Improved Nakagami imaging usinga combined estimator

Nakagami imaging is an ultrasonic tissue characterization technique which has shown promise for differentiating different breast masses. Two parameters describing the shape and the scale of the distribution of the RF-data envelope can be computed using sliding windows. Though the traditional moment-based estimator is theoretically unbiased, its performance is not stable and depends on the value of the parameters. In this work, we propose a framework combining two estimators to improve and homogenize the accuracy of the estimation. This is done by computing both parameters and deciding a posteriori which is the best by measuring the goodness of fitting between the estimated theoretical distribution and the observed one. We test this method both on synthetic images presenting real US speckle pattern structures and a Nakagami distribution with chosen parameters and on clinical US breasts images. Results show a considerable improvement of the accuracy and of the stability of the estimations over conventional estimation.