3.9 Parameter Tradeoffs: SNR and CNR

Parameter Tradeoffs: SNR and CNR

In magnetic resonance imaging (MRI), the quality of the image is governed by three primary competing factors: Signal-to-Noise Ratio (SNR), spatial resolution, and scan time. Adjusting any single parameter on the console inevitably triggers a compromise among these three properties. Understanding these relationships is essential for the ARRT registry exam and for daily clinical practice.

---

Signal-to-Noise Ratio (SNR) and Voxel Volume

Signal-to-Noise Ratio (SNR) is the ratio of the amplitude of the signal received by the radiofrequency (RF) coil to the amplitude of the background electrical noise. Signal represents the coherent information gathered from the patient's hydrogen protons, while noise represents incoherent, random electrical fluctuations generated by the patient's body and the scanner's receiver electronics.

The amount of signal collected is directly proportional to the size of the voxel, which represents the three-dimensional volume of tissue. The voxel volume is calculated as:

$\text{Voxel Volume} = \text{Slice Thickness} \times \left( \frac{\text{Field of View (FOV)}_{\text{frequency}}}{\text{Frequency Matrix}} \right) \times \left( \frac{\text{Field of View (FOV)}_{\text{phase}}}{\text{Phase Matrix}} \right)$

1. Slice Thickness

Slice thickness determines the depth of the voxel.

• Thicker Slices: Increasing slice thickness increases voxel volume, which increases the number of hydrogen protons within the voxel. This leads to a linear increase in signal. For example, doubling the slice thickness from 3 mm to 6 mm doubles the SNR.
• Tradeoff: Thicker slices reduce spatial resolution along the slice-select axis and introduce partial volume averaging, which occurs when multiple tissue types are averaged together within a single voxel, masking small structures.

2. Field of View (FOV)

The FOV represents the physical size of the imaging window.

• Larger FOV: Increasing the FOV increases the dimensions of the voxels. If the FOV is increased symmetrically (e.g., from 12 cm to 24 cm), the voxel size increases in two dimensions (frequency and phase). This causes the voxel volume, and therefore the SNR, to increase as the square of the FOV change: $\text{SNR} \propto (\text{FOV})^2$
• Tradeoff: Symmetrically doubling the FOV increases the SNR by a factor of 4, but significantly degrades in-plane spatial resolution.

3. Matrix Size

The matrix size determines the number of pixels across the FOV.

• Larger Matrix (e.g., $256 \times 256$ to $512 \times 512$): Increasing the matrix size divides the FOV into smaller pixels, reducing voxel volume. If both matrix dimensions are doubled, the voxel volume is reduced to $1/4$ of its original size. Consequently, the SNR is reduced to $25%$.
• Tradeoff: A larger matrix increases spatial resolution but severely degrades SNR.

---

Number of Excitations (NEX / NSA)

The Number of Excitations (NEX), also known as the Number of Signals Averaged (NSA), refers to how many times the same line of k-space is filled during the scan.

• The Physics of Averaging: MRI signal is coherent and adds up linearly when averaged. Noise is random and incoherent, adding up as the square root of the number of measurements. Therefore, the relationship between SNR and NEX is: $\text{SNR} \propto \sqrt{\text{NEX}}$
• Tradeoffs:
  • To double the SNR using NEX, you must increase the NEX by a factor of 4 ($\sqrt{4} = 2$). This quadruples the scan time.
  • Doubling the NEX (e.g., from 1 to 2) increases the SNR by approximately $41%$ ($\sqrt{2} \approx 1.41$) and doubles the scan time.
  • While NEX is an effective way to recover lost SNR, it is highly time-inefficient.

---

Receiver Bandwidth (rBW)

Receiver Bandwidth (rBW) is the range of frequencies that the scanner's receiver coil is tuned to collect during the frequency encoding readout.

• The Relationship: The noise collected during readout is proportional to the square root of the receiver bandwidth: $\text{SNR} \propto \frac{1}{\sqrt{\text{rBW}}}$

1. Narrow Receiver Bandwidth

Reducing the receiver bandwidth (e.g., from $\pm 32\text{ kHz}$ to $\pm 16\text{ kHz}$) narrows the frequency window, which filters out random background noise.

• Advantages: SNR increases by approximately $41%$ ($\sqrt{2}$).
• Disadvantages: Readout time is longer, which increases the minimum Echo Time (TE) and limits the number of slices available per TR. It also increases chemical shift and magnetic susceptibility artifacts.

2. Wide Receiver Bandwidth

Increasing the receiver bandwidth collects more frequencies, including more background noise.

• Advantages: Readout is faster, permitting shorter minimum TE values and faster echo spacing in FSE. It also reduces chemical shift and metal susceptibility artifacts.
• Disadvantages: SNR is decreased.

---

Influence of TR and TE on SNR

Pulse sequence timing parameters directly influence the available longitudinal magnetization and the decay of transverse magnetization, thereby shaping SNR.

1. Repetition Time (TR)

• Longer TR: Allows the longitudinal magnetization vector ($M_z$) to recover more fully toward alignment with $B_0$ before the next RF excitation pulse is applied. Fuller recovery results in a larger transverse magnetization vector ($M_{xy}$) upon excitation, which yields a stronger signal and higher SNR.
• Tradeoff: Increasing TR increases the total scan time.

2. Echo Time (TE)

• Longer TE: Allows the transverse magnetization ($M_{xy}$) to undergo more dephasing via T2 or T2* relaxation before the echo is read. By the time the signal is collected, its amplitude has decayed, resulting in a lower SNR.
• Tradeoff: TE is kept short to maximize SNR in T1 and PD-weighted sequences.

---

Contrast-to-Noise Ratio (CNR)

While SNR measures overall signal relative to background noise, Contrast-to-Noise Ratio (CNR) measures the difference in SNR between two adjacent tissues:

$\text{CNR} = \text{SNR}_{\text{tissue A}} - \text{SNR}_{\text{tissue B}}$

CNR is the ultimate determinant of whether pathology can be visualized. A scan can have extremely high SNR, but if the pathology and surrounding normal tissues have the same signal intensity, the CNR is zero, and the lesion remains invisible.

Optimization Strategies for CNR:

• Pulse Sequence Selection: Using T2-weighted sequences makes fluid/edema (long T2) bright against dark background tissue.
• Inversion Recovery (IR): Suppressing background tissues (e.g., using STIR to null fat or FLAIR to null CSF) increases the relative contrast of lesions.
• Contrast Media: Gadolinium-based contrast agents (GBCAs) selectively shorten the T1 relaxation times of tissues, rendering them bright on T1-weighted images and improving CNR.
• Coil Selection: Using dedicated surface or phased-array coils improves local SNR, which directly improves CNR.

---

Parameter Tradeoff Summary Table