Neural Signal Processing: MMN Demo

• The lecture demonstrates neural signal processing techniques with hearing screening and epileptic seizure detection demos.
• Explores different biopotentials, neuro-potentials, and signal processing methods for various applications.

Mismatch Negativity (MMN) Demo

• MMN demo involves presenting four identical sounds (standard) and one different sound (deviant).
• Deviant sound elicits a slight negativity compared to standard sounds, indicating mismatch negativity, reading data, and obtaining event-related potentials.

Experimental Setup for MMN and Auditory Brainstem Response (ABR)

• The subject wears a headband with electrodes, including one on the ear lobe.
• Data channels connect to an acquisition system wirelessly linked to a computer.
• Open BCI open-source GUI is used for three channels, and blink-thinking digital I/O indicates sound timing.

MMN Extraction Methodology

• The standard/deviant sound pattern is repeated 100 times to reduce neural variation.
• Averaging is used to extract neural responses.
• Ongoing research focuses on minimizing iterations (epochs) in event-related potential research.

Matlab Demonstration with EEG Lab

• EEG lab, a software for EEG analysis, is used with the Erp lab plug-in for evoked/event-related potentials.
• EEG lab is a combined platform for EEG, ECG, ecog, and seeg processing.

Data Import and Processing Ethics

• Privacy through data anonymization is crucial.
• Import data using EEG lab functions from Matlab arrays, including the sampling rate (e.g., 250 Hertz).
• Avoid using the same name for different subjects.

Event Information Import

• Event information can be imported from data channels, such as the fourth and fifth channels.
• Present two types of sounds, like 100 deviant and 400 standard sounds.

Power Line Interference

• Check for power line interference using Erp lab's amplitude spectrum plot, up to the Nyquist frequency (120 or 125 Hz for a 250 Hz sampling rate).
• In the Indian region, the power line frequency is 50 Hz.

Filtering Techniques

• Use filters, such as a Notch filter, to remove power line interference at specific frequencies, like 50 Hz (49-51 Hz cutoff) and 100 Hz (99-101 Hz cutoff).
• Apply a band pass filter (e.g., 3 to 30 HZ) with a stop band around -100 DB and a linear phase during the pass band.

Data Examination and Artifact Rejection

• Plot channel data (e.g., 20-second intervals) to observe repeating events.
• Remove non-neural artifacts like eye blinks; filtering allows selective frequencies, while artifact rejection allows selected magnitudes.
• Create an event list in EEG lab, specifying standard and deviant sounds, then create bin-based epochs with pre-stimulus (50 or 100 ms) and post-stimulus (400 ms) ranges.

Epoch Rejection and ERP Analysis

• Remove eye blinks by setting time interval (e.g., -100 to 396 ms) and microvolt range (e.g., -50 to 50).
• Artifact rejection criteria: Example: Out of 400 standard sounds, 330 passed artifact rejection.
• Compute average ERPs, excluding rejected epochs, and plot the final waveform to analyze.
• MMN generally appears within 100 to 300 milliseconds as negativity.
• Standard error of the mean (SEM) shows variability.

Batch File Processing

• Useful with multiple data sets.
• Commands performed in the session are stored in the history.

ABR Extraction Signal Processing

• ABR gets generated within 10 milliseconds of 0.1 microvolt amplitude.
• Raw data is noisy, requiring bandpass filtering and power line artifact removal.
• Response generated within 10 milliseconds demands a sampling rate of 10 kHz or more.
• Adaptive filtering is used: bandpass filter, spectrum check for neural/non-neural peaks, and rejection of peaks using band reject filters.

ABR Extraction Demo

• Uses an updated ABR subroutine requiring data input directory, sampling rate, event channel, and data channel inputs.
• Involves data reading, event list generation, epoch binning, and artifact rejection.
• Initial data is noisy with multiple power line interferences which are filtered.
• Final ABR waveform shows picks around 0.6 in amplitude.

Batched ABR Processing

• The final image displays time domain in milliseconds and amplitude.
• The process includes plotting raw data, raw spectrum, clear spectrum, a comparison, and the final waveform, identifying and removing non-neural components.

Rodent Experiment Setup

• Uses rodents with 10 channels and a biodegradable micro electrode array (MEA).
• Data is acquired from the rat's brain using an equation board (e.g., open BCI board) with a stimulation module.
• Baseline recording verifies electrode contacts and positioning.

Epilepsy Induction and Testing

• Epilepsy is induced using drugs to test anti-epileptic drug efficacy.
• Quantifies how quickly drugs restore the normal state after seizures.

Rodent Data Demonstration

• Data is analyzed using pure MATLAB-based analysis in the signal analyzer app.
• EEG lab requires data to be saved as a mat file for importing, specifying a 125 Hz sampling rate across 10 channels.

EEG Lab Observations

• A flat line on one channel indicates a possible connection issue.
• Caesar-like spikes and sharp waves are visible in the data.

Signal Analyzer App: Quantification

• Quantifies changes, time, frequency parameters, and features.
• Generates FFTs or amplitude spectrum and time-frequency analysis (spectrogram).

Baseline Generation and Spectrum Plotting

• Three boxes are available to input baseline, epilepsy, and after AD values, standardizing the Y-axis.
• Sampling rate of 125 was used while acquiring the data.
• Ensure limits are consistent when comparing spectrums for time domain amplitude.

Time Frequency Analysis with Spectrogram

• Displays time on the X-axis, frequency on the Y-axis, and amplitude/intensity as a heat map.
• Min/max ranges should match for accurate comparisons, with adjustable leakage and resolution.

Panner and Script Generation

• Use the Panner to focus on specific regions of interest.
• Capture and document results by copying the display.
• Generate reusable scripts for spectrum, spectrogram, and Panner settings.

Neural Signal Processing and Interpretation

• GUI-based eeglab works and batch file processing using ABR.
• Extract the response neural response and interpret to draw neural inferences.
• Spectrograms can be altered to spectrum, persistent spectrum, or scalogram based on data.