Frequency Agnostic Tissue Characterization in Ultrasound Imaging using Backscattered Signal Statistics

Abstract

Ultrasound (US) imaging-based tissue characterization (TC) is a vital tool for improving diagnostic accuracy by assessing tissue properties. However, existing methods often lack generalizability across varying US acquisition frequencies. This paper introduces a frequency-agnostic TC method that estimates backscattering statistical parameters and assesses the confidence of the underlying radio frequency (RF) data distribution using two approaches: (a) variational scale estimation and (b) test signal resampling. These parameters train a random forest for TC, validated on RF data acquired at 5–10 MHz. The method achieves Dice coefficients of 0.834±0.040 for hyperechoic regions and 0.751±0.051 for hypoechoic regions across 400 positional acquisitions.