Use of Fourier shape descriptors to improve the reproducibility of echographic measurements of arterial intima‐media thickness

Background A major source of error in the longitudinal assessment of the intima–media thickness (IMT) is the difficulty in retrieving the same echographic view of the vessel. Objective To present a method for increasing the reproducibility of IMT measurements by ultrasound in large arteries. Method The Fourier descriptor is a well-known means of describing an object's shape. By means of the discrete Fourier transform (DFT), the shape was represented in a frequency domain; the computational advantages of the DFT then permitted a measure of unlikeness between different shapes (the ‘distance’ measure; DM) to be defined and used as a criterion for reproducing the contour. When the sonographer compared successive images of a complex vascular segment, like the carotid bifurcation, the identity of the echographic cut was deduced from the identity of the vessel's contour. The best match of the baseline image was the view that minimized the contour DM. Results Preliminary studies in the carotid artery bifurcations of eight subjects showed that the DM responds to systematic variations in the ultrasound interrogation angle and reveals minimal changes in transducer position. Duplicate scans of 12 subjects were performed by three sonographers with different strategies for acquisition of the same images: a low DM was associated with a low difference in pairs of IMT measurements. Data were classified into two groups (normal or borderline vessels with a pooled mean IMT of 0.62 mm and overtly thickened segments with a pooled mean IMT of 1.31 mm). When minimization of the DM was the criterion for the acquisition of replicate scans, the mean absolute difference of paired data for the mean IMT of the distal common carotid artery was 0.03 ± 0.02 mm for the first group and 0.06 ± 0.03 mm for the second group. This is a significant reduction in comparison with non-quantitative alternative criteria for image reproduction. For the maximum IMT of the same segments the mean absolute differences were 0.07 ± 0.03 and 0.13 ± 0.06 mm in the first and second groups, respectively. Conclusion This method can be applied to the serial assessment of single atherosclerotic segments. The computational time is negligible. By reducing the scatter in sequential IMT data, longitudinal investigations (e.g. of the results of antihypertensive therapy) with shorter durations and smaller sample groups may be rendered feasible.