2022 Fiscal Year Annual Research Report
Metacognitive control of the neural signals that shape behaviour changes
| Project Area | Deciphering and Manipulating Brain Dynamics for Emergence of Behaviour Change in Multidimensional Biology |
|---|
| Project/Area Number |
22H05156
|
|---|
| Research Institution | Advanced Telecommunications Research Institute International |
|---|
Principal Investigator |
CORTESE Aurelio 株式会社国際電気通信基礎技術研究所, 脳情報通信総合研究所, 研究室長 (60842028)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
川人 光男 株式会社国際電気通信基礎技術研究所, 脳情報通信総合研究所, 所長 (10144445)
|
|---|
| Project Period (FY) |
2022-06-16 – 2027-03-31
|
| Keywords | reinforcement learning / metacognition / neural dynamics / cerebellum / decision-making / prefrontal cortex / neuroimaging / decoded neurofeedback |
|---|
| Outline of Annual Research Achievements |
Research stream 1 (Hoang, Toyama, Kawato, Cortese): We developed a new algorithm that can accurately extract fine-grained task and behaviour information from high-dimensional neural activity patterns (Hoang et al. 2023 eLife). Our new algorithm is coupled with reinforcement learning to generate insight into the neural representations occurring during specific behaviour changes. The algorithm was tested and validated on calcium imaging and behaviour data of mice performing a go/no-go task acquired by groups A02 Kitamura and Matsuzaki. Specifically, spike trains from a population of neurons are decomposed into components via tensor component analysis to infer low-dimensional latent structure in firing responses and further characterised with respect to temporal factors, task conditions, behaviour repertoire, and neurons.
Research stream 2 (Six, Misawa, Okamoto, Cortese): We designed the core behaviour tasks necessary for our experiments, which aim to induce efficient behaviour change by controlling low-dimensional neural representations. In short, tasks require participants to make optimal decisions to earn rewards while adapting to changing hidden rules or conditions. After IRB approval, we piloted the task in both its behaviour-only format and with fMRI (N=1 pilot participant). We presented the preliminary results at the Computational Neuroscience Winter Workshop (Rusutsu) as well as at other domestic conferences.
|
| Current Status of Research Progress |
Current Status of Research Progress
3: Progress in research has been slightly delayed.
Reason
Research stream 1 (Hoang, Toyama, Kawato, Cortese): Working on analysing and modelling calcium imaging data to characterise cerebellar spiking activity better. We are preparing our manuscript, running additional analyses, and getting ready to submit it within the following months. In parallel, we have started the discussion on designing a new computational model to accommodate findings of multiple neural subpopulations.
Research stream 2 (Six, Misawa, Okamoto, Cortese): We are working on the analysis flow of pilot data (behaviour and fMRI). Based on the results from our pilot experiment and analyses, we will update our task design and likely expand the scope of behaviour testing to have a wider and more reliable sampling of confidence / metacognitive abilities. These efforts will be necessary to successfully design and implement a neurofeedback study to change behaviour through the control of metacognitive neural signals.
|
| Strategy for Future Research Activity |
Research stream 1 (Hoang, Toyama, Kawato, Cortese): In our previous studies we found two populations of neurons in the cerebellum with very similar temporal response profiles and spatial distribution that, however, mapped on different behaviour functions. We plan to carry out computational analyses to test the hypothesis that these two neuron types form one population, with neurons changing their cue-response specificity due to long-term depression of parallel-fibre (PF) to Purkinje-cell synapses guided by climbing fibres (CF) inputs. First, we will trace the CF activity of individual Purkinje cells during learning. Second, to elucidate the mechanism of such transformation, we will construct a simple cerebellar neural network model of PCs, each receiving CF and PF inputs.
Research stream 2 (Six, Taylor, Misawa, Cortese): Based on task piloting developed in the previous year, we will extend our design and development of multiple behaviour tasks that will allow us to test behaviour change through additional measures. All tasks will share a common main structure, in which participants update their behaviour in response to hidden rule changes. We will collect participants’ confidence ratings and record eye and facial movements during decision-making. In this fiscal year, we will design and pilot the tasks to obtain clear and reliable behavioural measures and further test the recording components (camera, eye tracking device, fMRI for neural signals). In parallel, we will build a decoding pipeline to extract low-dimensional behavioural trajectories in fMRI neural space.
|
Research Products
(10 results)
