topography of the Subthalamic Nucleus to Inform Next-Generation Deep Brain Stimulation
in Movement Disorders
Herz DM
(2022)
Dynamic control of decision and movement speed in the human basal ganglia.
in Nature communications
Oswal A
(2021)
Neural signatures of hyperdirect pathway activity in Parkinson's disease.
in Nature communications
Song Y
(2024)
Inferring neural activity before plasticity as a foundation for learning beyond backpropagation.
in Nature neuroscience
Duchet B
(2023)
Mean-Field Approximations With Adaptive Coupling for Networks With Spike-Timing-Dependent Plasticity.
in Neural computation
Lefebvre G
(2022)
A Normative Account of Confirmation Bias During Reinforcement Learning.
in Neural computation
Herz DM
(2023)
Dynamic modulation of subthalamic nucleus activity facilitates adaptive behavior.
in PLoS biology
Duchet B
(2021)
Average beta burst duration profiles provide a signature of dynamical changes between the ON and OFF medication states in Parkinson's disease.
in PLoS computational biology
Tang M
(2023)
Recurrent predictive coding models for associative memory employing covariance learning.
in PLoS computational biology
Möller M
(2022)
Uncertainty-guided learning with scaled prediction errors in the basal ganglia.
in PLoS computational biology
Related Projects
Project Reference | Relationship | Related To | Start | End | Award Value |
---|---|---|---|---|---|
MC_UU_00003/1 | 01/04/2020 | 31/03/2025 | £1,280,000 | ||
MC_UU_00003/2 | Transfer | MC_UU_00003/1 | 01/04/2020 | 31/03/2025 | £2,361,000 |
MC_UU_00003/3 | Transfer | MC_UU_00003/2 | 01/04/2020 | 31/03/2025 | £1,126,000 |
MC_UU_00003/4 | Transfer | MC_UU_00003/3 | 01/04/2020 | 31/03/2025 | £2,269,000 |
MC_UU_00003/5 | Transfer | MC_UU_00003/4 | 01/04/2020 | 31/03/2025 | £2,274,000 |
MC_UU_00003/6 | Transfer | MC_UU_00003/5 | 01/04/2020 | 31/03/2025 | £2,177,000 |
Title | Behaviour and pupillometry in a bandit task |
Description | Behavioural data and pupil diliation obtained from humans performing learning and decision making task. |
Type Of Material | Database/Collection of data |
Year Produced | 2021 |
Provided To Others? | Yes |
Impact | No impact yet |
URL | https://data.mrc.ox.ac.uk/data-set/behaviour-and-pupillometry-bandit-task |
Title | Effects of hunger on model-based and model-free decision-making |
Description | Behavioural data from an experiment investigating effects of hunger on contributions of different neural systems to decision making. |
Type Of Material | Database/Collection of data |
Year Produced | 2021 |
Provided To Others? | Yes |
Impact | No impact yet. |
URL | https://data.mrc.ox.ac.uk/data-set/effects-hunger-model-based-and-model-free-decision-making |
Title | Human LFP recordings from STN during sequential conflict task |
Description | Local field potentials recorded from human subthalamic nucleus in a sequential decision making paradigm. |
Type Of Material | Database/Collection of data |
Year Produced | 2020 |
Provided To Others? | Yes |
Impact | No impact yet. |
URL | https://data.mrc.ox.ac.uk/data-set/human-lfp-recordings-stn-during-sequential-conflict-task |
Title | NeurOLAB |
Description | MATLAB toolbox for generating neural oscillations, simulating LFP signals and testing brain stimulation strategies. |
Type Of Material | Computer model/algorithm |
Year Produced | 2021 |
Provided To Others? | Yes |
Impact | The toolbox has been used to performed simulations described in our publication (PMID: 34358224). |
URL | https://github.com/gihan-weerasinghe/neurolab |
Title | The effects of hunger on experiential and explicit risk-taking |
Description | Set of behavioural data from an experiment investigating effects of hunger on risk taking. |
Type Of Material | Database/Collection of data |
Year Produced | 2021 |
Provided To Others? | Yes |
Impact | No impact yet. |
URL | https://data.mrc.ox.ac.uk/data-set/effects-hunger-experiential-and-explicit-risk-taking |
Description | Computational modelling of confirmation bias in reinforcement learning |
Organisation | University of Oxford |
Department | Department of Experimental Psychology |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Develop mathematical models of confirmation bias in reinforcement learning |
Collaborator Contribution | Professor Chris Sumerfield related computational models to data from humans in learning tasks |
Impact | The collaboration resulted in a joint publication (PMID: 34758486). This is an interdisciplinary collaborabion which combines: - Computational neuroscience (Rafal Bogacz) - Psychology (Chris Summerfield) |
Start Year | 2020 |
Description | Computational models of deep learning in the brain |
Organisation | University of Oxford |
Department | Department of Computer Science |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | We developed computational models of learning in cortical circuits. |
Collaborator Contribution | Professor Thomas Lukasiewicz evaluated how well the brain inspired models solve real world machine learing problems. |
Impact | The collaboration resulted in a joint publication (PMID: 33840988). This is an interdisciplinary collaborabion which combines: - Computational neuroscience (Rafal Bogacz) - Computer Science (Thomas Lukasiewicz) |
Start Year | 2020 |
Description | Learning reward uncertainty in the striatum |
Organisation | University of Oxford |
Department | Department of Experimental Psychology |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | - Formulation of experimental predictions based on computational models - Data analysis |
Collaborator Contribution | Experiments measuing activity of specific cell types in striatum, while mice learned about mean and variability of rewards associated with different options. |
Impact | Abstracts descrribing results so far have been submitted to CoSyNe and FENS conferences. This is an interdisciplinary collaborabion which combines: - Neurophysiology (Mark Walton, Pete Magill) - Computational neuroscience (Rafal Bogacz) |
Start Year | 2020 |
Description | Mathematical modelling of confidence during decision making |
Organisation | University of Oxford |
Department | Department of Experimental Psychology |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Mathematical models of confidence judement during decision making. |
Collaborator Contribution | Experiments analysing how human participants assign confidence to their decisions. |
Impact | A manuscript describing results obtained so far has been submitted to Psychological Review. This is an interdisciplinary collabotation combining: - Psychology (Nick Yeung) - Computational Neuroscience (Rafal Bogacz) |
Start Year | 2020 |
Description | Model of neural circuits predicting dynamic stimuli |
Organisation | University of Auckland |
Department | Auckland Bioengineering Institute (ABI) |
Country | New Zealand |
Sector | Academic/University |
PI Contribution | We provide expertise on mathematical model of cortical circuits |
Collaborator Contribution | Dr Mahyar Osanlouy has extended and simulated predictive coding model of information processing of visual cortex, learning representation of dynamic stimuli. |
Impact | A new computational model of cortical circuits has been developed. |
Start Year | 2020 |
Description | Modelling dopaminergic activity during perceptual decision making |
Organisation | University College London |
Department | Gatsby Computational Neuroscience Unit |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Development of computational models of dopaminergic activity while animals learn to make perceptual decisions. |
Collaborator Contribution | Experimental measurements of dopaminergic activity while animals learn to make perceptual decisions. |
Impact | An abstract desribing results so far has been submitted to FENS conference. This is an interdisciplinary collaborabion which combines: - Neurophysiology (Armin Lak, Oxford) - Computational neuroscience (Andrew Saxe, UCL; Rafal Bogacz) |
Start Year | 2021 |
Description | Modelling dopaminergic activity during perceptual decision making |
Organisation | University of Oxford |
Department | Department of Physiology, Anatomy and Genetics |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Development of computational models of dopaminergic activity while animals learn to make perceptual decisions. |
Collaborator Contribution | Experimental measurements of dopaminergic activity while animals learn to make perceptual decisions. |
Impact | An abstract desribing results so far has been submitted to FENS conference. This is an interdisciplinary collaborabion which combines: - Neurophysiology (Armin Lak, Oxford) - Computational neuroscience (Andrew Saxe, UCL; Rafal Bogacz) |
Start Year | 2021 |
Description | Neural bases of risky decision making |
Organisation | University of Oxford |
Department | Nuffield Department of Clinical Neurosciences |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Developing computational models of brain decision circuits predicting human behaviour during decision making involving risky option |
Collaborator Contribution | Prof Sanjay Manohar designed and supervised experimental studies testing predicitions of computational models. |
Impact | Outputs include 3 datasets from performed experiments which have been made available on the MRC Brain Network Dynamics Data Sharing Platform. The collaboration resulted in a joint publications: van Swieten, M. M., Bogacz, R., & Manohar, S. G. (2021). Hunger improves reinforcement-driven but not planned action. Cognitive, Affective, & Behavioral Neuroscience, 21(6), 1196-1206. Moeller, M., Grohn, J., Manohar, S., & Bogacz, R. (2021). An association between prediction errors and risk-seeking: Theory and behavioral evidence. PLoS computational biology, 17(7), e1009213. Moeller, M., Manohar, S., & Bogacz, R. (2022). Uncertainty-guided learning with scaled prediction errors in the basal ganglia. PLoS Computational Biology, 18(5), e1009816. This is an interdisciplinary collaborabion which combines: - Computational neuroscience (Rafal Bogacz) - Clinical Neuroscience (Sanjay Manohar) |
Start Year | 2020 |
Description | Study of subthalamic nucleus activity during decision making |
Organisation | University College London |
Department | Institute of Neurology |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Co-designing an experiment testing predictions of our computational model of decision making in the basal ganglia |
Collaborator Contribution | Co-designing and performing experiments |
Impact | Publications: Leimbach, F., Georgiev, D., Litvak, V., Antoniades, C., Limousin, P., Jahanshahi, M., & Bogacz, R. (2018). Deep brain stimulation of the subthalamic nucleus does not affect the decrease of decision threshold during the choice process when there is no conflict, time pressure, or reward. Journal of cognitive neuroscience, 30(6), 876-884. Patai, E. Z., Foltynie, T., Limousin, P., Akram, H., Zrinzo, L., Bogacz, R., & Litvak, V. (2022). Conflict Detection in a Sequential Decision Task Is Associated with Increased Cortico-Subthalamic Coherence and Prolonged Subthalamic Oscillatory Response in the ß Band. Journal of Neuroscience, 42(23), 4681-4692. Dataset: Human LFP recordings from STN during sequential conflict task https://data.mrc.ox.ac.uk/data-set/human-lfp-recordings-stn-during-sequential-conflict-task |
Start Year | 2014 |
Description | Subharmonic entrainment of neural oscillations by deep brain stimulation |
Organisation | University of California, San Francisco |
Department | School of Medicine (UCSF) |
Country | United States |
Sector | Academic/University |
PI Contribution | Mathematical modelling of effect of deep brain stimulation on cortical activity. |
Collaborator Contribution | Experimental test of predictions of mathematical models of effect of deep brain stimulation on cortical activity. |
Impact | A manuscript desribing results obtained so far has been submitted to Brain Stimulation. his is an interdisciplinary collaborabion which combines: - Neurology (Philip Starr, UCSF) - Neural Engineering (Tim Dennison, Oxford) - Computational neuroscience (Rafal Bogacz) |
Start Year | 2021 |
Title | DEEP BRAIN STIMULATION |
Description | There is provided a method of generating deep brain stimulation signals, the method comprising receiving a plurality of sensor signals from a corresponding plurality of sensors on or in a subject, and using the received sensor signals to generate a plurality of stimulation signals for application at a corresponding plurality of target sites in the brain of the subject. There is further provided a method of generating stimulation signals, the method comprising receiving a plurality of sensor signals from a corresponding plurality of sensors on or in a subject, and using the received sensor signals to generate a plurality of stimulation signals for application at a corresponding plurality of target sites on or in the subject using a model of the response of neurons in the subject to the stimulation signals that models neural tissue as a plurality of coupled populations of neurons. |
IP Reference | WO2022029445 |
Protection | Patent application published |
Year Protection Granted | 2022 |
Licensed | No |
Impact | This method is described in our pubication (PMID: 34358224). |
Company Name | NeuEdge |
Description | Develops AI accellerator hardware |
Year Established | 2022 |
Impact | None yet |
Description | Podcast: Why do we develop bad habits? |
Form Of Engagement Activity | A broadcast e.g. TV/radio/film/podcast (other than news/press) |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Public/other audiences |
Results and Impact | Podcast explaining mechanisms of habit formation in the brain |
Year(s) Of Engagement Activity | 2022 |
URL | https://podfollow.com/928408356/episode/81f13938d6221bec8cdc2938f0237f43a95dce56/view |
Description | STEM placements for local school pupils (in2science) |
Form Of Engagement Activity | Participation in an open day or visit at my research institution |
Part Of Official Scheme? | No |
Geographic Reach | Local |
Primary Audience | Schools |
Results and Impact | Each year our group hosts 1 or 2 pupils from local schools in Oxford. The placement scheme was tailored for pupils from local state-funded schools to support their progress into university degrees and careers in science, technology, engineering and mathematics (STEM). During their time in the Unit, the pupils worked alongside Unit scientists and received personalised mentoring to gain a wide variety of practical experiences and learn more about key concepts and challenges in neuroscience and medical research. In a series of integrated workshops with in2scienceUK, the pupils also received guidance on university applications, wider information about STEM careers, and training in transferable skills. The pupils recorded their experiences and progress in blogs and images. |
Year(s) Of Engagement Activity | 2016,2017,2018,2019,2022 |
URL | https://www.mrcbndu.ox.ac.uk/news/unit-hosts-school-pupils-fourth-year-stem-placement-scheme |
Description | Schools Open Day |
Form Of Engagement Activity | Participation in an open day or visit at my research institution |
Part Of Official Scheme? | No |
Geographic Reach | Local |
Primary Audience | Schools |
Results and Impact | Each year around 70 GCSE/A level students from 6 local schools within Oxfordshire area attended the school open day. A range of hands-on practical sessions, lab tours and talks ran by Unit members at all levels to provide the children with an insight into the nature and benefits of medical/brain research, and inspire them to pursue a career in science. Also in attendance in 2018 were the Mrs Jean Fooks - Lord Mayor of Oxford, Cllr Chris Wright - Chair of Garsington Parish Council and Cllr Elizabeth Gillespie - South Oxfordshire District Council. |
Year(s) Of Engagement Activity | 2016,2017,2018,2022 |
URL | http://www.mrcbndu.ox.ac.uk/news |