Ultrasound Image Computation
After the beam formed RF data passes through all its processing stages, it becomes ready as an input to ultrasound image computation processes. Different types of ultrasound images are used in clinical practice, each conveying a particular type of information to the clinician.
The A-Mode data is the most basic type of ultrasound image derived from RF data, and forms the basis for other types of ultrasound images. With the analogy to telecommunications in mind, the A-Mode data is the result of amplitude demodulation of the RF data. In simpler terms, it is a signal that captures the variation in the amplitude of the RF data. The A-Mode data is therefore a 1D array of data, and strictly speaking not an image.
The RF data is an alternating sinusoid whose amplitude and phase are changing. The alternations by themselves carry little information about the structure of tissue being image. On the other hand, the amplitude of the sinusoid shows how much reflection and backscattering has happened at a particular depth in tissue. Therefore the amplitude demodulation is carried out to take out the alternations, and leave the amplitude signal in tact.
Techniques for amplitude demodulation in ultrasound imaging are adapted from telecommunications. The two most popular methods are through the Hilbert Transform, and IQ demodulation. In version 6.0.x of the Exam software, the demodulation technique is IQ Demodulation. It also includes filtering of the RF data in the base band. The cutoff frequency and the order of the filter can be adjusted flexibly through Exam B-IQFILT.
Traditional B-Mode images are 2D cross-sectional images of tissue acquired using 1D transducer arrays. A cross sectional image of the tissue can be formed by placing the A-Mode data for successive scan-lines side by side to form a 2D array of data. However, a number of processing steps are carried out on this array, before it becomes a presentable B-Mode image. These steps are described here.
Logarithmic Compression and Adjustment in the Dynamic Range
The variation in the amplitude of the RF data is relatively high. If the same image is mapped linearly to a 0-255 gray scale image, many important tissue structures will only have image values of 0-9. In other words, the presence of a few very high amplitude points in the image overshadows everything else. In order to achieve a balance (or a more even histogram), the amplitude values are mapped nonlinearly by a logarithmic-looking function which adjusts the dynamic range.
In version 6.0.x of the Exam software, the dynamic range adjustment is performed by dividing the range of values into two portions at a pivot point. Below the pivot point a logarithmic compression is used, and above the pivot point a linear mapping is used. The parameters for both parts can be adjusted flexibly through Exam B-DYNRNG .
From a clinical perspective, it is very important that the geometry of the B-mode image shows the exact geometry of the tissue being imaged. For instance if a tumor has a spherical shape in reality, it is essential that its cross sectional B-mode image has a circular profile, and for instance, not an elliptical profile.
The B-mode image obtained by putting together the A-lines, however, does NOT have this property. As this image does not take into account the geometry of the transducer, and the manner in which the RF data has been acquired, the generated image, although containing information about the structure of tissue, is not geometrically correct. Scan conversion is the process through which the data is mapped to the actual geometry of the tissue.
As an example, assume that a curved transducer has been used to acquire the data. In this case, the samples of the data in the array come from a curvi-linear grid as shown in the figure. The coordinates of the data on the transducer grid are:
- The sample number, determined by the sampling frequency, decimation factors, etc.
- The line number, determined by the line density, etc.
The B-mode image in the transducer grid is called the pre scan converted B-mode image.
In order to obtain a geometrically correct image, the pre scan converted image needs to be interpolated into the display grid, which represents the display pixels on the LCD screen of the system. Therefore, scan conversion is essentially an interpolation process. The importance of the LCD pixels is that if the pixels are not square in shape, this should be taken into account in the image generation. Otherwise the final image would appear skewed on the screen. The figure shows a typical display grid (dashed lines) overlaid on the transducer grid.
In order to perform the scan conversion, the geometry of the transducer grid and the screen grid must be known, as well as their location with respect to each other. The Exam software has some parameters as inputs to scan conversion. See Exam B-SCVT .
The B-mode ultrasound images usually have a grainy look, unique to ultrasound images. The grainy pattern comes from multiple scattering of the waves from the scattereres in tissue and their constructive-destructive interference. These grains are called speckles in the jargon. To enhance the quality of the B-mode images, image processing algorithms are used to reduce the speckles. These algorithms are usually called speckle reduction algorithms.
The speckle patterns differ from one tissue type to another as well as from healthy to diseased tissue. They help the clinicians in the diagnosis. Therefore speckle reduction is an art which retains these information, while enhancing the appearance of the images. Ultrasonix uses proprietary software for the image enhancement, and cannot distribute it to its research customers. However the Amplio SDK provides some in-house developed speckle reduction filters.
M-Mode images are images that show the motion of different points of tissue along a SINGLE scan line as a function of time. To generate the M-Mode images, the A-mode data from successive acquisitions (frames) of the same scan line are placed side by side in an array as an image. The image is updated in real-time as newer data become available. As can be seen from this description, after the A-line has been computed, there is really not further processing needed on the data. Therefore the Exam software does not have any particular parameters related to the processing of the M-mode images. See Exam M.
Pulse(d) Wave Doppler Images
Pulsed Waved Doppler Imaging, or PW Mode, is a method to use ultrasound for determining blood velocity and direction. The RF data goes through following steps to generate the PWD images:
- The acquired RF echo signals go through a demodulation to generate in-phase and quadrature components (I/Q).
- I/Q signals then go through a high pass filter, also known as a wall filter, to remove strong slower motions.
- The wall filtered signals are then used to generate the spectrum using short time Fourier analysis.
Power/Color Doppler Images
- Exam Parameters for CDI processing:
See also: Elastography