MEG: Basic Data Processing and Time-frequency Analysis Stephan Grimault, PhD November 22, 2006
General outline 1) Basic Pre-processing and processing of MEG data basic ERF (ERP) analysis and activation map Pre-processing and processing: definitions and purpose Pre-processing steps Processing steps 2) Another way to analyze the data: Time-frequency Analysis
Basic ERF (ERP) analysis Protocol example Stimuli
Activation map (1)
Activation map (2)
Pre-processing & Processing: definition and purpose Pre-processing : - Clean the data to keep only the functional signals of interest (involves detection and correction of noise and acquisition artifacts) - optimize data processing Processing: - analysis of functional signals of interest - to evaluate to cognitive hypotheses
General outline 1) Basic Pre-processing and processing of MEG data basic ERF (ERP) analysis and activation map Pre-processing and processing: definitions and purpose Pre-processing steps Processing steps 2) Another way to analyze the data: Time-frequency Analysis
Pre-processing is not trivial: MEG measures magnetic fields from femtotesla (10-15 T) to picotesla (10-12 T) Earth s magnetic field: 4,710-5 T. small magnetic field measurements lead to artifacts
Pre-processing steps (1): Environmental Noise Reduction
Pre-processing steps (2): DC offset removal
Pre-processing (3): Frequency filtering: High-pass filter Low-pass filter
6 f 4 f f
6 f 4 f f
6 f 4 f f 0.1 f
6 f 4 f f
6 f 4 f f 0.05 f
6 f 4 f f 0.1 f 0.05 f
γ β α θ 24-40-100Hz 12-45Hz 8.5-12Hz 4-8.5Hz δ 0.05-4Hz
Sampling rate = 600Hz t = 0.0016 s Signal Theory frequency MAX = Sampling/2 Low Pass Filter = Sampling rate /4 : 600Hz 150Hz
Pre-processing (3): Frequency filtering example High Pass Filter Low Pass Filter
Pre-processing (4): Eye artifact (blinks, motion) 20pT
Eye blink correction ex. 1
Eye blink correction ex. 2
Pre-processing (5): Current Noise at 60 Hz (50 Hz Europe)
Pre-processing (6): ECG or watch artifact
Pre-processing (7): Dental artifacts
Pre-processing (8): Subject motion artifact
General outline 1) Basic Pre-processing and processing of MEG data basic ERF (ERP) analysis and activation map Pre-processing and processing: definitions and purpose Pre-processing steps Processing steps 2) Another way to analyze the data: Time-frequency Analysis
Processing (1): single trial recording
Processing (1): Why record N trial SNR N 1 10 100 200
Processing (2) : select condition
Processing (3) : averaging
Processing ends work of interpretation begins! Protocol example Bonne chance! Stimuli
Basic Pre-processing and Processing of MEG data: Summary Pre processing Environmental Noise Reduction Frequency filtering : High Pass, Low Pass Eye blink correction 60 Hz or 50 Hz (Europe) ECG or watch artifact Processing Recording N trials by condition Select condition (trigger) Averaging
General outline 1) Basic Pre-processing and processing of MEG data basic ERF (ERP) analysis and activation map Pre-processing and processing: definitions and purpose Pre-processing steps Processing steps 2) Another way to analyze the data: Time-frequency Analysis
Time Frequency map representation
Keep Information from thetrials
Average after keep information from Trials
TF map calculation : the wavelets
TF map calculation : the wavelets
Time frequency analysis: one step more
Time frequency analysis: two step more Localisation? Localisation Discussion?
Thanks to: Franco Lepore (CERNEC) Pierre Jolicoeur (CERNEC) Équipe MEG (Anne Sophie, Christophe,, Jean- Marc, Kevin, Manon, Mihaela)