Ultrasound fuzzy entropy imaging based on time-series signal for tissue characterization

Abstract

This study proposed a parametric imaging method of fuzzy entropy based on time-series signal, which can enhance the contrast and augment the accuracy of the detection of lesions in soft tissue. Fuzzy entropy assesses the irregularity of ultrasound time-series data through the detection of similarities within the signal waveform. Applying fuzzy entropy in ultrasound tissue characterization can provide more tissue information. The sliding windows are performed to traverse across the whole image with the step of one pixel. The signal in the sliding window calculates the fuzzy entropy (FE), and the FE value is assigned to the window’s center pixel. The entropy image will be constructed after obtaining all fuzzy entropies. The performance of the proposed FE imaging was compared with weighted Shannon entropy (WSE) imaging and B-mode imaging by simulations, ex vivo experiments, and clinical data. FE imaging got superior lesion area prediction with the AUC of 0.959 ± 0.042 (p < 0.0001), the accuracy of 99.88 ± 0.09 % (p < 0.0001), and the contrast-to-noise ratio of 7.806 ± 4.476 (p < 0.001) in the thyroid nodule diagnostic tests. The results show that FE imaging significantly increases the contrast of the images. Compared with WSE, the FE imaging method has better sensitivity to disorder degree and better recovery ability of tissue information when the amplitude of the signal decreases or the change of signal properties is small. FE imaging considers the characteristics of ultrasound time series. It has a better recovery ability for ultrasound signals, which is a promising technique for soft tissue diagnosis and thermal ablation monitoring.