Uncertainty estimation for temperature measurement with diagnostic ultrasound

Ultrasound therapies are promising, non-invasive applications with potential to significantly improve, e.g. cancer therapies like viro- or immunotherapy or surgical applications. However, a crucial step towards their breakthrough is still missing: affordable and easy-to-handle quality assurance tools for therapy devices and ways to verify treatment planning algorithms. This deficiency limits the safety and comparability of treatments.

To overcome this deficiency accurate spatial and temporal temperature maps could be used. In this paper, the suitability of temperature calculation based on time-shifts of diagnostic ultrasound backscattered signals (echo-time-shift) is investigated and associated uncertainties are estimated. Different analysis variations were used to calculate the time-shifts: discrete and continuous methods as well as different frames as a reference for temperature calculation (4 s before, 16 s before the frame of interest, base frame). A sigmoid function was fitted and used to calculate temperatures. Two-dimensional temperature maps recorded during and after therapeutic ultrasound sonication were examined. All experiments were performed in agar-graphite phantoms mimicking non-fatty tissue, with high-intensity focused ultrasound being the source of heating.

Continuous methods are more accurate than discrete ones, and uncertainties of calculated temperatures are in general lower, the earlier the reference frame was recorded. Depending on the purpose of the measurement, a compromise has to be made between the following: calculation accuracy (early reference frame), tolerance towards small movements (late reference frame), reproducing large temperature changes or cooling processes (reference frame at a certain point in time), speed of the algorithm (discrete (fast) vs. continuous (slower) shift calculation), and spatial accuracy (interval size for index-shift calculation). Within the range from 20 °C to 44 °C, uncertainties as low as 12.4 % are possible, being mainly due to medium properties.

Temperature measurements using the echo-time-shift method might be useful for validation of treatment plan algorithms. This might also be a comparatively accurate, fast, and affordable method for laboratory and clinical quality assessment. Further research is necessary to improve filter algorithms and to extend this method to multiple foci and the usage of temperature-dependent tissue quantities. We used an analytical approach to investigate the uncertainties of temperature measurement. Different analysis variations are compared to determine temperature distribution and development over time.