Window size analysis in ultrasonic temperature estimation with timeshifts

Abstract

In this work, ultrasonic timeshift method for temperature change estimation was investigated for 2D simulated in-silico synthetic ultrasonic signals. Digital phantom tissue was created in MATLAB environment and acoustic simulation was running on k-Wave toolbox for two different temperature conditions. First temperature distribution was assigned to tissue as uniform 37 °C. Second temperature distribution is Gaussian form with peak at tissue center as 45 °C and tails of Gaussian curve is 37 °C. Signal was analyzed with ultrasonic timeshift method for temperature change estimation. This method is based on four steps, calibration with tissue constant, finding timeshift with cross correlation algorithm, find slope of timeshift vector respect to timestep, and multiply tissue constant and slope of local timeshift vector. This multiplication gives temperature change of local point. In this work, window size of smoothing filter of timeshift vector and linear fitting to timeshift—timestep data was analyzed as parametrically with range 3λ to 10λ with a 1λ increment for both windows equally. As a result, window parameters as 5λ to 7λ give best results, maximum absolute error is 0.82 °C, 0.97 °C and 0.92 °C respectively and mean absolute error is ∼0.35 °C. As a verify, different analysis was performed on different temperature distribution with discrete two peak curves.