Large-Array Deep Abdominal Imaging in Fundamental and Harmonic Mode

Abstract

Deep abdominal images suffer from poor diffraction-limited lateral resolution. Extending the aperture size can improve resolution. However, phase distortion and clutter can limit the benefits of larger arrays. Previous studies have explored these effects using numerical simulations, multiple transducers, and mechanically swept arrays. In this work, we used an 8.8-cm linear array transducer to investigate the effects of aperture size when imaging through the abdominal wall. We acquired channel data in fundamental and harmonic modes using five aperture sizes. To avoid motion and increase the parameter sampling, we decoded the full-synthetic aperture data and retrospectively synthesized nine apertures (2.9-8.8 cm). We imaged a wire target and a phantom through ex vivo porcine abdominal samples and scanned the livers of 13 healthy subjects. We applied bulk sound speed correction to the wire target data. Although point resolution improved from 2.12 to 0.74 mm at 10.5 cm depth, contrast resolution often degraded with aperture size. In subjects, larger apertures resulted in an average maximum contrast degradation of 5.5 dB at 9-11 cm depth. However, larger apertures often led to visual detection of vascular targets unseen with conventional apertures. An average 3.7-dB contrast improvement over fundamental mode in subjects showed that the known benefits of tissue-harmonic imaging extend to larger arrays.

Fig. 1.

Pulse sequence and post-processing schemes. Each imaging frame consisted of 100 lines with pulse inversion (200 focused beams within imaging FOV in blue). Following immediately, 5 lines were acquired with repeated pulse inversion pairs (20 focused beams within repeat FOV in red). These frames were then repeated five times with different aperture sizes (N=element count). During post-processing, full synthetic aperture (FSA) data decoded from each frame (imaging FOV) were beamformed using a diverging wave model. The transmit and receive channels of the largest aperture FSA data were controlled to mimic various aperture configurations, including the ones physically transmitted. Images from this retrospective sweep were used in all analysis. Matched configurations (N=128, 192, 256, and 320) were used to assess the validity of retrospective sweep.

Fig. 2.

Images of the experimental setup. (a) The large linear array. (b) Side view of the ex vivo porcine abdominal wall samples. (c) Wire target imaging setup. Inset shows the top view. (d) Phantom imaging setup.

Fig. 3.

In vivo data inclusion criteria. Study began with 55 images from 13 patients with at least one segmentable hypoechoic target within 7–13 cm depth. Targets were then segmented at three depths (no more than 1 target per depth within an image). Noisy data were excluded by a temporal coherence (TC) threshold. F and H indicate the number of fundamental and harmonic images/targets available at various steps. The final aggregate contains fundamental images from all 13 subjects and harmonic images from 8 subjects.

Fig. 4.

Images of the wire target acquired through the water path and the wall samples. At each row, images are shown for five selected apertures. For each through-sample acquisition, images are shown for FSA beamforming with effective sound speed (SOS) and water sound speed. Beamforming SOS is reported on the left column. Images are displayed on a 60 dB dynamic range. The right panel shows the lateral profiles of the images at the depth of maximum envelope brightness. Only three profiles are displayed for clear visualization.

Fig. 5.

Full-width-half-maximum (FWHM) of the wire target PSFs. Measurements are shown for beamforming with (a) water sound speed and (b) effective sound speed. The water path profiles in (a) and (b) are identical. The transmit/receive F-number at the target depth in the control image is reported on (b). Circles indicate the smallest FWHM.

Fig. 6.

Cystic resolution profiles calculated from the wire target images for select aperture configurations. Profiles are reported for beamforming with water sound speed (top row) and effective sound speed (bottom row).

Fig. 7.

(a-c) Time-delayed RF signals from the wire target are displayed as a function of receive elements (FSA beamformed). Images correspond to (a) the control and Sample 2 with (b) water SOS beamforming and (c) bulk correction. Arrival time delay estimates for control and through-sample acquisitions using (d) water SOS beamforming and (e) bulk correction. Note that time delay estimates from 1D arrays are affected by aberration integration error [32]. These results are only presented to illustrate the effects of the abdominal wall on large arrays, and no phase correction was applied.

Fig. 8.

Impact of gross sound speed error as a function of aperture size. Water path wire target images are shown after FSA beamforming with various levels of sound speed mismatch $\left(\mathbf{\text{Δ}}\mathit{c}\right)$. Images correspond to a 128 (top) and 384 (bottom) element aperture. Images are displayed on a 60 dB dynamic range. The right panel reports FWHM calculated as a function of sound speed mismatch and aperture size. FWHM was calculated at the depth of maximum envelope brightness. A linear rise in FWHM is due to a gradual increase in depth and apparent wavelength with $\mathbf{\text{Δ}}\mathit{c}$ (under binomial approximation).

Fig. 9.

Harmonic images of wire targets at multiple depths of the ATS phantom beamformed with selected apertures. Rows 1–3 show control and acquisitions through Samples 1 and 3, respectively. The bulk correction was not applied to the through-sample acquisitions. White ROIs on the left column indicate point targets used for FWHM measurements. Images are displayed on a 60 dB dynamic range.

Fig. 10.

Point resolution of wire targets at multiple depths of the ATS phantom. FWHM is reported as a function of the transmit/receive F-number for fundamental and harmonic modes. Columns 1–4 correspond to control and through-sample acquisitions, respectively. Each figure corresponds to a point target. For example, rows 1–4 correspond to the first, third, fifth, and seventh point targets from the top of each image in Fig. 9 (white ROIs). Along the rows, targets were at similar depths (indicated in the plots) but not co-registered.

Fig. 11.

Impact of bulk correction (ATS phantom). Harmonic B-mode images for a 384-element aperture without (left column) and with (middle column) bulk correction (B.C.) are displayed for three through-sample cases. Fundamental and harmonic FWHM (after B.C.) as a function of transmit/receive F# is reported in the right column. Due to the spatial variation, effective sound speed was calculated from and applied to only one isolated target in the three images (the top point target in the three through-sample cases of Fig. 9). Beamforming (effective) sound speed is reported on each image.

Fig. 12.

Fundamental and harmonic images of anechoic lesions in the ATS phantom as a function of aperture size. The first and last two rows correspond to control and acquisitions through Sample 3. The left column shows ROIs used in the contrast measurement. Lesions from the first two rows are not co-registered with those from the last two rows. ROIs were segmented manually. All images are displayed on a 60 dB dynamic range.

Fig. 13.

Contrast of anechoic lesions in the ATS phantom as a function of aperture size. (a)-(d) show measurements from control and acquisitions through Samples 1, 3, and 4, respectively. Red, blue, green, and yellow lines represent measurements from ROIs A, B, C, and D, shown in Fig. 12. Color-coded texts indicate the depths of lesion centers. Solid and dashed lines represent fundamental and harmonic acquisitions, respectively.

Fig. 14.

FIELD II simulations of random aberrations. Left panel shows B-mode images of a 1-cm diameter lesion and a point target for various apertures and aberrator RMS values. Images are displayed on a 60 dB dynamic range. Lesion contrast (ROIs shown on top left) are plotted on the top right figure. Standard deviations were calculated over 100 (5 speckle × 20 aberrator) and 5 realizations for aberrated and control cases, respectively. Point target FWHM is reported on the bottom right figure. Standard deviation was calculated over 20 aberrator realizations. Note that large RMS values (>60 ns) have increased FWHM variance.

Fig. 15.

Fundamental B-mode images of the liver of three subjects. Images are shown for five selected apertures. The left column shows the ROIs used for contrast measurement. The white sectors show the three depths used in the target segmentation. Red, green, and yellow ROIs represent targets at those depths. The contrast calculated from these ROIs is listed in the images using matched colors. Black arrows indicate the targets discussed in the results section. The solid white ROI at 8 cm depth was used to calculate temporal coherence (TC) from the repeat FOV acquisition. TC is indicated on the left column. All images are displayed on a 60 dB dynamic range.

Fig. 16.

Harmonic B-mode images of the liver of three subjects. Acquisitions, display layouts and symbols are matched to those of Fig. 15.

Fig. 17.

In vivo contrast measurements. The first three rows show hypoechoic target contrast as a function of aperture size in 13 subjects. Red, green, and yellow lines represent ROIs located at 7–9, 9–11, and 11–13 cm, respectively. Fundamental and harmonic data are shown in separate figures. For clear visualization, up to two lines per imaging depth are shown in the first three rows. Due to TC thresholds, the harmonic analysis used fewer targets and subjects (see Fig. 3). The last row shows the maximum contrast gain or loss for individual targets at three depths. All available targets after TC thresholding are shown in the last row. For targets where the smallest aperture provides maximum (absolute) contrast, a negative loss is reported relative to the worst contrast (contrast at smallest aperture - worst contrast). For all other targets, a positive gain is reported relative to the smallest aperture (contrast at smallest aperture - best contrast). Average gains and losses computed over the relevant targets are reported in absolute dB for fundamental (F) and harmonic (H) modes.

Fig. 18.

Comparison of fundamental and harmonic in vivo target contrast for various apertures. All viable harmonic targets (after TC thresholding) are included with matched fundamental images. Numbers indicate average contrast gain in harmonic mode.

Fig. 19.

Assessment of retrospective aperture synthesis. (Left panel) Harmonic speckle texture inside liver and blood vessels. Images are shown for two apertures (128 and 320 elements) that were physically transmitted (left column) and retrospectively synthesized from the largest aperture transmit (right column). Images correspond to Example C in Fig. 16. Similarity of images in the 128-element case illustrates both the efficacy of aperture synthesis and stationarity of the target during the 0.46 s acquisition period. (Right panel) Comparison of contrast measured from physically transmitted and retrospectively synthesized configurations. All viable targets in subjects (top row) and phantoms (bottom row) are shown for the four matched configurations. Average absolute differences are reported in dB for fundamental (F) and harmonic (H) modes. Note that smaller physical apertures were temporally distant from the largest aperture frame (Fig. 1).