2020 Fiscal Year Research-status Report
Experimental and theoretical investigation of Bayesian inference across multiple layers in somatosensory cortex
| Project/Area Number |
19K20348
|
|---|
| Research Institution | Okinawa Institute of Science and Technology Graduate University |
|---|
Principal Investigator |
LI Yuzhe 沖縄科学技術大学院大学, 神経計算ユニット, 研究員 (30815436)
|
|---|
| Project Period (FY) |
2019-04-01 – 2022-03-31
|
| Keywords | calcium imaging / prism lens / functional connectivity / transfer entropy / multiple cortical layers / clustering |
|---|
| Outline of Annual Research Achievements |
1. The new novelly designed setup has been finalized. The new setup, which includes the 3D printed body parts, the motor and sensor circuits, and the control and recording programs, is designed and achieved by myself. The control program and recording program are developed based on python, which is able to achieve fixed sampling-rate recording and task control parallelly. The new control paradigm includes a fluctuation to the motor, which will introduce uncertainty to mice’s sensory input
2. Achieved clear multiple-layer calcium imaging, after failures on more than 80 mice.
3. Achieved analysis on multiple layer recordings of resting states, which includes transfer entropy, HMM analysis and probabilistic PCA for neuron clustering.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
1. It takes a bit more time in developing the new setups. The details of the models for 3D printing were modified many times according to testing with animals.
2. After a lot of failures, the multiple layer imaging is finally well taken.
3. Analysis on a dataset that contains sequential multiple layer scanning by 2-photon calcium imaging was done, which is a bit different from the dataset from this experiment, but the methods I developed for the analysis can be shared with this project.
|
| Strategy for Future Research Activity |
1. Animal training with tasks need to be done.
2. The Bayesian model and deep learning model analogue are worth to try on the neural recordings taken during the tasks.
|
| Causes of Carryover |
1. The new animal experiments and surgeries need to purchase more virus and lens.
2. An extra computer needs to be purchased for separating the task control and data recording to avoid frame dropping problem.
3. To accelerate training, a second setup needs to be made, which needs to purchase more printing materials, and electronic elements.
|
