Computationally efficient feature denoising filter and selection of optimal features for noise insensitive spike...

Area of Science:

• Neuroscience
• Signal Processing
• Computational Biology

Background:

• Accurate spike sorting is crucial for real-time neural data analysis.
• Existing feature extraction methods can be sensitive to noise, impacting classification accuracy.

Purpose of the Study:

• To develop a new feature extraction method for spike sorting that improves noise immunity and preserves spike information.
• To evaluate the performance and computational efficiency of the proposed method.

Main Methods:

• A feature denoising filter was applied to detected spikes.
• Six features, including a novel one, were extracted from filtered spikes.
• A separability index was used for optimal feature selection.

Main Results:

• The method achieved spike classification error as low as 5% at a noise level of 0.2.
• The three highest-performing features, including the new one, were identified.
• Computational complexity was reduced to 11% of the Principle Component Analysis (PCA) method.

Conclusions:

• The proposed feature extraction method enhances noise immunity and spike classification accuracy.
• This method offers a computationally efficient alternative for real-time spike sorting applications.