BME 5111/6011: Diffusion-Weighted Imaging and Functional MRI Study Guide

Diffusion-Weighted Imaging (DWI) Principles and Calculations

• Parameter Control of Diffusion Weighting ($b$):
      * The $b$-value, measured in $\text{s/cm}^2$ (though traditionally expressed as $\text{s/mm}^2$ in clinical practice), is controlled by altering several pulse sequence parameters:
          * The temporal gap ($T$) between the applications of the gradient pulses.
          * The amplitude ($g$) of the pulses.
          * The duration ($\delta$) of the pulses.
      * The Diffusion-Weighting Equation: The relationship between these parameters is defined by the following formula:
          * $b = \gamma^2 g^2 \delta^2 (T - \frac{\delta}{3}$
      * Constants and Constraints for Hydrogen ($H^1$):
          * Gyromagnetic ratio for Hydrogen: $\gamma = 2\pi \times 42.58 \times 10^6 \, \text{rad/(T} \cdot \text{s)}$
          * Maximum gradient slope available: $g = 8 \, \text{G/cm} = 8 \times 10^{-4} \, \text{T/cm}$
          * Time constraint for the experiment: T < 100 \, \text{ms}

• Computation of Gap ($T$) and Duration ($\delta$) for Specific $b$-values:
      * Case a: $b = 500 \, \text{s/cm}^2$
          * Substituting the constants into the equation:
              * $500 \, \text{s/cm}^2 = (2\pi \times 42.58 \times 10^6 \, \text{rad/(T} \cdot \text{s)})^2 \times (8 \times 10^{-4} \, \text{T/cm})^2 \times \delta^2 (T - \frac{\delta}{3}$
              * $500 \, \text{s/cm}^2 = 4\pi^2 (42.58^2 \times 10^{12}) \, \text{rad}^2 / (\text{T}^2 \cdot \text{s}^2) \times (64 \times 10^{-8}) \, \text{T}^2 / \text{cm}^2 \times \delta^2 (T - \frac{\delta}{3}$
              * Isolating the temporal components: $\delta^2 (T - \frac{\delta}{3}) = \frac{500 \, \text{s/cm}^2}{4\pi^2 (42.58^2)(64)(10^{-4}) \, \text{rad}^2 / (\text{cm}^2 \cdot \text{s}^2)} = 1.05 imes 10^{-8} \, \text{s}^3$
          * Using trial-and-error to find practical (though noted as physically impractical in context) values:
              * $\delta = 2 \, \text{ms}$
              * $T = 3.4 \, \text{ms}$
      * Case b: $b = 1000 \, \text{s/cm}^2$
          * To double the $b$-value from the previous step, the target product must be: $\delta^2 (T - \frac{\delta}{3}) = 2.11 imes 10^{-8} \, \text{s}^3$
          * Calculated parameters:
              * $\delta = 2 \, \text{ms}$
              * $T = 6.11 \, \text{ms}$
      * Case c: $b = 2000 \, \text{s/cm}^2$
          * To double the value again, the target product is: $\delta^2 (T - \frac{\delta}{3}) = 4.22 imes 10^{-8} \, \text{s}^3$
          * Calculated parameters:
              * $\delta = 2 \, \text{ms}$
              * $T = 11.4 \, \text{ms}$
      * Case d: $b = 3000 \, \text{s/cm}^2$
          * The target product for this high diffusion weighting is: $\delta^2 (T - \frac{\delta}{3}) = 6.33 imes 10^{-8} \, \text{s}^3$
          * Calculated parameters:
              * $\delta = 2 \, \text{ms}$
              * $T = 17.02 \, \text{ms}$

• Note on Units: There was a systematic error mentioned regarding the $b$ parameter; it is typically provided in $\text{s/mm}^2$. However, for the purposes of mastering the use of the diffusion equation, the calculations were performed in $\text{s/cm}^2$.

Practical Considerations in Gradient and Parameter Control

• Impact of Doubling Gradient Slope ($g$):
      * If the gradient slope is increased to $g = 16 \, \text{G/cm}$, the researcher has more flexibility in choosing $T$ and $\delta$ while maintaining a fixed $b$-value.
      * Optimization Strategy: While there is no strict physical rule mandating which parameter to alter, it is practical to assign a maximum value to the temporal gap ($T$).
      * By varying $\delta$ within a fixed $T$, researchers ensure there is sufficient time to insert a $180^{\circ}$ pulse between the two gradient pulses, which is necessary for certain refocusing sequences.

Diffusion Sensitivity and Signal Attenuation

• Low-Pass Filter Analogy for Diffusion ($D$):
      * Increasing the $b$-value acts as an increasingly strict low-pass filter cutoff for measuring diffusion.
      * This is due to the signal representation: $S = e^{-bD}$.
      * Effect of Higher $b$: As $b$ increases, the signal drop becomes much more pronounced for any given value of $D$.
      * Observed Signal Requirement: To maintain any given observed signal as $b$ increases, the tissue or substance must have progressively lower values of $D$. High $b$-values essentially "filter out" the signal from anything with high diffusion coefficients.

Physiological Basis of the BOLD effect in fMRI

• The BOLD (Blood Oxygen Level Dependent) Effect Process:
      * The sequence of physiological activities from stimulus onset to baseline recovery involves several stages:
          1. Neural Activation: Local neural tissue experiences increased metabolic activity due to higher excitation or inhibition rates. This results in an increase in glucose consumption to fuel the Adenosine Triphosphate (ATP) cycle.
          2. Oxygen Extraction Increase ("The Pre-Dip"): Increased glucose consumption leads to higher oxidative metabolism. The rate of oxygen extraction from the blood increases, leading to a higher concentration of deoxygenated or "reduced" hemoglobin (HbR) in the capillary bed. This causes a localized drop in Magnetic Resonance (MR) field homogeneity and $T_2^*$, resulting in a small, localized signal drop immediately following stimulation.
          3. Vasodilation and Over-compensation: Glial cells supporting the neurons release peptides. These peptides cause the dilation of adjacent capillaries, increasing the local blood volume. This phenomenon (documented by Sherrington and Roy in 1890) increases localized blood delivery. The increase in oxygenated blood delivery is so great that it over-compensates for the HbR, causing the relative concentration of [HbR] to drop. This improves field homogeneity, leading to a significant increase in the MR signal.
          4. Transition to Recovery: After several seconds of elevated delivery (supporting neural activity that may have only lasted for milliseconds), blood flow and volume begin a slow decrease back toward initial levels.
          5. Post-Stimulus Undershoot: Because the oxygen/glucose delivery lasted much longer than the neural event (seconds vs. milliseconds), the tissue eventually reaches a state where the need for hematocrit reduces significantly. Blood flow and volume drop below the initial baseline level, causing a period of slight over-extraction of oxygen/glucose.
          6. Return to Baseline: The undershoot period can persist for nearly $30 \, \text{seconds}$ before glucose extraction, oxygen extraction, blood flow, and blood volume finally return to baseline equilibrium.

Experimental Designs in Functional Neuroimaging

• Blocked Experimental Design:
      * Structure: Involves sustained periods (blocks) of stimulus presentation (e.g., flashing checkerboards, heat, words) alternating with periods of non-presentation.
      * Modeling: These are generally modeled using square waves.
      * Strengths: This approach is optimized for detection. It maximizes the ability to identify whether a specific area of the brain is responding to the stimulus (|h(t)| > 0).
      * Weaknesses: It is difficult to characterize the specific shape of the hemodynamic response ($h(t)$). Because the signal is an integration of constant stimulus responses, it is very hard to "de-convolve" the data to see the underlying rise, peak, and fall of a single response.

• Event-Related Experimental Design:
      * Structure: Involves the presentation of isolated, transient stimuli (e.g., single words, short bursts of sound) separated by enough time that the brain's response can return to baseline or be mathematically separated.
      * Sufficient Separation: Hemodynamic responses to individual events are assessed in isolation or via selective averaging.
      * Strengths: This design is preferred for the characterization of the hemodynamic response, allowing researchers to estimate its shape, amplitude, rise time, fall time, and undershoot duration.
      * Selective Averaging: This works effectively if the input sequence of stimuli is "uncorrelated."
      * Weaknesses: While averaging multiple events can eventually help with detection, the inherent variation in responses across the brain makes this method less certain for detection compared to the blocked design.