Review
A review of calibration techniques for freehand 3-D ultrasound systems

https://doi.org/10.1016/j.ultrasmedbio.2004.11.015Get rights and content

Abstract

Three-dimensional (3-D) ultrasound (US) is an emerging new technology with numerous clinical applications. Ultrasound probe calibration is an obligatory step to build 3-D volumes from 2-D images acquired in a freehand US system. The role of calibration is to find the mathematical transformation that converts the 2-D coordinates of pixels in the US image into 3-D coordinates in the frame of reference of a position sensor attached to the US probe. This article is a comprehensive review of what has been published in the field of US probe calibration for 3-D US. The article covers the topics of tracking technologies, US image acquisition, phantom design, speed of sound issues, feature extraction, least-squares minimization, temporal calibration, calibration evaluation techniques and phantom comparisons. The calibration phantoms and methods have also been classified in tables to give a better overview of the existing methods. (E-mail: [email protected])

Introduction

Ultrasound (US) is an appealing imaging modality because it is relatively inexpensive, safe, noninvasive, compact, portable and can image in real-time almost any body tissue. For these reasons, US is widely used and even gaining popularity in fields such as intraoperative imaging. Conventional US is a 2-D modality, in contrast to computed tomography (CT), magnetic resonance imaging (MRI) and other modalities that are volumetric. Three-dimensional (3-D) US is an emerging new technology that has many advantages over 2-D imaging: it allows the direct visualization of 3-D anatomy; 2-D slice views can be generated at arbitrary orientations; and volume and other measurements may be obtained more accurately. Measuring the volume of the prostate (Crivianu-Gaita et al. 1997; Hoffmann et al. 2003), monitoring fetal development (Kelly et al. 1994) or evaluating brain shift during neurosurgery (Comeau et al. 2000; Unsgaard et al. 2002) are examples of applications for 3-D US. For a more detailed list of applications, refer to Nelson and Pretorius (1998).

There are four general methods to construct an US volume. They are classified into the following categories: 1. constrained sweeping techniques, 2. 3-D probes, 3. sensorless techniques, and 4. 2-D tracked probe (also known as “freehand”) techniques. They can be described as follows:

  • 1

    The constrained sweeping systems are characterized by a spatially predefined, constrained sweeping of the entire 2-D probe body that can be accomplished with a motor attached to the probe. Slices are generally either acquired in a wedge (fan-like) pattern, in a series of parallel slices (translation, as for MRI/CT), or with a rotation around a central axis (Fenster and Downey 2000).

  • 2

    3-D US probes usually consist of 2-D arrays that allow explicit imaging in 3-D. These probes are relatively large and expensive in comparison with 2-D probes and their image resolution is not as good as their 2-D counterparts; refer to Light et al. (1998) for more information. Other 3-D probes can be either mechanically or electronically steered within the probe housing. An annular array producing a thin US beam can be accurately controlled by an internal mechanical motor in 2-D, to obtain a 3-D volume with high resolution. 2-D probes can also be electronically steered within the image plane to increase the field-of-view (FOV), as in Rohling et al. (2003).

  • 3

    The sensorless techniques attempt to estimate the 3-D position and orientation of a probe in space. Pennec et al. (2003), for example, proposed a system where a time sequence of 3-D US volumes is registered to play the role of a tracking system. Sensorless tracking can be done by analyzing the speckle in the US images using decorrelation (Tuthill et al. 1998) or linear regression (Prager et al. 2003). However, Li et al. (2002) found that it was impossible to accomplish real freehand scanning using only speckle correlation analysis. Although the Prager et al. (2003) results are encouraging, their sensorless approach is still far from the accuracy obtained with tracked probes.

  • 4

    Freehand systems allow image acquisition with unconstrained movement. They generally consist of a sensor (attached to a probe) that is tracked by a device that calculates the sensor’s position and orientation at any point in time. This information is used to compute the 3-D coordinates of each pixel of the US images. Locating US images within a tracked coordinate system opens up a new world of possibilities: the images can be registered to a patient and to images from other modalities (Arbel et al. 2001; Brendel et al. 2002; Comeau et al. 2000; Dey et al. 2002; Lindseth et al. 2003b; Lindseth et al. 2003c). All the tracking devices used for freehand systems work in a similar manner: the device tracks the position and orientation (pose) of the sensor on the probe, not the US image plane itself. So, an additional step must be added to compute the transformation (rotation, translation and scaling) between the origin of the sensor mounted on the probe and the image plane itself. The process of finding this transformation is called calibration and is the focus of this article.

The objective of this paper is to review what has been published in the past 10 years in the field of calibration techniques for freehand 3-D US systems. The first section of this article covers the tracking technologies and the second section covers the acquisition of US images. The third section introduces calibration and all aspects of the problem. The fourth section discusses the methods used to test the calibration. Finally, the fifth section summarizes the results obtained by the different research groups. The main contribution of this paper is the comprehensive review and classification of all the different calibration techniques.

Section snippets

Tracking

There are four common technologies to track medical instruments, 1. mechanical, 2. acoustical, 3. electromagnetic, and 4. optical. The role of a tracking system in the context of 3-D US is to determine the position and orientation of a sensor attached to the US probe. After calibration is performed, every pixel in each 2-D image is mapped in the 3-D coordinate system of the tracking device to reconstruct a geometrically correct volume. In the following paragraphs are brief descriptions of the

Image acquisition

There are two common solutions to transfer images from an US machine to a computer. The most popular technique is to connect the analog output (e.g., composite video, S-video) of an US machine to a frame-grabbing card on a computer (Comeau et al. 2000; Detmer et al. 1994; Meairs et al. 2000). The second method is to directly acquire digital images from the US machine, often by connecting them through a network cable (Barratt et al. 2001a; Berg et al. 1999; Lindseth et al. 2003c). Barry et al.

Calibration

This section defines, in a more detailed and graphic manner, what is involved for calibration. We begin with a short summary of the calibration process and then describe each step in detail. Figure 2 illustrates the coordinate systems used for locating the image plane of an US probe in space. Three coordinate systems are represented in Fig. 2, that of the position sensing device (called “world coordinate system”), that of the sensor mounted on the probe and that of the image plane. In Fig. 2,

Calibration evaluation

This section summarizes the main techniques to evaluate the in vitro precision and accuracy of a calibration transformation. These tests are sometimes performed on the same phantoms as the one used for calibration (Detmer et al. 1994), but preferably on a different phantom (Lindseth et al. 2003c; Treece et al. 2003) to minimize bias. Precision is estimated when multiple measurements of the same phenomenon are compared with themselves. Accuracy is estimated when the measurements are compared

Results

It would be very practical if all authors would agree on which calibration technique is the best. Yet, the criteria for defining what is best differs among groups, depending mostly on the type of application of the tracked US images. It would probably be safe to say that most authors agree on the fact that accuracy and precision are the most important evaluation criteria. However, as was seen in the previous section, these criteria can be evaluated in many different ways. Other factors, such as

Conclusion

US probe calibration is an obligatory step to build 3-D volumes from 2-D images acquired in a freehand US system. Calibration finds the transformation that relates the image plane to the sensor attached on the probe. Many authors have included a brief literature review in their articles, but this paper is the first comprehensive overview of what has been done in the field. Calibration is a process with multiple components that were covered in the various sections of this review. A very brief

Acknowledgements-

This work was supported by a grant by IRIS-PRECARN TULIP project (Canada) and by grants from the Research Council of Norway and the Norwegian Ministry of Health and Social Affairs.

References (109)

  • D.F. Leotta

    An efficient calibration method for freehand 3-D ultrasound imaging systems

    Ultrasound Med Biol

    (2004)
  • D.F. Leotta et al.

    Performance of a miniature magnetic position sensor for three-dimensional ultrasound imaging

    Ultrasound Med Biol

    (1997)
  • F. Lindseth et al.

    A robust and automatic method for evaluating accuracy in 3-D ultrasound-based navigation

    Ultrasound Med Biol

    (2003)
  • F. Lindseth et al.

    Probe calibration for freehand 3-D ultrasound

    Ultrasound Med Biol

    (2003)
  • K. Martin et al.

    Measurement of the speed of sound in ethanol/water mixtures

    Ultrasound Med Biol

    (2001)
  • S. Meairs et al.

    Reconstruction and visualization of irregularly sampled three- and four-dimensional ultrasound data for cerebrovascular applications

    Ultrasound Med Biol

    (2000)
  • D.M. Muratore et al.

    Beam calibration without a phantom for creating a 3-D freehand ultrasound system

    Ultrasound Med Biol

    (2001)
  • T.R. Nelson et al.

    Three-dimensional ultrasound imaging

    Ultrasound Med Biol

    (1998)
  • N. Pagoulatos et al.

    A fast calibration method for 3-D tracking of ultrasound images using a spatial localizer

    Ultrasound Med Biol

    (2001)
  • X. Pennec et al.

    Tracking brain deformation in time sequences of 3D US images

    Pattern Recog Lett

    (2003)
  • R.W. Prager et al.

    Sensorless freehand 3-D ultrasound using regression of the echo intensity

    Ultrasound Med Biol

    (2003)
  • R.W. Prager et al.

    Rapid calibration for 3-D freehand ultrasound

    Ultrasound Med Biol

    (1998)
  • R.N. Rohling et al.

    Three-dimensional spatial compounding of ultrasound images

    Med Image Anal

    (1997)
  • G.M. Treece et al.

    High-definition freehand 3-D ultrasound

    Ultrasound Med Biol

    (2003)
  • J.W. Trobaugh et al.

    Frameless stereotactic ultrasonographyMethod and applications

    Comput Med Imaging Graph

    (1994)
  • J.W. Trobaugh et al.

    Three-dimensional imaging with stereotactic ultrasonography

    Comput Med Imaging Graph

    (1994)
  • D.V. Amin et al.

    Calibration method for determining the physical location of the ultrasound image plane

  • M.E. Anderson et al.

    The impact of sound speed errors on medical ultrasound imaging

    J Acoust Soc Am

    (2000)
  • T. Arbel et al.

    Automatic non-linear MRI-ultrasound registration for the correction of intra-operative brain deformations

  • K.S. Arun et al.

    Least-squares fitting of two 3-D point sets

    IEEE Trans Pattern Anal Machine Intell

    (1987)
  • R.A. Beasley et al.

    Registration of ultrasound images

    SPIE Proc

    (1999)
  • P. Besl et al.

    A method for registration of 3-D shapes

    IEEE Trans Pattern Anal Machine Intell

    (1992)
  • N. Bilaniuk et al.

    Speed of sound in pure water as a function of temperature

    J Acoust Soc Am

    (1993)
  • W. Birkfellner et al.

    Systematic distortions in magnetic position digitizers

    Med Phys

    (1998)
  • J.M. Blackall et al.

    An image registration approach to automated calibration for freehand 3D ultrasound

  • E.M. Boctor et al.

    A rapid calibration method for registration and 3D tracking of ultrasound images using spatial localizer

    SPIE Proc

    (2003)
  • L.G. Bouchet et al.

    Calibration of three-dimensional ultrasound images for image-guided radiation therapy

    Phys Med Biol

    (2001)
  • B. Brendel et al.

    Registration of 3D CT and ultrasound datasets of the spine using bone structures

    Comput Aided Surg

    (2002)
  • R.A. Brown

    A stereotactic head frame for use with CT body scanners

    Invest Radiol

    (1979)
  • J. Canny

    A computational approach to edge detection

    IEEE Trans Pattern Anal Machine Intell

    (1986)
  • J.C. Carr

    Surface reconstruction in 3D medical imaging

    (1996)
  • J.C. Carr et al.

    Design of a clinical free-hand 3D ultrasound system

    SPIE Proc

    (2000)
  • F. Chassat et al.

    An experimental protocol for accuracy evaluation of 6D localizers for computer-assisted surgeryApplication to four optical localizers

  • P. Cinquin et al.

    Computer assisted medical interventions

    IEEE Eng Med Biol Magazine

    (1995)
  • R.M. Comeau et al.

    Integrated MR and ultrasound imaging for improved image guidance in neurosurgery

    SPIE Proc

    (1998)
  • R.M. Comeau et al.

    Intraoperative ultrasound for guidance and tissue shift correction in image-guided neurosurgery

    Med Phys

    (2000)
  • R. Deriche

    Optimal edge detection using recursive filtering

  • R. Deriche

    Fast algorithms for low-level vision

    IEEE Trans Pattern Anal Machine Intell

    (1990)
  • D. Dey et al.

    Automatic fusion of freehand endoscopic brain images to three-dimensional surfacesCreating stereoscopic panoramas

    IEEE Trans Med Imaging

    (2002)
  • D.W. Eggert et al.

    Estimating 3-D rigid body transformationsA comparison of four major algorithms

    Machine Vision Applic

    (1997)
  • Cited by (137)

    View all citing articles on Scopus
    View full text