Speckle removal from heterogeneous-tissue signals using independent component analysis

Abstract

In order to diagnose fibrosis quantitatively, a technique to extract fiber structure and suppress speckle from the echo data acquired using available clinical ultrasonic diagnostic instruments would be very valuable. To meet this need, we propose to apply independent component analysis (ICA) to separate the signals returning from clinical tissue fiber structures from the speckle signals. To use ICA and to remove speckle, data acquired from a normal tissue-mimicking phantom are acquired to characterize speckle and used to correct for speckle in clinical echo-signal data. Results showed satisfactory separation of speckle signals and fiber-tissue echo signals. However, results indicated that the ability of the ICA algorithm to correctly image fiber tissue structures depended greatly on whether the speckle in the clinical and tissue-mimicking images had same properties.