## Biological Systems Engineering

#### Date of this Version

5-2013

#### Citation

IEEE Transactions on Ultrasonics, Ferroelectrics, and Fr 898 equency Control, vol. 60, no. 5, May 2013. Pages: 898 - 908. DOI: 10.1109/TUFFC.2013.2647.

#### Abstract

We have previously investigated the change of apparent lateral speckle size caused by the direction and spatial rate of scanner A-line acquisition (scan velocity). An algorithm which measures the lateral component of blood flow velocity was developed based on the increase in speckle size resulting from relative motion between moving scatterers and the scan velocity. In this paper, the change of the apparent dominant angle of the speckle pattern in a straight vessel was investigated and a new method of two-dimensional blood flow velocity estimation is introduced. Different scan velocities were used for data acquisition from blood flow traveling at an angle relative to the ultrasound beam. The apparent angle of the speckle pattern changes with different scan velocities because of misregistration between the ultrasound beam and scatterers. The apparent angle of the speckle pattern was resolved by line-to-line cross-correlation in the fast-time (axial) direction on a region-of-interest (ROI) in each blood flow image and used to spatially align the ROI. The resulting lateral speckle size within the aligned ROI was calculated. The lateral component of the blood flow is shown to be closest to the scan velocity which gives the maximum speckle size and the apparent angle of speckle pattern collected by this scan velocity is the best estimate for the actual angle of blood flow. These two components produce two-dimensional blood flow velocity estimations. This method was studied through both computer simulation and experiments with a blood flow phantom. Nine scan velocities were used to collect blood flow data with velocities ranging from 33 to 98 cm/s and four beam-to-flow angles. In simulated plug blood flow, the mean bias of angle estimation is below 2% with an average standard deviation of 3.6%. In simulated parabolic blood flow, the angle of blood flow is overestimated because of speckle decorrelation caused by flow gradients and the estimation bias increases with decreasing beam-to-flow angle, which has an average value of 8.8% and standard deviation of 10%. Because of the complexity of flow profiles in the blood flow phantom, the angle of blood flow is also overestimated and the mean bias is increased by a factor of two compared with simulated parabolic flow. For the velocity estimation results, the mean bias is below 5% with an average standard deviation of 4.6% in the simulated plug blood flow. In the simulated parabolic flow and blood flow phantom, the velocity is underestimated because of speckle decorrelation. The mean bias of velocity estimation in the simulated parabolic flow is −6% with an average standard deviation of 11.2%. In the blood flow phantom, the mean bias of the velocity estimation is −5% with a higher average standard deviation of 21.5%. This method can resolve the angle and amplitude of two-dimensional blood flow simultaneously. The accuracy of the estimation can be further improved by using more scanning velocities.

#### Included in

Bioresource and Agricultural Engineering Commons, Civil and Environmental Engineering Commons

## Comments

Copyright by the authors. Used by permission.