Ultrasound Entropy Imaging for Detection and Monitoring of Thermal Lesion During Microwave Ablation of Liver.

Xiejing Li,Xin Jia,Ting Shen,Mengke Wang,Guang Yang,Hua Wang,Qinli Sun,Mingxi Wan,Siyuan Zhang
DOI: https://doi.org/10.1109/jbhi.2022.3167252
IF: 7.7
2022-01-01
IEEE Journal of Biomedical and Health Informatics
Abstract:Ultrasonic B-mode imaging offers non-invasive and real-time monitoring of thermal ablation treatment in clinical use, however it faces challenges of moderate lesion-normal contrast and detection accuracy. Quantitative ultrasound imaging techniques have been proposed as promising tools to evaluate the microstructure of ablated tissue. In this study, we introduced Shannon entropy, a non-model based statistical measurement of disorder, to quantitatively detect and monitor microwave-induced ablation in porcine livers. Performance of typical Shannon entropy (TSE), weighted Shannon entropy (WSE), and horizontally normalized Shannon entropy (hNSE) were explored and compared with conventional B-mode imaging. TSE estimated from non-normalized probability distribution histograms was found to have insufficient discernibility of different disorder of data. WSE that improves from TSE by adding signal amplitudes as weights obtained area under receiver operating characteristic (AUROC) curve of 0.895, whereas it underestimated the periphery of lesion region. hNSE provided superior ablated area prediction with the correlation coefficient of 0.90 against ground truth, AUROC of 0.868, and remarkable lesion-normal contrast with contrast-to-noise ratio of 5.86 which was significantly higher than other imaging methods. Data distributions shown in horizontally normalized probability distribution histograms indicated that the disorder of backscattered envelope signal from ablated region increased as treatment went on. These findings suggest that hNSE imaging could be a promising technique to assist ultrasound guided percutaneous thermal ablation.
What problem does this paper attempt to address?