3D IMAGING OF MUSCLES AND BONES

by Nicola Fry, Adam Shortland. One Small Step Gait Laboratory, Guy’s Hospital, London , England .

Adam and Nicky with the 3D probe

Researchers at the One Small Step Gait Laboratory in Guy’s Hospital, London , have been developing a 3D ultrasound system for the measurement of musculoskeletal morphology using a Vicon 612, a primitive ultrasound scanner and video frame grabber.

Progressive muscular and bony deformities are features of cerebral palsy and other childhood disorders. It is during adolescence that the functional mobility of children with these conditions decline, with many losing their ability to walk in their late teens and early twenties. It is likely that deterioration in muscle properties and the compromise of the skeletal levers leads to this loss of mobility. If we had a good way of quantifying deformity at different stages in the development of the child we may be able to:

- Understand more about the natural history of deformity development.

- Intervene at an earlier stage to maintain function.

- Assess the effects of our interventions on local musculoskeletal deformity.

Medical ultrasound imaging has been around since the late 1960s, and is the most common imaging modality in the hospital environment. It has a wide range of applications from assessing blood flow in the carotid artery to imaging tumours in the liver. 2D ultrasound is a useful but largely qualitative technique and in some areas it has been superseded by tomographic CT scanning and MRI. 3D ultrasound is a recent development but has great potential in quantitative imaging. 3D ultrasound is used routinely to assess the growing foetus or to inspect internal abdominal organs. These systems have exceptional accuracy but in general have a small scanning window. 3D scanning of limb segments requires that the position and the orientation of the probe can be registered over large distances. There are such freehand 3D systems available using magnetic tracking technology, but if you have a motion laboratory, a video frame grabber, a friend who is an ultrasonographer, and a bit of time on your hands, you could make one yourself!

Ultrasound probe with 3D localiser

The following step-by-step instructions are targeted at those in the laboratory with an engineering bent:

STEP 1: Find or buy an analogue 2D B-mode ultrasound machine with a 5 or 7.5 MHz linear array probe. Perfectly adequate, new examples can be bought for as little as $15000, but you may be able to ‘borrow’ one from your local radiology department. Often, these analogue machines are being thrown away in favour of their hi-resolution digital counterparts. Make sure your ultrasound machine has a composite video output. It is wise to include your new purchase in a hospital-wide quality assurance scheme.

STEP 2: Construct a 3D localiser from the Vicon marker kit. We attach 4 x 14cm marker stems to a small Perspex block. It’s best to keep the markers a good distance apart (at least 20cm) to reduce errors. The Perspex block should be machined so that it locks to the surface of your ultrasound probe.

STEP 3: Fix the Perspex block to the probe and secure it (we use tie-wrap) (see Figure). Connect the video output of the ultrasound scanner to your analogue video framegrabber (or to your firewire card using a DV bridge).

STEP 4: Before you can image in 3D, you’ll have to find the equation of the plane of the ultrasound image with respect to a local co-ordinate system defined by the probe-mounted marker set (3D localiser). This is a bit of a challenge, but you should find the paper of Fry et al. most helpful. Once you’ve worked the calibration out, you will be able to find the global co-ordinates of any pixel within the ultrasound image as you move the probe through space.

STEP 5: You need to write a computer program to construct a 3D image block of voxels from the ultrasound video images and the co-ordinates of your 3D localiser using an interpolation algorithm. There are a number of algorithms you could choose from (for review see Rohling et al. 1998); we use a nearest pixel algorithm and it seems to work well. If you were very clever you could write a plug-in for Vicon Workstation!

STEP 6: To investigate and display sections through your 3D volume you will need to write a suite of programs, or you can buy a program such as Analyse or IDL. In principle, you could write the code in MATLaB but the memory management of the very large matrices involved may be a limitation.

STEP 7: If you are not well-acquainted with ultrasound machines then ask a radiographer to help you out and go on a course to familiarise yourself with the physics of ultrasound and its safe usage. Get yourself a musculoskeletal ultrasound atlas and a good anatomy book with lots of pictures. Now play with your new toy, making sure you have all the necessary ethical and institutional approvals.

Applications

There are a number of ways in which you can use your system to inform treatment decision-making and even get better motion analysis results!

Muscle morphology

The need for an alternative to the passive range of motion examination to measure muscle deformity is clear. Passive ROM is the primary tool used for treatment selection across the world yet the examination has poor intra- and inter-rater reliability and cannot distinguish without ambiguity the muscles or muscular components responsible for a limitation in joint range.

There is poor correlation between the static measurements of the clinical examination and the dynamic measurements from analysis of the child’s movement, possibly because the factors that govern active and passive movements are not equivalent. Unambiguous state-ments describing the dimensions of muscles would help us to separate the contributions of deformity, weakness and neurological deficit to the patient’s motor problems, directly informing treatment recommendations.

The Figure illustrates a 3D ultrasound scan of the calf musculature of a normally-developing child and a child with spastic diplegic cerebral palsy of similar age. The hope is that by measuring the response of muscles to different treatments (e.g. strengthening, Botulinum toxin and surgery) in research studies, we will be able to determine the optimum treatment for those attending our gait laboratory for clinical and ultrasonic assessment.

Longitudinal slices through the calf muscles of (a) child with spastic diplegic cerebral palsy and (b) a normally developing child. Skin surface is at the top of the images, knee at the right and ankle on the left.

Longitudinal and transverse sections of a 3D volume containing the femoral head.

The surface of bone can be identified under ultrasound scanning easily because it reflects so much of the incident acoustic power. This means that reconstructions of the bone surface can be made with 3D ultrasound. We have used this technique to estimate the center of the femoral head with respect to a pelvic reference frame formed from the anterior and posterior iliac spines, and compared the results to those from MRI. In the medio-lateral and anterior-posterior directions there is exceptional agreement between the two methods (within a few millimeters), but work needs to be done to improve the accuracy of the method in the inferior-superior direction.

Our intention is to extend the technique to create partial surface reconstructions of the skeleton to enable a precise relationship between anatomical and technical reference frames to be found for individual subjects.

Bony morphology

n the clinical environment, tibial torsion, femoral anteversion and patella alta are measured by palpation of bony landmarks. Certainly, the lack of reliability of these measures may encourage a conservative approach to intervention being adopted. We have conducted 3D ultrasound studies in vitro and in vivo which indicate that 3D ultrasound can measure femoral anteversion and tibial torsion reliably, and in good agreement with the standard 3D CT method or MRI methods. Routine, serial measurements of bony torsion made possible with ultrasound will enable us to evaluate the natural history of skeletal deformity and its adaptation after surgical intervention.

Longitudinal section through a 3D ultrasound volume containing the patella and the proximal tibia.

References

Fry NR, Childs CR, Eve LC, Gough M, Robinson RO, Shortland AP. Accurate measurement of muscle belly length in the motion analysis laboratory: potential for the assessment of contracture. Gait and Posture 2003: 17: 119-124

Rohling RN, Gee AH, Berman L. Automatic registration of 3-D ultrasound images. Ultrasound Med Biol. 1998 Jul;24(6):841-54.