NONINVASIVE TWO-DIMENSIONAL TEMPERATURE IMAGING FOR GUIDANCE OF THERMAL THERAPY Emad S. Ebbini University of Minnesota Twin Cities Department of Electrical and Computer Engineering ABSTRACT Two-dimensional temperature estimation using pulse-echo di- agnostic ultrasound has been previously described [5, 6, 8]. Measurement models in both time and frequency domains have been proposed. In [5], spectral shifts in the echo data have been shown to be proportional to changes in the local tissue temperature. In [7, 8], echo shifts were shown to be proportional to local change in tissue temperature. For both the spectral and echo shift methods, the proportionality was shown to be proportional to the local changes in the speed of sound and local tissue expansion. However, temperature images suffer from some artifacts due to the distortion of the imaging beam as it traverses the heated region. In particu- lar, temperature imaging artifacts due to the thermoacoustic lens effects have been reported [7, 10]. In addition, tissue inhomogeneity leads to nonuniform speckle patterns which also lead to errors in the estimated temperature profiles. A re- constructive imaging method employing a physics-based 2D filter and a projection method is presented. The method of projection onto convex sets (POCS) is used. Experimental data was obtained during controlled temperature heating of in vitro bovine muscle using a helical RF ablation probe. Re- constructions using the POCS-based iterative algorithm are shown to produce artifact-free temperature fields. Both spa- tial and temporal characteristics of the reconstructed temper- ature conform well with the extent of the heating source and the temporal dynamics of the controlled temperature. These results demonstrate that noninvasive temperature imaging of a relatively large heating region (nearly 3 cm in diameter) can be reliably monitored using pulse-echo ultrasound with ap- propriate signal and image processing. 1. INTRODUCTION Noninvasive temperature estimation continues to attract atten- tion as a means of monitoring and guidance for minimally- invasive thermotherapy. Currently, minimally-invasiveRF ab- lation is the most commonly used form of thermal therapy, but Funded in Part by Army Grant DAMD 17-01-1330. Parts of this work were funded under Grant CA 66602 from the National Institutes of Health. other techniques are being used or investigated. For exam- ple, high-intensity focused ultrasound (HIFU) is being evalu- ated clinically as a form of noninvasive thermal therapy. Suc- cessful implementation of noninvasive temperature estima- tion will be a boon for thermal therapy as it becomes less in- vasive and as the heating sources become more sophisticated. Currently, magnetic resonance imaging (MRI) and ultra- sound have both been proven to have the temperature sen- sitivity and spatial resolution necessary to provide noninva- sive temperature feedback. The main limitation for MRI is the cost and the potential complication of the heating proto- col as the heating equipment must be MR compatible. Ultra- sound is relatively less expensive, portable, and can be used in conjunction with almost any heating protocol without adding any significant constraints. Therefore, noninvasive tempera- ture estimation based on ultrasound echo data continues to be an important problem in the area of image-guided minimally- invasive thermal therapy. 2. TEMPERATURE ESTIMATION The temperature change estimation method in this paper is based on the thermal dependence of the ultrasound echo that accounts for two different physical phenomena: local change in speed of sound and thermal expansion of the propagat- ing medium due to changes in temperature. The speed of sound c is a function of temperature. In most tissue media around body temperature, c increases with temperature in the range (see [2]). In fatty tissues, c decreases with increasing temperature[4]. In [5], we have described a temperature measurement equa- tion of the form: Δf k (T ) ≈ k 2 ∂c(T ) ∂T 1 d(T ) |T=T 0 − ∂d(T ) ∂T |T =T 0 c(T ) d 2 (T ) ΔT, (1) which relates the change in tissue temperature to changes in harmonic frequencies related to the mean scanner spacing [6]. Strictly speaking, T 0 is function of space and time. However, in practice, the proportionality constant in Equation 1 is con- stant for a range of ΔT up to 15 ◦ C from normal body tem- perature. Therefore, it is reasonable to assume T 0 constant throughout the temperature range of interest. Note that for 884 0-7803-9577-8/06/$20.00 ©2006 IEEE ISBI 2006