Adaptive Burst Detection in Neural Organoids
Developing threshold methods that adapt to individual organoid characteristics rather than using fixed parameters
The Challenge
Traditional burst detection methods use fixed thresholds across all recordings, but we’ve found that midbrain organoids show significant baseline variability in their electrical activity patterns. A one-size-fits-all approach misses critical neural dynamics.
Our Approach
We developed an adaptive RMS-based threshold method that:
- Calculates baseline activity characteristics for each individual organoid
- Adjusts detection thresholds based on organoid-specific noise floors
- Accounts for temporal drift in recording quality
- Preserves biological variability while reducing technical noise
Key Findings
Our analysis revealed that 6-OHDA treatment doesn’t simply reduce activity—it fundamentally reorganizes network dynamics. Treated organoids show:
- Increased burst-to-burst variability
- Loss of stereotyped temporal coordination
- Network fragmentation patterns
Next Steps
We’re now implementing these methods in our containerized analysis pipeline for deployment on the NRP Kubernetes cluster. This will enable large-scale analysis across multiple experimental batches.
This work is part of my dissertation research in the Sharf Lab at UC Santa Cruz.