A generalized Hermite pyramid for ultrasonic image analysis I. Kidney and liver

Abstract

Ultrasonic images of the kidney and of the liver are subjected to a multiscale analysis in a generalized Hermite pyramid framework. The gradient images of the multiscale decompositions of the images of healthy and sick kidneys, and of theintraoperativeand conventionally imaged livers, exhibit differences, in the structures of gray level regions, which can be interpreted by a medical doctor. These are used as inputs to an unsupervised classifier to automatically classify the images into homogeneous groups, which are found,in the case of the ultrasonic images examined, to correspond to the different physical characteristics of tissues of the organs under study. The main contribution of the paper is believed to be the multiscale tissue characterization along with its display in a manner that has utility as a diagnostic aid to the clinician.