Ultrasound-based Noninvasive Temperature Estimation In Radiofrequency Ablation Of Tumors

Hyperthermia is described as a therapy capable of destroying local tumors by heating the target tissue to a certain high temperature.It has been widely used in clinical tumor treatment.Radiofrequency ablation(RFA)has become one of the most common heating methods using thermal effect of radio frequency current to form a high temperature thermal coagulation area and kill the tumor cells by releasing the radio frequency alternating current.Temperature regulation in target zone during ablation is critical to induce irreversible coagulation and to avoid necrosis of the normal tissue.In recent years,many noninvasive temperature estimation methods have been proposed.Among them,ultrasound-based temperature estimation has become one of the most popular research directions in this field,as its various advantages of simple operation,safety,high resolution and real-time data processing.The current study focused on noninvasive temperature estimation based on ultrasound image.The correlations between features of ultrasound image and temperature were explored.Radiofrequency ablation experiment was designed and conducted,in which animal kidneys both in vitro and in vivo were used as biological materials.During the ablation,ultrasound images and temperature data were collected in real time for processing and analysis.After preprocessing,image features such as the average gray value,the gray-level co-occurrence matrix,and the gray-level gradient co-occurrence matrix were extracted and used to analyze their relationships with temperature.It is demonstrated that the average gray value of ultrasound images has a certain correlation with temperature,but the correlations in different temperature intervals are different,so it is difficult to establish a unified regression model;the parameters of the gray-level co-occurrence matrix have certain correlations with temperature in high temperature range,but those in low temperature range are very weak,so it is difficult to reflect the change of temperature overall.In order to achieve accurate prediction of each temperature interval during RFA,two noninvasive temperature estimation methods based on ultrasound image texture analysis were proposed.One of the methods was based on the multiple parameters of the gray gradient-level co-occurrence matrix.It is showed that some parameters of the gray-level gradient cooccurrence matrix,such as hybrid entropy and inverse difference moment,have high linear correlations with temperature in the entire temperature interval,which proves that they have feasibility to be the temperature characterization parameters in radiofrequency ablation.So several parameters highly correlated with temperature are selected to perform multiple regression analysis to achieve a temperature prediction model with a higher degree of fitting.The other method is to reprocess the ultrasound subtraction image by wavelet analysis to enhance the image texture features and enhance the linear correlation between features and temperature.In order to verify the feasibility of the proposed methods,the established regression models are used to predict the temperature of the ablation area and compare with the measured temperature in experiment.The results show that the parameters obtained by these two methods both have significant increase in the degree of correlation,and the linear relationships are significant.Good linear fit can be achieved for multiple sets of experimental data,and the degree of fit are above 0.95,which demonstrates that the methods proposed can be used to achieve high-precision noninvasive temperature estimation.It provides effective temperature monitoring in clinical radiofrequency ablation therapy based on ultrasound image.