Spike Sorting

1
Spike Sorting

• Goal Extract neural spike trains from MEA
  electrode data
• Method 1 Convolution of template spikes
• Method 2 Sort by spikes features

2
Cluster Cutting

• Advantages
• Better separation
• Requires less information
• Disadvantages
• Computationally intensive

3
Remap2pin02 Spikes
4
Selected Features

1. Max peak height
2. Voltage difference between max and second max
3. Sum of max positive and max negative peaks
4. Time between max positive and max negative peaks
5. Max width of a polarization

5
Features

1. Max peak height -- Color
2. Voltage difference between max and second max
   -- Z-axis
3. Sum of max positive and max negative peaks --
   Y-axis
4. Time between max positive and max negative peaks
   -- X-axis
5. Max width of a polarization -- Size

6
Features Plot
7
Remap2pin02 Spikes
8
Training Features Plot
9
Training Features Plot
10
Training Features Plot
11
Future Direction

• Optimal feature choice
• Training algorithm
• Bayesian clustering
• Nearest neighbor
• Support Vector Machine
• Neural Network

12
Conclusion

• Data suggests we should be able to isolate
  individual neural firing patterns from MEA data
• Use MEA data to model and study network of
  neurons in culture