Exercise 2

Topic: Filtering by harmonic imaging.

The quality of an ultrasound image is often corrupted by “acoustic noise”, this is particularly a problem when we are imaging patients that have much subcutaneous fat. The most important causes of acoustic noise are:

1. Reverberations: Multiple echoes between fat/muscle tissues. These echoes will reach the transducer with a larger time delay, which gives “fake” echoes further down in the image.

2. Sidelobes in transmitted/received beam which gives signal from scatterers at other angles than the focus direction.

The acoustic pressure of the ultrasound beam could be several MPascal without damaging the tissue of the patient. These high intensities cause harmonic distortion in the transmitted wave. The distortion increases with propagation distance, and is highest at the focus of the beam. Harmonic components in the wave have considerable lower sidelobe level and less multiple echoes from tissue in the body than components of the fundamental frequency.

Modern ultrasound transducers have 60-100 % relative bandwidth. We are able to receive and isolate second harmonic echoes that is caused by nonlinear propagation, if we send a pulse (without second harmonic distortion) with frequency in the lower part of the frequency band. This is a method to reduce the acoustic noise, and improve the quality of the image.

The Matlab file data_Oving2.mat  contains received RF-signal from a sector scanning of a heart using a phased array transducer. The transducer bandwidth is 1.0 – 4.0 MHz.

RF:  2D array of the received signal

SXmit: transmitted acoustic signal measured with a hydrophone near the transducer

t: “time-axis”, an array containing sample times for s and rf(:,k). (k is the beam number)

dfi: angle between the beams (radians)

Info: information (text)

The task is to construct a linear time invariant filter that isolates the second harmonic parts of the received signal. The filter shall be used to make a gray scale ultrasound image. Compare the quality of the image with and without filtering. It is not necessary to scan-convert the images. An appropriate format could be depth at the vertical axis and angle at the horizontal axis. 

Hint: Use logarithmic compression of the detected RF-signal over a dynamic range of ~40dB. The Matlab commando (gray(256)) gives 256 gray scales for pixel values 0-255.

Who makes the best image?