Comparison between different speckle tracking methods in ultrasound tongue images

Get Complete Project Material File(s) Now! »

Basic physics of medical ultrasound imaging

In brief, ultrasound image imaging is based upon the echo-location principle. However, the connection between the medical ultrasound sound and the echo-location principle was not made until the mature of the underwater acoustics, which made use of SONAR (Sound Navigation and Ranging) to measure the depth of water at sea. The inventions of SONAR and medical ultrasound imaging can be traced to the sinking of Titanic [30] when scientists tried to detect icebergs underwater using echo ranging. Nevertheless, at that time, there were no practical ways to implement the ideas until the discovery of piezoelectricity. In 1916 and 1917, by making use of technologies of piezoelectricity, Paul Langevin and Constantin Chilowsky invented a high-power echo-ranging system to detect the submarines [31].
The recognition that ultrasound could benefit medical diagnoses can be traced to World War I. Afterward, ultrasound is progressively applied to the therapy and the surgery. In the 1970s, medical ultrasound witnessed a rapid expanding with the advent of 2D real-time systems. Color flow systems occurred in 1990s. Presently, active research field includes contrast agents, molecular imaging, tissue characterization, integration with other modalities, such as photo-acoustic imaging [32].

Ultrasound-tongue tissue interaction

Since last several decades, B-mode ultrasound imaging has been employed to visualize the motion of the tongue with considerable success [29]. When the ultrasound waves travel through a medium, some effects may occur, which include attenuation, reflection, refraction and scattering. The interaction between the sound waves and the medium is determined by the acoustic properties of the medium and the ultrasound imaging system. In this section, firstly, we analyze the interaction between ultrasound and the tongue tissue. Then we present our data recording equipment used throughout this thesis.

Ultrasound tissue interaction

The quality of an ultrasound tongue image is of utmost importance in determining its usefulness. The overall quality of the ultrasound tongue image is the end product of a combination of many factors originating not only from the imaging system but also from the stability of the recording system and the performance of the operator. All of the components within the ultrasound tongue imaging system, including the transducer, image processing, display and recording devices, impact on the ultimate quality of the ultrasound tongue image. It is necessary to analyze the interaction between the ultrasound and tongue tissue, which can be helpful to make better use of ultrasound imaging.
When we are employing ultrasound to measure the motion of the tongue, the transducer typically is placed under the chin (as can be seen from the Figure 2-3). The ultrasound waves will transmit through the tongue body until it is reflected back.

Ultrasound tongue image distortions

Due to the sensitivity of the ultrasound imaging quality, there are several image distortions in ultrasound tongue images. In more detail, the distortions include speckle noise contamination, double edges, contour discontinuities, contour invisibility and inconsistent transducer placement and so on. Here, the focus is mainly on the influence of speckle and the contour invisibility. Speckle noise: Due to the inherent contamination with the speckle noise, the analysis of ultrasound tongue image poses a great challenge. Although there are more attempts to employ this kind of information to follow the motion, speckle noise processing is still an open issue in ultrasound image processing. Indeed, the speckle noise degrades the ultrasound tongue image, concealing the fine structures [36], which leads to the difficulties in the motion tracking of the tongue.
Hidden from the view (or faint contour): When the tongue tissue goes perpendicular to the propagation direction of ultrasound waves, the quality of image is good. While, when the tongue tissue goes parallel to the propagation direction of ultrasound waves, the quality of the image is poorly and missing occurs. Take the /i/ as an example, due to the natural motion of the tongue, part of the tongue is invisible. This kind of invisibility increased the difficulties of ultrasound tongue interpretation.

READ  Local connectivity and convolutional networks

Ultrasound tongue data acquisition

The ultrasound data acquisition devices used in this thesis belong to our SSI system’s data acquisition part, which is comprised of a helmet to hold the ultrasound probe to capture the movement of the tongue at the frame rate of 60 Hz, a VGA, CMOS industrial camera for the lips, a microphone to record acoustic speech signals, an electroglottograph (EGG) to measure and record vocal fold contact movement during speech production. All the different modes of data are recorded synchronously [37].
The lightweight, adjustable helmet (Figure 2-4) is fitted with a micro-convex, 1 inch diameter, 128 element ultrasound probe for tongue imaging [37]. An adjustable platform is used to hold the ultrasound transducer in contact with the skin beneath the chin. The ultrasound machine chosen is the Terason T3000, a system which is lightweight and portable yet retaining high image quality, and allowing data to be directly exported to a PC via Firewire. Our entire SSI system can be placed in a small carrying case, thus enabling everyday applications.

Table of contents :

Table of Contents
List of Figures
List of Tables
Chapter 1 Introduction
1.1 Silent speech interface concepts
1.2 Related work
1.3 The thesis work
1.4 Structure of the thesis
Chapter 2 Principles of ultrasound tongue imaging
2.1 Introduction
2.2 Basic principles of medical ultrasound imaging
2.2.1 Basic physics of medical ultrasound imaging
2.2.2 Ultrasound pulse
2.2.3 Ultrasound scan types
2.3 Ultrasound-tongue tissue interaction
2.3.1 Ultrasound tissue interaction
2.3.2 Ultrasound tongue image distortions
2.3.3 Ultrasound tongue data acquisition
2.4 Conclusion
Chapter 3 Speckle tracking in ultrasound tongue images
3.1 Introduction
3.2 Speckle tracking in ultrasound tongue images
3.2.1 Fundamentals of speckle tracking
3.2.2 Deformation registration
3.2.3 Optical flow
3.2.4 Local invariant feature
3.2.5 Comparison between different speckle tracking methods in ultrasound tongue images
3.3 Similarity-based automatic speckle tracking re-initialization
3.3.1 Ultrasound image similarity measurement
3.3.2 Ultrasound image-based speckle tracking re-initialization
3.4 Conclusion
Chapter 4 Contour tracking in ultrasound tongue images
4.1 Introduction
4.2 Active contour model with Contour group-similarity constraint
4.2.1 Active contour model with contour group-similarity constraint
4.2.2 Automatic re-initialization during contour tracking
4.2.3 Experiments and results
4.3 A comparative study on the different contour tracking algorithms
4.3.1 Comparison of contour tracking methods with re-initialization
4.3.2 Similarity-based contour extraction
4.4 Conclusion
Chapter 5 Physics-based 3D tongue motion modeling
5.1 Introduction
5.2 Physics-based 3D tongue modeling
5.2.1 Theoretical foundations of motion-driven based 3D tongue modeling
5.2.2 Interface overview
5.3 Speckle tracking-based tongue motion simulation
5.3.1 Speckle tracking-based tongue motion visualization
5.3.2 Experimental results
5.4 Contour-guided 3D tongue motion visualization
5.4.1 Contour-based 3D tongue motion visualization
5.4.2 Experimental results
5.5 Conclusion
Chapter 6 Conclusions
6.1 Conclusions
6.2 Perspectives


Related Posts