Signal and Noise in fMRI

What is signal? What is noise?

I. Introduction to SNR

Signal, noise, and the General Linear Model

Signal-Noise-Ratio (SNR)

Signal Size in fMRI

Slide 7

Slide 8

Differences in SNR

Slide 10

Slide 11

Effects of SNR: Simulation Data

SNR = 2.0

SNR = 1.0

SNR = 0.5

SNR = 0.25

SNR = 0.125

Slide 18

What are typical SNRs for fMRI data?

Effects of Field Strength on SNR

Theoretical Effects of Field Strength

Measured Effects of Field Strength

Excitation vs. Inhibition

II. Properties of Noise in fMRI

Can we assume Gaussian noise?

Types of Noise

Why is noise assumed to be Gaussian?

Is noise constant through time?

Slide 28

Is fMRI noise Gaussian (over time)?

Is Signal Gaussian (over voxels)?

Variability

Variability in Subject Behavior: Issues

Response Time Variability

Intersubject Variability

Variability Across Subjects

Young Adults

Elderly Adults

Slide 38

Slide 39

Effects of Intersubject Variability

Slide 41

Implications of Inter-Subject Variability

Spatial Variability?

Standard Deviation Image

Spatial Distribution of Noise

Low Frequency Noise

High Frequency Noise

III. Methods for Improving SNR

Fundamental Rule of SNR

Slide 50

Trial Averaging

Example of Trial Averaging

Slide 53

Slide 54

Increasing Power increases Spatial Extent

Slide 56

Effects of Signal-Noise Ratio on extent of activation: Empirical Data

Active Voxel Simulation

Effects of Signal-Noise Ratio on extent of activation:
Simulation Data

Explicit and Implicit Signal Averaging

Caveats

Accurate Temporal Sampling

Slide 63

Slide 64

Slide 65

Slide 66

Slide 67

Slide 68

Accurate Spatial Sampling

Partial Volume Effects

Partial Volume Effects

Partial Volume Effects

Partial Volume Effects

Partial Volume Effects

Where are partial volume effects most problematic?

Activation Profiles

Temporal Filtering

Filtering Approaches

Power Spectra