Correlations Between B-mode Ultrasound Image Texture Features and Tissue Temperature in Hyperthermia

The noninvasive estimation of tissue temperature is a key problem in hyperthermia.The original ultrasonic RF signal was used to reconstruct B-mode images,eliminating device-dependent variances to explore the use of B-mode image texture parameters in monitoring temperature and coagulation zones in hyperthermia.The water bath heating experiments were carried out on in-vitro porcine liver samples,and the ultrasound RF signal at different temperatures was collected.Then the Bmode ultrasonic images were reconstructed,and the texture features(gray level histogram,gray level cooccurrence matrices,and gray level-gradient co-occurrence matrix) were extracted.The correlations between temperature and texture parameters of B-mode images were analyzed.Resultsindicate that some texture parameters are linearly correlated with the temperature in the heated tissue at a temperature from 20 ℃ to 60 ℃.Five parameters with large correlations with temperatures include mean gray scale of gray level histogram,entropy of gray level co-occurrence matrix,and hybrid entropy,inverse difference and correlation of gray level-gradient co-occurrence matrix.These five texture features with large correlations with temperatures can be used for tissue temperature estimation in clinical microwave tumor ablation.