Noninvasive Temperature Estimation Based on Texture Features of Ultrasound Images for RFA
Abstract BackgroundHyperthermia is one of the most common therapy for tumor, in which temperature monitoring is essential. Invasive temperature monitoring is mostly used in clinic at present, which leads to limited points under monitoring and extra hurt to patients. Therefore, noninvasive temperature estimation is a key and tough problem in hyperthermia. In this paper, a noninvasive temperature estimation method for hyperthermia based on B-mode ultrasonic image with wavelet transform and texture analysis is proposed.MethodsAnimal tissue both in vitro and in vivo(pig kidneys in vitro and rabbit kidney in vivo) were taken as experimental materials. Ultrasound images and temperature data were collected from radio frequency ablation experiment under approved protocol. Image subtraction and wavelet transform were performed on the ultrasound images to enhance the change during ablation, after which texture features were extracted to analyze the relationship with temperature.ResultsTexture features hybrid entropy and energy extracted from gray-level gradient co-occurrence matrix of ultrasound images processed had higher linear correlation with temperature than traditional gray value method. From in vitro and in vivo experiments, the linear correlation coefficients obtained between hybrid entropy and temperature were 0.979±0.028 and -0.935±0.037, respectively, and those between energy and temperature were 0.932±0.056 and -0.915±0.034, respectively. In addition, the slopes of linear model obtained in vitro were relatively consistent, while those under in vivo condition had certain individual differences.ConclusionsThe results demonstrated that texture features hybrid entropy and energy of ultrasound image processed with the method proposed in this paper can be used for temperature estimation with a smaller error because of high linear correlation with temperature both in vitro and in vivo. Compared with most previous research which only based on biomaterial in vitro, the in vivo experiment conducted in this paper verified the feasibility of the noninvasive temperature estimation method proposed in practice.