The Beamformer class
When instantiating a beamformer class with the settings from the Reader, we
expect to store values for all the beamformer.setups and the
beamformer.options.
Beamformer setups
The setups are here to store the main information. The list of the setups is defined within beamformers/setups.py, but these are:
Setup name |
Type |
Default |
Description |
|---|---|---|---|
emitted_probe received_probe |
array_like |
These are the probes used to acquire the data, they both have the shape (3, nb_transmissions, nb_elements). The first dimension is the x, y and z positions of the elements, and, for each transmissions, the concerned elements. This configuration allows to have distinct emission and reception probe (which is useful for STA for example). Stored in m. |
|
delays |
array_like |
The acquisition delays for each cycle for each element. If we are sending tilted plane waves, the delays are stored here (in s). |
|
transmissions_idx |
array_like |
The indices for the transmissions, as integers, if you don’t want to perform the beamforming on all transmissions (single plane wave vs full compounded for example). |
|
sound_speed |
float |
The speed of sound used to compute the delays, or simply an estimation of the medium sound of speed if there’s no delays (in m/s). |
|
f_number |
float |
The f-number to use for the beamforming. |
|
t0 |
float |
The initial time to add to the transmission times (in s). |
|
sampling_freq |
float |
The sampling frequency of the acquisition system (in Hz). |
|
central_freq |
float |
The central frequency of the probe (in Hz). |
|
signal_duration |
float |
The duration of the emitted signal, it is used to adjust the original t0 to the center of the emission pulse. |
|
bandwidth |
float |
The bandwidth of the probe, in %. |
|
prf |
float |
The Pulse Repetition Frequency (in Hz). |
|
emitted_thetas received_thetas |
array_like |
[0.] |
For convex probes, this indicates the thetas of the probe for each emitted / received probe elements. |
Each of these setups can be updated at any time using:
1beamformer.update_setup('name_of_setup', value)
Beamformer options
The options are here to define the beamforming options, not all of them are implemented in all the beamformers though:
Option name |
Type |
Default |
Description |
|---|---|---|---|
interpolation |
int str |
‘linear’ |
The interpolation method to use (can be either ‘none’, ‘linear’, ‘quadratic’ or ‘cubic’). The latter two are only implemented in Numpy (CPU). |
reduction |
int str |
‘sum’ |
The reduction method, either ‘sum’ or ‘mean’. This is defining the way we are reducing the delays at the end of the beamforming process. |
rx_apodization |
int str |
‘boxcar’ |
The reception apodization method, either ‘boxcar’ or ‘tukey’, only ‘boxcar’ is supported by Numpy. |
rx_apodization_alpha |
float |
0.1 |
The coefficient for the apodization method. |
compound |
int bool |
True |
If set to True, we compound the data along the transmission axis. Can be combined with reduce. |
reduce |
int bool |
True |
If set to True, we reduce the data along the received probe elements. Can be combined with compound. Only supported for DAS beamformer. |
fix_t0 |
int bool |
True |
If set to True, the t0 is automatically set to
|
emitted_aperture |
int bool |
True |
If set to True, we consider the emitted elements also have an aperture. Else case, only the received elements are affected by the f-number (like MUST). |
Same as for the setups, you can update the options anytime using:
1beamformer.update_option('name_of_option', value)
Compatibility
As specified above, all the options are not supported by all the beamformers, or the libraries (Numpy / Numba for CPU, and cupy for GPU). This section is listing the current support. If nothing is specified, they are all supported.
Compatibility of the interpolation option:
Library |
None |
Linear |
Cubic |
Quadratic |
|
|---|---|---|---|---|---|
CPU |
Numpy |
OK |
OK |
OK |
OK |
Numba |
OK |
OK |
|||
GPU |
cupy |
OK |
OK |
Compatibility of the rx_apodization option:
Library |
boxcar |
tukey |
|
|---|---|---|---|
CPU |
Numpy |
OK |
|
Numba |
OK |
OK |
|
GPU |
cupy |
OK |
OK |
Compatibility of the compound option:
Beamformer |
Reduction |
Comments |
|---|---|---|
DAS |
Sum |
OK |
Mean |
The ‘no compound’ option is supported for Delay And Mean, but note that all the transmissions will be preserved. This means that if, for a given pixel, one of the transmission is not solicited (due to the aperture of the element, or the transmission angle or whatever), its dedicated transmission will be set to 0, but there won’t be any information about if it has been solicited or not. Therefore, if you want to perform the mean of these transmissions by yourself, the result might be different from the one with the ‘compound’ option. |
|
FDMAS |
Both |
The FDMAS algorithm performs the non-linear root / square operations on the reduced raw-focused data (summed or averaged). This means that the sum of the beamformed transmissions is not equal to the compounded beamforming. |
p-DAS |
Both |
Same as the FDMAS algorithm, the non-linear operations of p-DAS are also performed on reduced data. |
Compatibility of the reduce option:
Beamformer |
Reduction |
Comments |
|---|---|---|
DAS |
Sum |
OK |
Mean |
The ‘no reduction’ option is supported for Delay And Mean, but note that all the probe elements will be preserved. This means that if, for a given pixel, one of the probe element is not solicited (due to the aperture of the element defined with the f#), its dedicated element will be set to 0, but there won’t be any information about if the element has been solicited or not. Therefore, if you want to perform the mean of these elements by yourself, the result might be different from the one with the ‘reduce’ option. |
|
FDMAS |
Both |
No support, as the FDMAS reduction is a core part of the algorithm, this would lead to the same result as the DAS |
p-DAS |
Both |
No support, for the same reason as FDMAS |