A neural network approach for estimating large K distribution parameters

The K distribution has been proposed in the literature as a general speckle model for ultrasonic backscatter. The shape parameter of this distribution can be used to provide clinically important information on tissue density and regularity. A neural approach for parameter estimation is proposed, specifically for large values of the shape parameter. Experimental results on simulated images show that this approach compares favorably with other methods. Thus, neural networks can be used in conjunction with other approaches to accurately model speckle, and thereby to classify tissue.