Investigation of modified non-speckle removal method based on non-speckle probability image for quantitative tissue characterization by ultrasound envelope statistics

Abstract Echo-envelope statistics are a powerful indicator for quantitative tissue characterization. However, previous studies have reported that envelope statistics, particularly the shape parameter of the envelope statistics model, which contains important information on tissue characteristics, are susceptible to high-intensity and point-like non-speckle (NS) signals originating from small blood vessels in the liver. Therefore, this study proposes and investigates a modified NS removal method utilizing an NS probability image obtained from information across multiple windows. The performance of the proposed method was evaluated through Field II numerical simulation. Compared to the previously proposed method based on single-window information, the proposed method improved the estimation accuracy of statistical moments and liver fibrosis parameters by excluding signals with NS probabilities exceeding appropriate threshold. These results demonstrate the potential of the proposed method to improve the accuracy of envelope-statistics-based tissue characterization in the presence of NS signals.