Pixel-based small-window parametric ultrasound imaging for liver tumor characterization

Characterizing liver tumors remains a challenge in clinical practice. Ultrasound parametric imaging based on statistical distribution can enhance image contrast compared with B-mode imaging, requiring scatterers following specific distributions. This study proposes a pixel-based small-window parametric ultrasound imaging method using weighted horizontally normalized Shannon entropy (WhNSE) and fuzzy entropy (FE) to improve detectability liver tumor.

Pixel-based parametric imaging requires a sliding window to traverse across the B-mode image with the step of one pixel, while calculating the entropy by the pixel values in the window. The entropy is assigned to the center pixel of the sliding window. The entropy image is obtained after getting the entropy values of all pixels. FE and WhNSE are two novel entropies first applied to parametric imaging. The detection abilities of regions of interest (ROI) and the contrast-to-noise ratio (CNR) were evaluated through simulations and clinical explorations.

In simulations, FE imaging showed the highest improvement in detecting hyperechoic ROIs, with a CNR gain up to 457.31% (p < 0.01) in simulations. WhNSE imaging demonstrated the best performance in hyperechoic ROI detection, with a CNR of 1.607 ± 0.816 (p = 0.05), significantly higher than B-mode images.

The proposed pixel-based parametric imaging method based on fuzzy entropy and weighted horizontally normalized Shannon entropy both effectively enhance the contrast and detectability of ultrasound images. The imaging enhancement method of the pixel-based fuzzy entropy imaging with proper parameters got better detection performance, due to the consideration of the relationship of neighboring pixels.