3D Spatial and Temporal Spectrum Doppler Display for Ultrasound
Paul Liu, Chaowei Tan, Dong C. Liu
Introduction
In cardiovascular applications of medical ultrasound, determining the velocities and velocity distribution of blood flow assists diagnosis by providing information on the direction, speed, and turbulence of the flow. Spectrum Doppler uses repeated emission bursts focused at the same spatial location known as the range gate to calculate the velocity distribution within that gate. The information is then visualized as a spectrogram, which is a time varying plot with time on one axis and velocity on the other.
The drawback to conventional spectrum Doppler is that spatial information about the flow is lost as data within the range gate is summed. Hence, turbulence of the flow, which for example may be indicative of atherosclerosis of the carotid, is deduced from the width of the spectrum, or the spread of the velocities present in the gate. However, laminar flow within a vessel follows a parabolic velocity profile with respect to the distance from the vessel wall. Thus, spatial information of velocity may be as essential as velocity distribution.
We propose a 3-D visualization that includes the additional dimension of space. Using multiple gates to capture the velocity distribution at different locations along the cross-section of a vessel will give spatial in addition to distributional and temporal information.
Overview of the methods
Basic method
User selectable range gate with sufficient size to cover the entire blood vessel.
Division of aforementioned range gate into M sub-gates, as in Fig. 1.
Spectrogram of summed time series within each sub-gate is computed using spectral method of choice, such as short-time Fourier transform.
M spectrograms are displayed in 3D format by stacking the M slices of the spectrogram and then reconstructing the full 3D volume, as in Fig. 2. Each voxel represents the spectral density magnitude for its corresponding sub-gate and frequency (velocity). Voxel intensity is the power spectral density of the corresponding spectrogram slice. Opacity must be properly chosen to display the full range of spectral volume.
Improvement
Number of gates may be reduced if we perform interpolation, i.e. trilinear during rendering.
To allow display of an arbitrary 2D slice in the volume. This allows visualization of the spatial distribution profile (space and velocity dimensions) or any other combination of dimensions in a separate image.
A real time implementation of such a system with the 3D volume updating in time would be considered a 4D visualization Measurement package, such as traces of 2D slice, surface extraction, curvature or gradient calculation Modes.
To combine with B and/or C mode.
To combined with existing methods. For example, to combine with compounding, coded excitation, spectral estimation with missing samples, etc. Coded excitation is especially relevant as finer resolution is generally needed to discern the parabolic profile along the borders of the vessel in the case of laminar flow. Conventional pulse wave Doppler uses 10 to 15 cycles, which potentially blurs the spatial dimension and will result in a image without the noticeable parabolic profile.
Experiment and Results
Simulation (Done at August 2009)
To obtain actual in-vivo conventional spectrogram.
To use computer program to scale 64 copies of the actual spectrogram into a parabolic shape. This is done at this step because actual data from multiple gates could not be obtained at that time. (August 2009 when the simulation was done).
These 64 frames are then rendered in 3-D with particular hand-tuned transfer function for opacity control. The result is in Fig. 3.
Sub-system (Done at Jan 18 2010)
We have implemented the sub-system onto our ultrasound scanner. In Fig. 4, we image the carotid of a healthy subject.
We divide the displayed gate into multiple (16 as of this implementation) sectioned sub-gates in Fig. 4a and display the conventional spectrogram of one sub-gate.
The spatial distribution at a particular time is then displayed in Fig. 4b indicating a roughly parabolic profile.
Visualization is then performed by stacking slices of the spectrograms from each gate into a 3-D grid as shown in Fig. 4c.
The full 3-D volume is then reconstructed in Fig. 4d to give a spectral distribution dependent on both time and space.
The improved system for practical application, as in Fig. 5.
Fig. 5. A practical application using this 3D Spatial and Temporal Spectrum Doppler function
Conclusion
The visualization of flow distribution in time and space in a 3D form is new and also the visualization of arbitrary 2D slices from the 3D volume.
Allows clinicians a view of velocity distribution both in time, which may include several systolic/diastolic cycles and also space, which will allow verification of laminar or turbulent flow. Furthermore, we will allow 2D profiles in any combination of two dimensions.