Search Contact information
University of Cambridge Home Department of Engineering
University of Cambridge > Engineering Department

Three dimensional ultrasonic elasticity imaging

EPSRC Grant EP/E030882/1

October 2007 - March 2011


Overview

This is a project funded by the EPSRC to look at various aspects of 3D ultrasonic elasticity imaging. Ultrasonic strain (or elasticity) imaging is a technique by which the mechanical properties of tissue can be visualised, giving access to information that would not be available in a conventional ultrasound scan. Strain images are derived from measurements of displacement between pairs of RF ultrasound images. Our research focuses on the quasistatic approach to elasticity imaging, where displacements are caused by small hand-induced changes in axial probe pressure. The project is a collaboration between our group and a group led by Dr Jeff Bamber at the Institute of Cancer Research (ICR), Sutton.

The objectives of the project are mainly 3D elasticity imaging, although many of our developments are applicable to both 2D and 3D imaging. 3D strain imaging offers many of the same advantages as conventional 3D ultrasound, for example more complete visualisation of the anatomy and more accurate volume measurement. In addition, the availability of volumes of RF data potentially allows displacements to be tracked more accurately by using 3D windows and taking account of displacements in the out of plane direction. Also, the ability to calculate 3D displacement vectors may lead to developments in quantitative imaging of mechanical properties other than strain.

The project looks at several potential clinical applications of ultrasonic strain imaging, including applications in breast cancer and endometrial cancer detection.

The project objectives are:


Achievements

3D acquisition

We have developed three distinct approaches to 3D strain data acquisition:

Freehand 3D acquisition -
A volume of data is acquired by gradually moving a conventional 2D ultrasound probe elevationally to sweep out a volume while acquiring 2D strain data [1]. A position sensor records the location of each image in 3D space, as in ordinary freehand 3D ultrasound. Each adjacent image pair provides one 2D strain image and together they make up a volume of axial strain data.
Volume pair acquisition -
Whole volumes of data are recorded using a mechanically-swept 3D probe [2, 12]. The probe is held stationary while recording each volume and the axial compression is changed by a small amount between volumes. Strain data is then calculated by comparing two overlaid volumes.
Hybrid 3D acquisition -
A technique combining features of the first two [9, 22, 30]. A mechanically-swept 3D probe sweeps out only a single volume, acquiring several RF frames at each motor step, while applying a continuous up and down axial motion. This is similar to the freehand method but with the elevational motion controlled by the probe motor.

In developing these techniques, we have also compared the strain image quality achievable with each. An important part of this project was to make use of volume data to improve the accuracy of the strain estimates and therefore enhance the image quality. The volume-pair scanning approach does this by using 3D windows for displacement estimation and by tracking displacements in 3D rather than just in-plane. In comparison, the other methods are restricted to 2D in-plane displacement estimation and are therefore at an obvious disadvantage in terms of image quality. However, we have found that the practicality of the scanning technique is also an important factor affecting image quality, as the figures below show.

quality comparison 1

The figure shows a comparison of image quality for the first two methods [5, 14, 26]. The best quality, on the left, is from a freehand 3D scan. In order to obtain this image, the probe was moved slowly and carefully in the elevational direction, resulting in a high density of good quality strain estimates but at the cost of a difficult scanning technique. The worst quality, on the right, is also a freehand 3D scan. Here, the probe was moved more quickly in the elevational direction, so the data was easier to record, but the quality is noticeably worse. In the middle, the data was recorded using the volume-pair method, with a frame density less than either of the freehand scans. Despite this, the quality is reasonable and for comparable quality the scanning technique is easier than that of the freehand method.

quality comparison 2

The above figure is a comparison of the volume-pair method to the hybrid method [9, 22, 30]: the hybrid method is shown on the left and the volume-pair on the right. The feature of the freehand method that makes recording of high quality data difficult is the manual control of the elevational motion. The hybrid method avoids this difficulty and so is an easier scanning technique. We have found that the hybrid method is most reliable at producing high quality data.

Signal processing developments

We have made several developments in the algorithms and processing used to generate the strain images. This has led to better strain images in both 2D and 3D. Key results are shown below.

hybrid displacement
estimation

In the displacement estimation process, the accuracy of axial displacements is affected to some extent by the accuracy of lateral and elevational displacements. Initialising an axial search at the wrong lateral or elevational location can result in peak-hopping errors in the axial displacement estimates. We have developed a multi-level tracking algorithm in which the lateral and elevational displacements are estimated at lower resolution for improved accuracy and robustness of the more important axial displacements [6, 19, 25]. The figure above shows the lateral displacement and axial strain for this method (the hybrid method) compared to several other displacement estimation strategies (details in [6]) in a scan of the testis. There are fewer peak-hopping errors using the hybrid method.

weighting scheme

There are several post-processing steps in generating a strain image, specifically normalisation, spatial filtering (smoothing) and temporal filtering (persistence). The final image is affected by the way in which each of these processes weights the raw data. We have investigated several different weighting schemes for each stage of the processing [18, 29]. The figure above shows several different ways of weighting the data in the first row and the effect these have on each stage of the processing in the following rows. The best result (4th column) is obtained by using the SNRe weighting for normalisation and the combined variance weighting for the subsequent stages. Details of each weighting scheme are in [29].

shear strain normalisation

Finally, we have developed a normalisation method suitable for shear strain imaging [8, 23, 28]. Existing axial strain normalisation methods are based on normalising by the average value or a similar function varying over the image. These are not suitable for shear strain because the average value is usually close to zero. The figure above shows normalised and unnormalised shear strain for a stiff inclusion. The normalised image sequence is more constant over time, even when the compression changes to uncompression, and so is more suitable for producing a time averaged image. Also, the arrangement of dark and light regions of shear strain around the inclusion is able to show that this is a stiff rather than a soft inclusion.

Clinical use

The methods above have been developed into a clinical system which we have used to investigate several possible application areas in 2D and 3D imaging [7]. The figures below show areas where we have found strain imaging to be clinically useful.

axillary lymph nodes

Standard practice in breast cancer treatment is to scan the axillary lymph nodes using conventional ultrasound, in order to identify cases where the cancer has spread to the lymph nodes. We have investigated the use of strain imaging to examine the lymph nodes [21]. The above figure shows a benign and a malignant case, both of which appear benign in the B-mode images but are clearly different in the strain images. The arrows show the locations of the lymph nodes. The results of this study suggest we get improved sensitivity of malignant lymph node detection using strain imaging.

endometrium

Conventional trans-vaginal ultrasound imaging is used to assess cases of post-menopausal bleeding, which can be caused by both benign and malignant conditions. We have used strain imaging to compare benign and malignant cases [11]: the figure above shows an example of each with benign polyps on the left and deeply invasive endometrial carcinoma on the right. The arrows show the boundary between the endometrium and myometrium. In the benign case, the softer endometrium is distinct from the myometrium, whereas they are indistinguishable in the strain image for the malignant case. This may be useful as a way to confirm benign disease.

3D fibroids

We have also investigated 3D strain imaging in trans-vaginal ultrasound and have found the enhanced visualisation to be useful for imaging fibroids [24]. The 3D visualisation may be useful for distinguishing sub-mucosal and intra-mural fibroids. The figure above shows a 3D view and three orthogonal views with fibroids clearly visible as stiffer (darker) regions.

All of the data generated from these clinical trials, including the examples above, are stored in our clinical database [13].


Related publications

Journal articles

[1] J. E. Lindop, G. M. Treece, A. H. Gee and R. W. Prager.
3D elastography using freehand ultrasound.
Ultrasound in Medicine and Biology, 32(4):529-545, April 2006.

[2] G. M. Treece, J. E. Lindop, A. H. Gee and R. W. Prager.
Freehand ultrasound elastography with a 3-D probe.
Ultrasound in Medicine and Biology, 34(3):463-474, March 2008.

[3] A. Gee, J. Lindop, G. Treece, R. Prager and S. Freeman.
Stable, intelligible ultrasonic strain imaging.
Ultrasound, 16(4):187-192, November 2008.

[4] G. M. Treece, J. E. Lindop, A. H. Gee and R. W. Prager.
Uniform precision ultrasound strain imaging.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 56(11):2420-2436, November 2009.

[5] R. J. Housden, A. H. Gee, G. M. Treece and R. W. Prager.
3-D ultrasonic strain imaging using freehand scanning and a mechanically-swept probe.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 57(2):501-506, February 2010.

[6] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
A hybrid displacement estimation method for ultrasonic elasticity imaging.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 57(4):866-882, April 2010.

[7] L. Chen, S. J. Freeman, A. H. Gee, R. J. Housden, R. W. Prager, R. Sinnatamby and G. M. Treece.
Initial clinical experience of an ultrasonic strain imaging system with novel noise-masking capability.
British Journal of Radiology, 83:702-706, August 2010.

[8] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
A normalization method for axial-shear strain elastography.
IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 57(12):2833-2838, December 2010.

[9] R. J. Housden, L. Chen, A. H. Gee, G. M. Treece, C. Uff, J. Fromageau, L. Garcia, R. W. Prager, N. L. Dorward and J. C. Bamber.
A new method for the acquisition of ultrasonic strain image volumes.
Ultrasound in Medicine and Biology, 37(3):434-441, March 2011.

[10] G. M. Treece, J. E. Lindop, L. Chen, R. J. Housden, R. W. Prager and A. H. Gee.
Real-time quasistatic ultrasound elastography.
To appear in Journal of the Royal Society Interface.

[11] E. Neale, R. Housden, G. Treece, R. Crawford, A. Gee, E. Sala and R. Prager.
A pilot study using Trans-vaginal Real-time Ultrasound Elastography to evaluate the post-menopausal endometrium.
To appear in Ultrasound in Obstetrics and Gynaecology.

Conference papers

[12] G. M. Treece, J. E. Lindop, A. H. Gee and R. W. Prager.
Freehand strain imaging with a 3D probe.
In Proceedings of the Sixth International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 66, Santa Fe, NM, USA, November 2007.

[13] A. H. Gee, L. Chen, S. Freeman, G. M. Treece, R. W. Prager and L. H. Berman.
A clinical database for evaluating freehand quasistatic strain imaging systems.
In Proceedings of the Seventh International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 25, Lake Travis, TX, USA, October 2008.

[14] R. J. Housden, A. H. Gee, G. M. Treece, R. W. Prager, G. P. Berry, L. Garcia and J. C. Bamber.
A comparison of 3D strain image quality using freehand 3D imaging and a mechanically-swept 3D probe.
In Proceedings of the Seventh International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 53, Lake Travis, TX, USA, October 2008.

[15] G. M. Treece, J. E. Lindop, A. H. Gee and R. W. Prager.
Fast, variable smoothing of strain data using nonparametric regression.
In Proceedings of the Seventh International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 57, Lake Travis, TX, USA, October 2008.

[16] L. Chen, G. M. Treece, J. E. Lindop, A. H. Gee and R. W. Prager.
A multi-directional displacement tracking algorithm for strain imaging.
In Proceedings of the Seventh International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 60, Lake Travis, TX, USA, October 2008.

[17] A. H. Gee, G. M. Treece, L. Chen and R. W. Prager.
The use of quality metrics in ultrasonic strain imaging.
To appear in Proceedings of Acoustical Imaging 30, Monterey, March 2009.

[18] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
Development of a weighting scheme for strain estimation.
In Proceedings of the Eighth International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 35, Vlissingen, Zeeland, The Netherlands, September 2009.

[19] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
A hybrid displacement estimation method for strain imaging.
In Proceedings of the Eighth International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 66, Vlissingen, Zeeland, The Netherlands, September 2009.

[20] R. J. Housden, A. H. Gee, G. M. Treece and R. W. Prager.
Freehand strain image normalization for convex probes.
In Proceedings of the Eighth International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 87, Vlissingen, Zeeland, The Netherlands, September 2009.

[21] K. Taylor, S. O'Keeffe, M. G. Wallis, R. J. Housden and G. M. Treece.
Ultrasound elastography as an adjuvant to conventional ultrasound in the pre-operative diagnosis of malignant axillary lymph nodes in patients with suspected breast cancer: a pilot study.
In Proceedings of the Ninth International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 36, Snowbird, UT, USA, October 2010.

[22] R. J. Housden, C. Uff, A. H. Gee, G. M. Treece, R. W. Prager and J. C. Bamber.
A new method for the acquisition of strain image volumes.
In Proceedings of the Ninth International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 49, Snowbird, UT, USA, October 2010.

[23] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
A normalization algorithm for axial-shear strain elastography.
In Proceedings of the Ninth International Conference on the Ultrasonic Measurement and Imaging of Tissue Elasticity, page 68, Snowbird, UT, USA, October 2010.

[24] E. Neale, R. Housden, G. Treece, A. Gee and R. Prager.
Can fibroids be demonstrated on 3-dimensional elastography?
Submitted to ISUOG world congress.

Technical reports

[25] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
A hybrid displacement estimation method for ultrasonic elasticity imaging.
Technical Report CUED/F-INFENG/TR 615, Cambridge University Department of Engineering, November 2008.

[26] R. J. Housden, A. H. Gee, G. M. Treece and R. W. Prager.
3D ultrasonic strain imaging using freehand scanning and a mechanically-swept probe.
Technical Report CUED/F-INFENG/TR 623, Cambridge University Department of Engineering, January 2009.

[27] G. M. treece, J. E. Lindop, A. H. Gee and R. W. Prager.
Uniform precision ultrasound strain imaging.
Technical Report CUED/F-INFENG/TR 624, Cambridge University Department of Engineering, March 2009.

[28] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
A normalisation method for axial-shear strain elastography.
Technical Report CUED/F-INFENG/TR 645, Cambridge University Department of Engineering, April 2010.

[29] L. Chen, R. J. Housden, G. M. Treece, A. H. Gee and R. W. Prager.
A data weighting scheme for quasistatic ultrasound elasticity imaging.
Technical Report CUED/F-INFENG/TR 651, Cambridge University Department of Engineering, May 2010.

[30] R. J. Housden, C. Uff, L. Chen, A. H. Gee, G. M. Treece, R. W. Prager and J. C. Bamber.
A new method for the acquisition of ultrasonic strain image volumes.
Technical Report CUED/F-INFENG/TR 656, Cambridge University Department of Engineering, August 2010.


People / Contacts

Cambridge UniversityCUEDMedical Imaging Group  |

© Department of Engineering at the University of Cambridge
Information provided by Dr. R. James Housden (rjh80)

Last updated: May 2011