A Framework for Temperature Imaging using the Change in Backscattered Ultrasonic Signals

Hyperthermia is a cancer treatment that elevates tissue temperature to 40 to 43oC. It would benefit from a non-invasive, safe, inexpensive and convenient thermometry to monitor heating patterns. Ultrasound is a modality that meets these requirements. In our initial work, using both prediction and experimental data, we showed that the change in the backscattered energy: CBE) is a potential parameter for TI. CBE, however, was computed in a straightforward yet ad hoc manner. In this work, we developed and exploited a mathematical representation for our approach to TI to optimize temperature accuracy. Non-thermal effects of noise and motion confound the use of CBE. Assuming additive white Gaussian noise, we applied signal averaging and thresholding to reduce noise effects. Our motion compensation algorithms were also applied to images with known motion to evaluate factors affecting the compensation performance. In the framework development, temperature imaging was modeled as a problem of estimating temperature from the random processes resulting from thermal changes in signals. CBE computation was formalized as a ratio between two random variables. Mutual information: MI) was studied as an example of possible parameters for temperature imaging based on the joint distributions. Furthermore, a maximum likelihood estimator: MLE) was developed. Both simulations and experimental results showed that noise effects were reduced by signal averaging. The motion compensation algorithms proved to be able to compensate for motion in images and were improved by choosing appropriate interpolation methods and sample rates. For images of uniformly distributed scatterers, CBE and MI can be computed independent of SNR to improve the temperature accuracy. The application of the MLE also showed improvements in temperature accuracy compared to the energy ratio from the signal mean in simulations. The application of the framework to experimental data requires more work to implement noise reduction approaches in 3D heating experiments. The framework identified ways in which we were able to reduce the effects of both noise and motion. The framework formalized our approaches to temperature imaging, improved temperature accuracy in simulations, and can be applied to experimental data if the noise reduction approaches can be implemented for 3D experiments