ultrasound_metrics.channel.noise

ultrasound_metrics.channel.noise#

Estimate the signal-to-noise ratio (SNR) of an ultrasound transducer system.

This module provides methods to calculate SNR using repeated pulse-echo measurements of a phantom or tissue.

Notes

Motivation:

We would like to understand the power/SNR tradeoffs, e.g.: - How much does AFE power improve SNR? - How much does combining channels via summation improve SNR? - How much do increasing chip-repeats improve SNR?

Implemented Methods:
Repeated Measurements Variance (best for a static phantom):

  • Acquire multiple identical frames (same configuration)

  • Calculate the mean signal across frames (this represents the true signal)

  • Calculate the standard deviation across frames (this represents noise)

  • SNR = mean signal / standard deviation of noise

Signal Differences in Adjacent Frames:

  • Since tissue moves slowly and we acquire ultrafast ultrasound:

  • Calculate frame-to-frame differences

  • Assuming the tissue moves much slower than the frame rate, frame-to-frame differences primarily represent transmit and receive-channel noise

  • Compare mean signal amplitude to this difference (noise) to estimate SNR

Alternative Methods (Not Implemented):
Transmit-Off Noise Measurements:

  • Collect data with the transmitter turned off (tx-off)

  • This is all noise, so it’s easy to calculate SNR.

  • Downside: noise may be nonlinear or signal-dependent, so this would not capture the practical SNR.

Spectral Analysis:

  • Assumes that the signal is narrowband and that the noise is flat.

  • This is not true for our data, so we don’t implement this.

Differences from B-Mode SNR:

In ultrasound or image analysis, we often calculate SNR in the B-mode image. For example, checking the strength of a reflector in the image.

You might take a spatial region analysis approach, e.g.:

  • Select regions devoid of signal (e.g. water) to calculate noise power

  • If a reflector is present, you can also calculate signal power

This is a great way to calculate image-related SNR. However, it does not directly characterize the electronic noise, e.g. thermal or quantization noise. Scatterer SNR probably belongs in a different submodule of this same repository.

Attributes#

key

Classes#

NoiseEstimationMethod

Enumeration of noise estimation methods.

Module Contents#

class ultrasound_metrics.channel.noise.NoiseEstimationMethod[source]#

Bases: str, enum.Enum

Enumeration of noise estimation methods.

REPEATED_MEASUREMENTS[source]#

Use repeated measurements variance method.

ADJACENT_FRAMES[source]#

Use adjacent frames difference method.

Initialize self.  See help(type(self)) for accurate signature.
ADJACENT_FRAMES = 'adjacent_frames'[source]#
REPEATED_MEASUREMENTS = 'repeated_measurements'[source]#
ultrasound_metrics.channel.noise.key[source]#