In this paper, we have evaluated the use of smooth windows for ultrasound elastography. In ultrasound elastography, local tissue strain is estimated using operations such as cross-correlation on local segments of RF data. In this process, local data segments are selected by multiplying the RF data by a rectangular window. Such data truncation causes non-ideal spectral behavior, which can be mitigated by using smooth windows. Accordingly, we hypothesize that the use of smooth windows may improve the elastographic signal-to-noise ratio (SNRe) and contrast-to-noise ratio (CNRe) of strain images. The effects of using smooth windows have not been fully characterized for time-domain strain estimators. Thus, we have compared the elastographic performance of rectangular, Hanning, Gaussian, and Chebyshev windows used in conjunction with cross-correlation based algorithm and adaptive stretching algorithm using finite element method (FEM) simulation, experimental phantom, and in vivo data. Smooth windows are found to improve the SNRe by up to 3.94 for FEM data and by up to 1.76 for phantom data which represent 76% and 60.52% improvements, respectively. CNRe improves by up to 12.23 for FEM simulated data and by up to 4.28 for phantom data which represent 213.07% and 248.2% improvements, respectively. Mean structural similarity (MSSIM) was used for assessing the image perceptual quality and smooth windows improved it by up to 0.22 (85.98% improvement) for simulated data. We have evaluated these parameters at 1-6% applied strains for the experimental phantom and at 1%, 2%, 4%, 6%, 8%, and 12% applied strains for FEM simulation. We observed a maximum deterioration in axial resolution of 0.375 mm (which is on the order of the wavelength, 0.3 mm) due to smooth windows. "Salt-and-pepper" noise from false-peak errors has also been reduced. Smooth windows increased the lesion-to-background contrast (by increasing the CNRe by 213.07%) of a low contrast lesion (10-dB). For the in vivo cases, use of smooth windows resulted in better depiction of lesions, which is important for lesion classification. In this work, we have used an ATL Ultramark 9 scanner with an L10-5 (7.5 MHz) probe for the phantom experiment and a Sonix SP500 scanner with an L14-5/38 probe (10 MHz) for in vivo data collection.
All Science Journal Classification (ASJC) codes
- Acoustics and Ultrasonics
- Data window
- Spectral leakage