Nakagami-m parametric characterization of contrast-enhanced ultrasound: In vivo validations

Background, Motivation and Objectives: The distribution of echoes from simulated and in vitro nonlinear scattering microbubbles obeys the Nakagami (NK) model, of which shape parameter (m) has the potential to suppress the artifacts from systems and operators in echo amplitude-coding contrast-enhanced ultrasound (CEUS) (Med. Phys. 2019, 46: 5467-5477). However, the feasibility of m estimation in in vivo CEUS is still controversial and its sensitivity and discriminability are limited, since raw envelopes are regulated by complex radiofrequency (RF) and video-frequency (VF) processing. This study aimed to illustrate this regulation mechanism and overcome these limitations. Statement of Contribution/Methods: The regulation mechanism of eight harmonic detection techniques and logarithmic compression on NK distributions was investigated in both B and pulse-inversion (PI) modes. A window-modulated compounding moment estimator (WMCME) was developed to estimate the CEUS m values. Sensitivity and discriminability of m-coding CEUS were respectively quantified by m-contrast-to-tissue ratio (mCTR), m-contrast-to-noise ratio (mCNR), and -2 dB half-lengths of axial and lateral autocorrelation functions, which were validated via in vivo perfusion experiments of rabbit kidneys. Results, Discussion and Conclusions: Regulated by the RFVF processing, the distributions of CEUS also obeyed the NK model, of which NK-fitted correlation coefficient was .99pm 0.01 (p < 0.05 in t-test and p < 0.001 in Kolmogorov-Smirnov test). In all m-coding CEUS methods, logarithmic m-coding PI scheme well-characterized the ring-like perfusion features and details within the renal cortex. The m CTR, m CNR, -2 dB axial and lateral half-lengths were up to 7.9pm 1.5 mathrm{dB}, 34.4pm 1.7 mathrm{dB}, 1.02pm 0.02 and .91pm 0.02 mathrm{mm}, respectively, which were higher than or comparable with amplitude-coding CEUS. The NK model can characterize CEUS even if the envelope distributions were regulated by the RF VF processing. The logarithmic m-coding PI scheme significantly improved the sensitivity, discriminability, and robustness of m estimation in CEUS. This scheme provided an option to remove those artifacts in echo amplitude-coding CEUS and more distinctly characterize the inherent microvasculature marked by microbubbles, which could provide a valuable technique to improve precise diagnoses and therapy.