H-scan ultrasound imaging for the classification of thyroid tumors

Author(s)
Mawia Khairalseed; Rosa Laimes; Joseph Pinto; Jorge Guerrero; Himelda Chavez; Claudia Salazar; Gary R. Ge; Roberto J. Lavarello; Kenneth Hoyt
Year of publication:
2022
Type of publication:
Conference paper
Conference / Journal Name:
2022 IEEE International Ultrasonics Symposium (IUS)
Publisher:
IEEE

Abstract

H-scan ultrasound (US) imaging is a high-resolution soft tissue characterization technique. It is a matched filter approach derived from analysis of scattering from incident pulses in the general form of nth-order Gaussian-weighted Hermite polynomial functions (GHn). The purpose of this study was to evaluate the potential of H-scan US imaging to distinguish scatterer distributions of biopsy-confirmed human benign and malignant thyroid lesions (N = 16 per group). Image data was acquired using a Sonix Touch US system (Analogic Ultrasound) equipped with an L14-5 linear array transducer. To generate H-scan US images, three convolution filters (i.e., GH n , n = 2,4, and 8) were applied in parallel to the radiofrequency (RF) data sequences to measure the relative strength of the backscattered US signals. To examine any spatial differences in H-scan US image intensity, textural features like contrast, energy, homogeneity, etc., were extracted from same-sized region-of-interests (ROIs) confined to each individual tumor. Experimental results verified the utility of H-scan US imaging of human thyroid tumors. Results from texture analysis of the H-scan US images indicated that there were statistically significant differences between benign and malignant tumor types for select features (p < 0.03).