Deconvolution of in-Vivo Ultrasound B-Mode Images

Abstract

An algorithm for deconvolution of medical ultrasound images is presented. The procedure involves estimation of the basic one-dimensional ultrasound pulse, determining the ratio of the covariance of the noise to the covariance of the reflection signal, and finally deconvolution of the rf signal from the transducer. Using pulse and covariance estimators makes the approach self-calibrating, as all parameters for the procedure are estimated from the patient under investigation. An example of use on a clinical, in-vivo image is given. A 2 x 2 cm region of the portal vein in a liver is deconvolved. An increase in axial resolution by a factor of 2.4 is obtained. The procedure can also be applied to whole images, when it is ensured that the rf signal is properly measured. A method for doing that is outlined.

Restoration of medical ultrasound images using two-dimensional homomorphic deconvolution

“…The partial effect of radial deconvolution only or lateral deconvolution only was clearly visible both for the strong specular reflections of the vessel walls and the weaker reflections of the parenchyma. The increased density of the speckle in the radial direction after radial deconvolution only was similar to that observed in in vivo liver images deconvolved radially using linear Kalman filtering [15].…”

“…To achieve an approximate linear spatial invariance of the 2D PSF of the ultrasound signal, the length of the recorded beams was limited to about 2.5 cm, and saturation of the analog-to-digital converter was avoided as far as possible (see Jensen et al [15] for a short discussion of this problem). Equally important, the 2D homomorphic deconvolution was applied to the polar coordinate representation of the ultrasound image recorded by sector scanners.…”

“…Furthermore, the necessity to do 2D deconvolution of the original radio frequency image to obtain the best possible results is well known [6], [20]. However, as stated by Loupas et al [20] and later by Jensen et al [15] very few methods have so far been convincingly shown to work reliably for exclusive lateral or radial deconvolution of in vivo ultrasound images, Manuscript received March 21, 1994; revised January 31, 1995. This study was supported by grants from The Research Council of Norway.…”

“…A major reason for this slow progress for in vivo images is the change in the pulse bandwidth and pulse shape which takes place in the body. The tissue layers that are interposed between the transducer and the target behave as a transmission channel introducing unknown distortions in the ultrasonic signal [12], [15]. This change makes it necessary to estimate the PSF in the image to be deconvolved.…”

“…This change makes it necessary to estimate the PSF in the image to be deconvolved. Jensen et al [15] use a onedimensional ARMA model for the pulse and estimate the model parameters by a prediction error algorithm. Kalman filtering is used in the deconvolution process.…”