Signal Processing and Deep Learning Frameworks for Spatiotemporal Analysis of In-Vivo Neural Activity
Mentors
Project Description
This project focuses on developing advanced signal processing and machine learning algorithms for a newly developed computational miniature mesoscope that enables cortex-wide neural activity imaging in freely behaving mice. The system captures rich, multi-scale data spanning single-neuron calcium dynamics as well as resting-state vascular and hemodynamic signals. The student will work on algorithmic methods for denoising, motion correction, spatiotemporal demixing, and cross-modal analysis to extract interpretable neural and vascular activity patterns from large-scale in-vivo datasets. The project is conducted in close collaboration with a neuroscience group, with the goal of enabling quantitative studies of functional connectivity, neurovascular coupling, and brain-wide activity organization.
Research Goals
Weeks 1-2: Onboarding, background reading, and familiarization with the mesoscope system and datasets; introduction to existing analysis pipelines.Learning Goals
Timeline
Weeks 3-4: Development of core preprocessing modules (motion correction, denoising, background removal).
Weeks 5-7: Implementation of machine learning methods for neural activity extraction, segmentation, and spatiotemporal analysis.
Week 8-9: Integration of neural and vascular data analysis; validation on experimental datasets in collaboration with neuroscience partners.
Weeks 10: Final analysis, documentation, and presentation of results.