Subsample interpolation strategies for sensorless freehand 3D ultrasound

Abstract

Freehand 3D ultrasound can be acquired without a position sensor by deducing the elevational probe motion from the inter-frame speckle decorrelation. However, a freehand scan involves lateral and axial, as well as elevational, probe motion. The lateral sampling is determined by the A-line separation and is relatively sparse: lateral motion tracking therefore requires sub-sample interpolation. In this paper, we investigate the resilience of lateral interpolation techniques to simultaneous lateral and elevational probe motion. We propose a novel interpolation strategy and, through a series of in vitro experiments, compare its performance with that of established alternatives. The new technique is shown to be superior, limiting interpolation errors to around 5% of the length of the freehand reconstruction.

Sensorless Reconstruction of Unconstrained Freehand 3D Ultrasound Data

“…The principal sources of the bias are lateral interpolation between A-lines [10], compensation for coherent scattering [5] and speckle decorrelation caused by probe rotation instead of elevational motion. The last of these would result in systematic overestimation of elevational separationas is evident in many of the results in Figure 13 -and helps explain the somewhat larger drift for, say, data set 2 compared with previous results obtained with purely translational probe motion [5,10]. There is also some drift in the elevational yaw and tilt estimates (see, for example, the reslice of data set 1 in Figure 15), though we have not cluttered the results by presenting quantitative details here.…”

“…Secondly, there are other causes of speckle decorrelation (noise, in-plane motion, transducer rotation, tissue compression, physiological motion) which might be confused for elevational motion. While most of these remain a matter for future research, we have studied the effects of in-plane motion in detail [10]. We showed how to estimate in-plane motion to sub-sample accuracy using novel Gaussian interpolation techniques.…”

“…By "within range", we mean that the majority of the patches on B i are offset by an amount within the search range of the in-plane alignment [10] and are close enough elevationally to correlate to some extent with a corresponding patch on A. Frames that are elevationally too distant from A will be totally decorrelated and not allow any meaningful estimation of elevational or in-plane separation.…”

“…For all frame pairs within range of each other, we estimate the spatial separation between corresponding patches as three degrees of freedom: axial, lateral and elevational translation. This is performed as accurately as possible, making use of the technique in [5] to compensate for coherent scattering and the methods in [10] to calculate the lateral shift and subsequent correlation to sub-sample precision.…”

Sensorless Reconstruction of Freehand 3D Ultrasound Data

“…While most are a matter for future research, the effects of in-plane motion have been studied in detail. In [8], it is shown how to interpolate the in-plane motion estimates to sub-sample precision and then correct the elevational correlation values by allowing for the sub-sample in-plane motion.…”