Speckle Classification for Sensorless Freehand 3D Ultrasound

Despite being a valuable tool for volume measurement and the analysis of complex geometry, the need for an external position sensor is holding up the clinical exploitation of freehand three-dimensional ultrasound. Some sensorless systems have been developed, using speckle decorrelation for out-of-plane distance estimation, but their accuracy is still not as good as that of sensor-based systems. Here, we examine the widely held belief that accuracy can be improved by limiting the distance measurements to patches of ultrasound data containing fully developed speckle. Without speckle detection, we observe that scan separation is systematically underestimated by 33.1 % in biological tissue. We describe a number of speckle detectors and show that they reduce the underestimate to about 25 %. We conclude that speckle classification can improve the quality of distance estimation, but not sufficiently to achieve accurate, metric reconstruction of the insonified volume