Non-invasive real-time temperature monitoring of tissue is a difficult problem and hot issue in hyperthermia. In this paper, fresh in vitro porcine liver tissues were heated in water-bath and the ultrasound images were acquired. The correlation of temperature and features, including gray level histogram, gray level co-occurrence matrix and gray level-gradient level co-occurrence matrix extracted from differential ultrasound images was analyzed. The results showed that 11 features of image were high correlated with tissue temperature from 26℃ to 45℃. These features included the mean values, standard deviation and entropy based on gray level histogram, the angular second moment, contrast, inverse difference moment and correlation in four directions (0°,45°,90°,135°), un-uniformity, energy, mixed entropy and inertia. The proposed 11 image features can help to construct the model of temperature measurement, it may be used for coagulation detecting in hyperthermia.
|