During limb movement nerves have to slide and stretch, but the extent of such nerve movements has not been accurately assessed in vivo. This is a question of possible clinical interest since nerve entrapment syndromes, where such sliding may be altered, are common and present a significant health problem.

A method to measure transverse nerve movement from ultrasound images by measuring the distance between the nerve border and fixed structures was developed previously (Lynn et al. 2001). Since nerves are continuous structures, it is not possible to measure longitudinal movement in this way. To study nerve sliding in vivo, a non-invasive method was required. Doppler ultrasound can detect nerve movements (Hough et al. 2000), but the velocity of nerve sliding is at the lower limit of the method. Using frame-by-frame cross-correlation analysis of high-frequency ultrasound images, we have developed a method to measure longitudinal movement of peripheral nerves (Dilley et al. 2001).

Sequences of ultrasound images are captured at 10 frames s^-1 using a Diasus ultrasound system (Dynamic Imaging, UK) with a 10-22 MHz, 26 mm linear array transducer and analysed off-line using software developed in Matlab (Mathworks, USA). Resolution of the images is 0.093 mm pixel^-1 and image size was 280 X 440 pixels. The analysis software employs cross-correlation to determine relative movement between adjacent frames in sequences of ultrasound images. The programme calculates the correlation coefficient between the pixel grey levels for selected regions of tissue in two adjacent images. In the compared frame, the co-ordinates of the region of interest are offset along the horizontal image plane by a pixel at a time within a predetermined range (typically ± 15 pixels, i.e. ± 1.35 mm). The correlation coefficient is calculated for each individual pixel shift and the cross-correlogram is plotted. The region of interest is also offset along the vertical image plane (± 2 pixels) to compensate for any small vertical movements of the nerve. At each vertical offset a cross-correlogram is plotted in the horizontal image plane. A quadratic equation fitted to the maximum coefficient and its two adjacent values in each correlogram allows the estimation of the pixel shift to sub-pixel precision. The peak correlation coefficient in the two planes is used to estimate the shift in region of interest between frames. The program calculates the relative movement for each pair of adjacent frames in ultrasound sequences. The frame interval is increased to improve the accuracy of the correlation algorithm at low velocities (< 2 pixels). Best values are assessed from consistency of pixel shifts at different frame intervals and the a coefficient of the quadratic equation that defines the narrowness of the peak in the cross-corellogram.

Preliminary in vitro measurements on phantoms gave accurate estimates for movements, and reliable results were also obtained when the transducer was moved over the surface of the forearm at constant velocity.

Median nerve sliding has been examined during a range of limb movements in subjects and patients in studies approved by the local ethical committee. For example, sliding of the median nerve in the forearm and upper arm was determined during 40 deg wrist extension in 11 control subjects. The mean nerve movement in the forearm was 4.1 ± 0.94 mm (mean ± S.D.) and upper arm 1.7 ± 0.55 mm with the arm in 90 deg shoulder abduction. These results were consistent with nerve sliding estimates on cadavers and also expected values based on the anatomy of the wrist. Estimates of additional nerve strain were determined from measurements at different locations and indicated an additional strain of 1.1 % on the median nerve in the proximal forearm following 40 deg wrist extension. Movement of flexor digitorum profundus was also examined in seven subjects during 30 deg extension of the index finger from full flexion. Mean excursion in the distal forearm was 14.2 ± 4.4 mm.

In summary, the present method can be used to reliably measure longitudinal sliding from high-frequency ultrasound images of peripheral nerves, as well as other soft tissues (e.g. tendon, muscle). This method is currently being used to investigate altered nerve sliding in patients with nerve entrapment.

This work was supported by the Arthritis Research Campaign.