Advertisement

Inter-transducer variability of ultrasound image quality in obese adults: Qualitative and quantitative comparisons

Open AccessPublished:September 29, 2022DOI:https://doi.org/10.1016/j.clinimag.2022.09.010

      Highlights

      • The quality of ultrasound images varies among different transducers and scanners
      • Ultrasound images quanlity can be quantitatively assessed
      • Optimal ultrasound images can be acquired in high BMI patients using the specially designed transducer

      Abstract

      Purpose

      Acquiring high-quality ultrasound images of deep abdominal organs and vasculatures in obese adults (BMI >30 kg/cm2) is considered challenging. The aim of the study was to assess the inter-transducer variability in B-mode and color Doppler image quality from four commercial ultrasound transducers through qualitative and quantitative analyses.

      Methods

      Four curvilinear transducers on three ultrasound scanners were used to acquire B-mode and color Doppler images of deep abdominal structures in 15 obesity ≥ class II (BMI >35 kg/cm2) adults. Using visual-qualitative assessment and an offline image processing software, visual-qualitative score and quantitative mean pixel values of B-mode images, and color area ratios of color Doppler images were calculated. Differences in these values among the transducers were analyzed using one-way ANOVA. The intra- and inter-observer reliability of visual-qualitative assessment and offline image processing was tested using the intraclass correlation coefficient (ICC).

      Results

      Differences in visual-qualitative score, mean pixel value of B-mode images, and color area ratio of color Doppler images among the four transducers were significant (p < 0.001). Transducer −4 produced the highest quality of B-mode (45–53% improvement) and color Doppler (22–73% improvement) images among the transducers. Intra-observer repeatability and inter-observer reproducibility were higher with performing offline image processing than visual-qualitative assessment (ICC: 0.97–0.99 versus ICC: 0.76–0.97).

      Conclusion

      There was significant image quality variability between different transducers. Transducer −4, a transducer designed specifically for high BMI patients, had the highest quality B-mode and color Doppler images compared to the other transducers lending to improved ultrasonographic visualization in obese patients.

      Keywords

      1. Introduction

      Image quality of abdominal ultrasound based on resolution and penetration greatly influences the diagnostic interpretation of pathology and performance of ultrasound-guided interventional procedures.
      • Huang D.
      • Yusuf G.
      • Daneshi M.
      • Ramnarine R.
      • Deganello A.
      • Sellers M.
      • Sidhu P.
      Contrast-enhanced ultrasound (CEUS) in abdominal intervention.
      Currently, commercially available ultrasound transducers have limitations in acquiring high quality images of deep abdominal organs and vasculatures. Abdominal image quality further decreases as patients' body mass indexes (BMI) increase.
      • Jeeji A.
      • Ekstein S.
      • Ifelayo O.
      • Oyemade K.
      • Tawfic S.
      • Hyde R.
      • et al.
      Increased body mass index is associated with decreased imaging quality of point-of-care abdominal aortic ultrasonography.
      It is considered technically challenging to acquire high quality ultrasound images while visualizing deep abdominal structures in obese populations (BMI >30 kg/cm2) due to increased ultrasound attenuation following increased tissue depth. Challenging abdominal structures to visualize include the abdominal aorta and the liver near the diaphragm, which both lie deep in the abdomen. The vasculature of the kidney is difficult to assess using Doppler technique due to slow flow through the microvasculature in the cortex and deep location of the renal artery.
      • Machann J.
      • Horstmann A.
      • Born M.
      • Hesse S.
      • Wolfgang Hirsch F.
      Diagnostic imaging in obesity.
      • Cwik G.
      • Gierblinski I.
      Errors and mistakes in the ultrasound diagnosis of the pancreas.
      Additionally, thickened subcutaneous fat along the abdominal wall results in ultrasound sound wave distortion and scattering, as well as subsequent loss of contrast and image sharpness. These obtainable images are qualitatively described as “limited” or “degraded”.

      Browne, J., Watson, A., Hoskins, P. Investigation of the effect of subcutaneous fat on image quality performance of 2D conventional imaging and tissue harmonic imaging.

      • Ihnatsenka B.
      • Boezaart A.P.
      Ultrasound: basic understanding and learning the language.
      As such, additional uses of magnetic resonance imaging (MRI) and/or computed tomography (CT) become necessary to have optimal and adequate images of deeply located structures for diagnostic interpretation in these populations.
      • Uppot R.N.
      Technical challenges of imaging and image-guided interventions in obese patients.
      This leads to increased healthcare cost, radiation exposure, and inefficient use of time.
      Although qualitative assessments are useful in determining optimization of anatomic structures, quantitative comparisons may clearly establish standards of imaging quality across various equipment and patient demographics. For instance, image quality can be quantified by measuring the ability of the transducer to detect reflected ultrasound waves as an emulation of the variability in refraction, such as the tissue echogenicity in B-mode image, and depiction of slow flow with color Doppler technique. Pixel counting analysis measures backscatter that is useful for determining the echo-intensity of grayscale images. Color area ratio measures the percentage of color pixels representing Doppler signals (flow in vessels) in a total region of interest.
      • Gao J.
      • Thai A.
      • Erpelding T.
      Comparison of superb microvascular imaging to conventional color Doppler ultrasonography in depicting renal cortical microvasculature.
      Recently studies are focused on comparison of shear wave elastography using scanners made by various ultrasound manufacturers. Yet, differences in quality of B-mode and color Doppler images in patients with high BMI have not been investigated between transducers.
      • Gress V.S.
      • Glawion E.N.
      • Schmidberger J.
      • Kratzer W.
      Comparison of liver shear wave elastography measurements using Siemens acuson S300, GE logiq E9, Philips EPIQ7 and Toshiba aplio 500 (software versions 5.0 and 6.0) in healthy volunteers.
      Furthermore, ultrasound probe selection greatly varies in the visibility and image quality for deep organs and structures. It is estimated that up to 30% of variation can occur when comparing tissue Doppler imaging between different machines, thus affecting clinical results and repeatability.
      • Cruz-Lemini M.
      • Valenzuela-Alcaraz B.
      • Figueras F.
      • Sitges M.
      • Gomez O.
      • Martinez J.M.
      • et al.
      Comparison of two different ultrasound systems for the evaluation of tissue Doppler velocities in fetuses.
      Even with adequately acquired images, systemic variability occurs when comparing image quality, artifacts, and Doppler sensitivity among vendors, transducers, and image processing software.
      • Brattain L.
      • Telfer B.
      • Dhyani M.
      • Grajo J.
      • Samir A.
      Machine learning for medical ultrasound: status methods, and future opportunities.
      It is important to identify clinically useful transducers that adequately visualize anatomic details and depict blood flow in high BMI populations for improving diagnostic accuracy and repeatability between users. This is especially true when patients require ongoing monitoring, follow-ups, or second opinions. It may also improve outcomes of ultrasound-guided procedures in patients with high BMI.
      This study aimed to compare the ability to acquire high-quality B-mode and color Doppler images of deep abdominal structures in obese populations (BMI >35 kg/cm2), among four different commercial ultrasound transducers using visual-qualitative assessment and quantitative analysis.

      2. Material and methods

      2.1 Participants

      The Institutional Review Board of the University approved the study (IRB# 2020-0007) and all participants provided written informed consent prior to ultrasound scanning. Inclusion criteria: Age of 20y or older; able to understand and provide written informed consent; body mass index (BMI) >35 kg/cm2; no major surgeries in the liver, kidney, and abdominal aorta; medically stable without cardiac, renal, or hepatic failure; tolerant to ultrasound.

      2.2 Ultrasound image acquisition

      All participants fasted for 6 h prior to ultrasound scanning. During scanning, subjects were positioned in lateral decubitus or supine position to best view deep abdominal structures. We used four commercial curvilinear transducers on three ultrasound scanners (transducer-1 on ultrasound scanner A; transducer-2 on scanner B; transducer-3 and transducer-4 on scanner C) in the study. The scanning protocol and machine settings (scanning frequency [3.0 MHz for B-mode and 2.5 MHz for color Doppler], grayscale gain: 0 dB, single image focus, dynamic range: 65, time gain compensation, color gain, pulse repetition frequency, harmonic imaging) were standardized and preset.
      Commercialized four curvilinear transducers on three ultrasound scanners were used to acquire the following images. One of the transducers was designed with advanced Multi-D beam forming technique specifically for high BMI patients (Table 1).
      • 1.
        B-mode images of the liver/diaphragm and proximal abdominal aorta (coronal view);
      • 2.
        Color Doppler images of the portal vein, entire right renal artery (proximal to distal, from abdominal aorta to kidney hilus), and kidney cortex (near and mid portion of the kidney in sagittal plane).
      Table 1Material and technology of four curvilinear transducers.
      TransducerMaterialElementsBandwidthTechnology
      1Single crystal piezoelectric1921.0–6.0 MHzWider bandwidth transducer has better harmonic imaging, axial resolution, and higher sensitivity, which improves penetration and contrast resolution.
      2Hanafy lens1921.5–6.0 MHzThe transducer has a uniform slice thickness throughout the depth, therefore maintains image uniformity and contrast resolution in both near and far field.
      3Single crystal piezoelectric1801.0–5.7 MHzWider bandwidth transducer has better harmonic imaging, axial resolution, and higher sensitivity, which improves penetration and contrast resolution.
      4Multi-D array2881.0–3.5 MHzAdvanced Multi-D beam forming transducer controls the beam thickness, beam formation in both transmissing and receiving phases to enable deep penetration. High element density transducer allows dynamic elevation aperture control as needed.
      The listed images were acquired from all participants for each of the four transducers. Ultrasound images (B-mode and color Doppler) with Digital Imaging and Communication in Medicine (DICOM) format were transferred from ultrasound scanners to a desktop computer for offline visual-qualitative assessment and image processing.
      A single ultrasound expert with 30 years of experience in abdominal ultrasound performed all ultrasound scans to prevent inter-observer variations in image acquisition.

      2.3 Visual-qualitative assessment of ultrasound images

      Visual-qualitative assessment of ultrasound images of deep organ (the right lobe of the liver/diaphragm) and vasculature (coronal view of the proximal abdominal aorta) was conducted using a modified Brightness Mode Quality Ultrasound Imaging Examination Technique (B-QUIET) method.
      • Bahner D.P.
      • Adkins E.J.
      • Nagel R.
      • Way D.
      • Werman H.A.
      • Royall N.A.
      Brightness mode quality ultrasound imaging examination technique.
      Modified B-QUIET method included eight components of visual-quantitative assessment: anatomy identification, resolution, depth, gain, near field, far field, receding edge (right side of image), and leading edge (left side of image). Because our study used de-identified ultrasound images, the original B-QUIET score was modified to exclude subscale for date/time, body marker, and patient's information. Based on the modified B-QUIET scores, we assessed B-mode image quality of the proximal abdominal aorta (coronal view) and liver/diaphragm. The image quality was scored as 1: unacceptable; 2: needs improvement; 3: acceptable; and 4: optimal for interpretation.
      • Bahner D.P.
      • Adkins E.J.
      • Nagel R.
      • Way D.
      • Werman H.A.
      • Royall N.A.
      Brightness mode quality ultrasound imaging examination technique.
      The average of all scores in eight components was used for analysis. Each of two operators (M.M. and J.L.) performed visual-qualitative assessment on the same ultrasound images recorded from 15 participants two times (7 days apart). The two operators conducted all visual-qualitative assessments independently.

      2.4 Offline image processing

      There are various factors influencing the resolution and brightness of displayed ultrasound images. Technical variations in transducers and machine design of signal processing contribute to differences in brightness on B-mode images and color signals on color Doppler images. Raw radiofrequency (RF) ultrasound data are scanner-independent and can provide the most accurate ultrasound data to quantify backscatter coefficient and attenuation coefficient of the tissue. However, RF data is currently not available for radiologists to interpret clinical ultrasound examinations. In clinical patient care, radiologists view B-mode and color Doppler ultrasound images transferred from commercialized ultrasound scanners to the picture archiving communication system (PACS). Therefore, we used standardized machine settings to acquire ultrasound images and the corresponding off-line software and protocol to process all ultrasound images.
      • Gao J.
      • Mennitt K.
      • Belfi L.
      • Zheng Y.Y.
      • Chen Z.
      • Rubin J.M.
      Green tagging in displaying color doppler aliasing: a comparison to standard color mapping in renal artery stenosis.
      We used an offline image processing software, GetColorPixels (Shanghai Jiaotong University, Shanghai, China) to quantify all B-mode and color Doppler images.
      • Gao J.
      • Thai A.
      • Erpelding T.
      Comparison of superb microvascular imaging to conventional color Doppler ultrasonography in depicting renal cortical microvasculature.
      • Gao J.
      • Mennitt K.
      • Belfi L.
      • Zheng Y.Y.
      • Chen Z.
      • Rubin J.M.
      Green tagging in displaying color doppler aliasing: a comparison to standard color mapping in renal artery stenosis.
      A standardized size of the region of interest (ROI) to assess mean pixel counts representing backscatter (echo-intensity) of B-mode images (Fig. 1) and color area ratio (area of color/total area of ROI) of color Doppler images (Fig. 2). ROIs were maintained for each anatomical landmark throughout the study. The size of a rectangular ROI to calculate the color area ratio in the kidney cortex was 0.25 cm2. The ROIs for all other landmarks (liver-subcostal, liver-intercostal, proximal abdominal aorta, main portal vein, and right renal artery) were set to 0.5 cm2.
      Fig. 1
      Fig. 1a-b. Grayscale image of the proximal abdominal aorta in coronal view (aorta). Using offline image processing software, mean pixel in the region of interest (0.5 cm2) is 19.11 (a), 38.33 (b), 35.5 (c), and 70.59 (d) using transducer-1 (a), −2 (b), −3 (c), and − 4 (d), respectively. The highest mean pixel value is in the image obtained using Transducer-4 (d).
      Fig. 1
      Fig. 1a-b. Grayscale image of the proximal abdominal aorta in coronal view (aorta). Using offline image processing software, mean pixel in the region of interest (0.5 cm2) is 19.11 (a), 38.33 (b), 35.5 (c), and 70.59 (d) using transducer-1 (a), −2 (b), −3 (c), and − 4 (d), respectively. The highest mean pixel value is in the image obtained using Transducer-4 (d).
      Fig. 1
      Fig. 1a-b. Grayscale image of the proximal abdominal aorta in coronal view (aorta). Using offline image processing software, mean pixel in the region of interest (0.5 cm2) is 19.11 (a), 38.33 (b), 35.5 (c), and 70.59 (d) using transducer-1 (a), −2 (b), −3 (c), and − 4 (d), respectively. The highest mean pixel value is in the image obtained using Transducer-4 (d).
      Fig. 1
      Fig. 1a-b. Grayscale image of the proximal abdominal aorta in coronal view (aorta). Using offline image processing software, mean pixel in the region of interest (0.5 cm2) is 19.11 (a), 38.33 (b), 35.5 (c), and 70.59 (d) using transducer-1 (a), −2 (b), −3 (c), and − 4 (d), respectively. The highest mean pixel value is in the image obtained using Transducer-4 (d).
      Fig. 2
      Fig. 2a-b. To assess color Doppler sensitivity in depicting small vessels in the kidney cortex, we measured color area ratio (color area/area in total region of interest) in color Doppler image of the longitudinal right kidney. The region of interest for measuring color area ratio is place in near/mid portion of the kidney cortex. Color area ratio calculations were 0.19 (a), 0.35 (b), 0.38 (c), and 0.56 (d) using Transducer-1, -2, -3, and -4, respectively.
      Fig. 2
      Fig. 2a-b. To assess color Doppler sensitivity in depicting small vessels in the kidney cortex, we measured color area ratio (color area/area in total region of interest) in color Doppler image of the longitudinal right kidney. The region of interest for measuring color area ratio is place in near/mid portion of the kidney cortex. Color area ratio calculations were 0.19 (a), 0.35 (b), 0.38 (c), and 0.56 (d) using Transducer-1, -2, -3, and -4, respectively.
      Fig. 2
      Fig. 2a-b. To assess color Doppler sensitivity in depicting small vessels in the kidney cortex, we measured color area ratio (color area/area in total region of interest) in color Doppler image of the longitudinal right kidney. The region of interest for measuring color area ratio is place in near/mid portion of the kidney cortex. Color area ratio calculations were 0.19 (a), 0.35 (b), 0.38 (c), and 0.56 (d) using Transducer-1, -2, -3, and -4, respectively.
      Fig. 2
      Fig. 2a-b. To assess color Doppler sensitivity in depicting small vessels in the kidney cortex, we measured color area ratio (color area/area in total region of interest) in color Doppler image of the longitudinal right kidney. The region of interest for measuring color area ratio is place in near/mid portion of the kidney cortex. Color area ratio calculations were 0.19 (a), 0.35 (b), 0.38 (c), and 0.56 (d) using Transducer-1, -2, -3, and -4, respectively.
      Finally, the thickness of the subcutaneous fat was measured using B-mode images (DICOM format) on images of the proximal abdominal aorta and right lobe of the liver/diaphragm.

      2.5 Statistical analysis

      Modified B-QUIET scores, mean pixel values, and color area ratios were collected using Microsoft Excel (Microsoft Inc.). All variables were expressed by the mean and standard deviation (SD). Differences in subcutaneous fat thickness, visual-qualitative assessment scores, and mean pixel value of B-mode images, color area ratio of color Doppler images were compared using one-way analysis of variance (ANOVA) and post-hoc to determine differences among four transducers and each paired transducer. Results of the echo-intensity of proximal abdominal aorta and color area ratios of renal cortex were also displayed with box-and-whisker plots (Figs. 3 and 4). To test intra- and inter-observer reliability in performing visual-qualitative assessment and offline image processing, single operator processed visual-qualitative assessment and B-mode and color Doppler images two times in 15 participants and two observers independently performed visual-qualitative assessment and image processing on the same 15 participants. Intra-observer repeatability and inter-observer reproducibility were analyzed using the intraclass correlation coefficient (ICC). A p-value of <0.05 was considered statistically significant. All statistical analyses were conducted using commercial software of SPSS (SPSS, 28.0 version, IBM) and Excel (Microsoft, Pota Lato, CA).
      Fig. 3
      Fig. 3Box-and-whisker plots show a significant difference in grayscale pixel value across Transducer 1–3 and Transducer-4 in one-way ANOVA and paired groups (post-hoc test) analysis. No significant difference was found between Transducer-1 and Transducer-2 or between Transducer-2 and Transducer-3 in post-hoc testing. The mean (x) and standard deviation (SD) of mean pixel value of the proximal aorta (coronal view) imaged using Transducer-1 (blue box), -2 (orange box), -3 (gray box), and -4 (yellow box) measures 30.61 ± 19.56, 28.03 ± 12.58, 27.99 ± 13.47, and 69.02 ± 22.34, respectively. Note: ns, non-significant; ***, P < 0.001.
      Fig. 4
      Fig. 4Box-and-whisker plots show significant differences in color area ratio between Transducer-1 and Transducer-2, between Transducer-3 and Transducer-4 in one-way ANOVA and paired groups (post-hoc test) analysis. No significant difference was found between Transducer-2 and Transducer-3 in post-hoc testing. The mean (x) and standard deviation (SD) of color area ratio of the kidney cortex imaged using Transducer-1 (blue box), -2 (orange box), -3 (gray box), and -4 (yellow box) measures 0.16 ± 0.13, 0.37 ± 0.26, 0.46 ± 0.21, and 0.59 ± 0.16, respectively. Note: Color area ratio = color area/total area in the region of interest. ns, non-significant; **, P < 0.01; ***, P < 0.001.

      3. Results

      From November 2020 to May 2021, we successfully performed B-mode and color Doppler ultrasonography of the abdomen on 15 adult volunteers (9 men, 6 women; mean age 54y, age range 25y to 73y) who were classified into class II and III obesity (mean BMI 38.34 kg/cm2, BMI range 35–46.2 kg/cm2). The material and technology used for designing four transducers are listed in Table 1. The subcutaneous fat thickness on liver/diaphragm and proximal abdominal aorta images was not significantly different across the four transducers (Table 2). However, the differences in visual-qualitative assessment score of the B-mode ultrasound images were statistically significant across the four transducers (Table 3).
      Table 2Thickness of subcutaneous fat and visual-qualitative assessment of ultrasound images.
      Transducer1234(p)
      p values are based on one-way analysis of variance (ANOVA).
      Subcutaneous fat thickness (cm)
      Liver/diaphragm1.5 ± 0.411.51 ± 0.421.53 ± 0.41.55 ± 0.40.98
      Proximal abdominal aorta1.5 ± 0.411.51 ± 0.411.54 ± 0.381.54 ± 0.40.99
      Visual-qualitative assessments were conducted using a modified B-QUIET method.12
      Visual Quality Assessment score
      Liver/diaphragm2.02 ± 0.561.95 ± 0.762.23 ± 0.613.29 ± 0.47<0.001
      Proximal abdominal aorta1.87 ± 0.612.08 ± 0.652.50 ± 0.653.23 ± 0.69<0.001
      Note: The method uses a 4-point Likert scale for each of the eight components (anatomy identification, resolution, depth, gain, near field, far field, receding edge [image right side], leading edge [image left side]). 4-point scale is defined as 1: unacceptable; 2: needs improvement; 3: acceptable; and 4: optimal. Each operator scored the individual ultrasound image using the modified B-QUIET method. The average of eight components of visual-qualitative assessment scores was used for analysis.
      low asterisk p values are based on one-way analysis of variance (ANOVA).
      a Visual-qualitative assessments were conducted using a modified B-QUIET method.
      • Bahner D.P.
      • Adkins E.J.
      • Nagel R.
      • Way D.
      • Werman H.A.
      • Royall N.A.
      Brightness mode quality ultrasound imaging examination technique.
      Table 3Comparison of mean pixel value and color area ratio among the four ultrasound transducers.
      Transducer1234(p)
      p value is based on one-way analysis of variance.
      Gray mean pixel value
      Liver-Subcostal52.20 ± 28.3936.93 ± 24.1937.12 ± 18.6586.35 ± 19.56<0.001
      Liver-Intercostal55.76 ± 26.0152.33 ± 28.6450.08 ± 17.5694.30 ± 18.92<0.001
      Proximal abdominal aorta30.61 ± 19.5628.03 ± 12.5827.99 ± 13.4769.02 ± 22.34<0.001
      % difference (pixel value)#1−45.55%−53.02%−52.81%
      Color area ratio
      Color are ratio = color area/total area in the region of interest.
      Main portal vein0.12 ± 0.100.31 ± 0.160.37 ± 0.190.50 ± 0.14<0.001
      Right renal artery0.19 ± 0.150.41 ± 0.180.53 ± 0.230.66 ± 0.18<0.001
      Kidney cortical perfusion0.16 ± 0.130.37 ± 0.260.46 ± 0.210.59 ± 0.16<0.001
      % difference (color area ratio)#2−73.14%−37.71%−22.29%
      Note: #1 represents total increase (%) in mean pixel value in 3 B-mode images obtained by Transducer-4 compared with the respective probe; #2 represents lower (%) color area ratio in 3 color Doppler images obtained by 3 transducers compared with that obtained by the Transducer-4. % difference (pixel value#1 or color area ratio#2) = (transducer x − transducer-4) / transducer-4 × 100%.
      low asterisk p value is based on one-way analysis of variance.
      a Color are ratio = color area/total area in the region of interest.
      Mean pixel value representing echo-intensity was used for determining amplitude of backscatter indicating the quality (resolution and penetration) of B-mode images. The color area ratio, a quantitative assessment of the sensitivity of color Doppler in depicting Doppler flow signals, was used for determining the quality of color Doppler images. Differences in echo-intensity of B-mode images and color area ratio of color Doppler images among the four ultrasound transducers were highly significant (p < 0.001, Table 3). Of note, Transducer-4 produced the highest quality of B-mode and color Doppler images among the transducers and consistently yielded higher grayscale pixel intensity (45–53% improvement) and color area ratios (22–73% improvement) when compared to Transducer 1, 2, and 3. In comparison to Transducer 1–3, the improvement rate (%) of mean gray pixel for Transducer-4 was 45.55%, 53.02%, 52.81%, respectively, and the improvement rate (%) of mean color area ratio for Transducer-4 was 73.14%, 37.71%, 22.29%, respectively (Table 3). Mean pixel value of the proximal abdominal aorta tabulated for Transducer-4 (Fig. 1a-d) demonstrated higher image resolution and penetration when compared to Transducers 1–3. In depicting kidney cortical microvasculature using color Doppler, the sensitivity of Transducer-4 was also significantly higher compared to Transducer 1–3 (Fig. 2a-d). Intra-observer repeatability and inter-observer reproducibility in performing visual-qualitative assessment and offline image processing were moderate to excellent (ICC: 0.76–0.97) and excellent (ICC: 0.97–0.99) (Table 4), respectively.
      Table 4Repeatability and reproducibility of off-line image processing and visual-qualitative assessment
      Average measuresICC
      ICC, intraclass correlation coefficient using a consistency definition.
      95% confidence intervalF test with true value 0
      Lower boundUpper boundValueSignificance
      Quantitative image processing
      Operator-1:Operator-1:pixel count0.990.980.9976.26<0.001
      Operator-2:Operator- 2:pixel count0.980.970.9957.79<0.001
      Operator-1:Operator-2:pixel count0.970.950.9833.85<0.001
      Operator-1:Operator-1:Color ratio0.990.990.99130.73<0.001
      Operator-2:Operator-2:Color ratio0.990.980.9976.26<0.001
      Operator-1:Operator-2:Color ratio0.970.960.9939.15<0.001
      Visual-qualitative assessment
      Operator-1:Operator-1 (AO)0.970.950.9832.67<0.001
      Operator-2:Operato-2 (AO)0.940.890.9615.30<0.001
      Operator-1:Operator-2 (AO)0.760.600.864.15<0.001
      Operator-1:Operator-1 (liver)0.940.890.9615.21<0.001
      Operator-2:Operato-2 (liver)0.890.810.938.91<0.001
      Operator-1:Operator-2 (liver)0.840.730.906.05<0.001
      Note: Inter-observer reliability test: measurements/scores performed by operator-1 to measurements/scores performed by operator-2 on the same images independently; intra-observer reliability tests: measurement/scores 1 to measurement/scores 2 on the same images obtained by the operator-1 (operator-2). Pixel count, gray mean pixel value in B-mode ultrasound image; Color ratio, color area ratio = color area/total area in the region of interest on color Doppler image. Visual image quality assessment is presented by ultrasound image quality score evaluated using a modified Brightness Mode Quality Ultrasound Imaging Examination Technique (B-QUIET). AO, proximal abdominal aorta; liver, the area of liver right lobe/diaphragm.
      a ICC, intraclass correlation coefficient using a consistency definition.

      4. Discussion

      We have observed that Transducer-4 possessed the highest quality of B-mode grayscale and color Doppler images when compared to the other three transducers. This transducer was specifically designed to image deep structures in patients with high BMI (>30 kg/cm2) and has proven to produce superior B-mode image penetration and color Doppler sensitivity among the tested transducers with 45–53% improvement in mean pixel value and 22–73% improvement in mean color area ratio. Through image processing, mean pixel value quantifies the amplitude of backscatter from highly attenuated deep tissues. Color area ratio represents the ability of color Doppler in depicting weak Doppler signals produced by slow flow within the kidney cortical microvasculature.
      • Gao J.
      • Thai A.
      • Erpelding T.
      Comparison of superb microvascular imaging to conventional color Doppler ultrasonography in depicting renal cortical microvasculature.
      Both are effective means of comparable quantification of image quality across transducers. We have also demonstrated excellent intra-observer repeatability and inter-observer reproducibility in performing offline image processing for quantifying grayscale pixels in B-mode images and color area ratios in color Doppler images, which proved to be superior to visual-quantitative assessment of image quality.
      In clinical ultrasonography, common factors affecting ultrasound image quality are the experience of the operator, the body habitus of the patient, and software/hardware designed for image processing and production (scanner/transducer) made by manufacturers.
      • Farsalinos K.
      • Daraban A.
      • Unlu S.
      • Thomas J.
      • Badano L.
      • Voigt J.U.
      Head-to-head comparison of global longitudinal strain measurements among nine different vendors.
      In this study, one experienced operator scanned all participants to avoid inter-observer variation in ultrasound image acquisition and the difference in the subcutaneous fat thickness on acquired ultrasound images among the four transducers was not significant. To test the ability of the transducers to acquire high quality images in technically challenging populations, we recruited participants with obesity classes II and III as increased thicknesses of subcutaneous tissue and intraperitoneal fat in obesity result in high ultrasound attenuation and poor beam penetration.
      • Uppot R.N.
      Impact of obesity on radiology.
      • Uppot R.N.
      • Dushyant S.V.
      • Hahn P.F.
      • Gervais D.
      Impact of obesity on medical imaging and image-guided intervention.
      Previously suggested solutions to imaging obese subjects included the use of lower frequency transducer to improve ultrasound penetration.
      • Uppot R.N.
      • Dushyant S.V.
      • Hahn P.F.
      • Gervais D.
      Impact of obesity on medical imaging and image-guided intervention.
      However, using low frequency transducers to image deep organ structures incurs the tradeoff of decreased resolution of B-mode image and sensitivity of color Doppler.
      • Uppot R.N.
      Impact of obesity on radiology.
      • Uppot R.N.
      • Dushyant S.V.
      • Hahn P.F.
      • Gervais D.
      Impact of obesity on medical imaging and image-guided intervention.
      The subcutaneous tissue thickness reflects the distribution of adipose tissue, which directly affects ultrasound image quality due to strong beam distortion and phase aberration in the subcutaneous fat and subsequent decrease in the penetration of the sound beam to internal organs.
      • Browne J.E.
      • Watson A.J.
      • Hoskins P.R.
      • Elliott A.T.
      Investigation of the effect of subcutaneous fat on image quality performance of 2D conventional imaging and tissue harmonic imaging.
      Therefore, we quantitatively compared the image quality acquired from participants with similar thickness of subcutaneous fat (Table 2). Our study provides strong evidence that Transducer-4, which was designed for the high BMI population, can generate high quality images in ultrasound resolution and penetration. As obesity continues to rise worldwide, using the specialized transducer may reduce the burden of additional imaging (CT and/or MRI) in this population.
      Based on quantitative comparisons, Transducer-4 showed significantly higher quality in both B-mode and color Doppler images while the other transducers demonstrated relatively lower image quality to varying degrees. For example, Transducer-1 had similar values of grayscale pixel counts compared to Transducer-2 and Transducer-3; however, the quality of color Doppler imaging was comparatively much lower as assessed by the differences of color area ratio. Transducer-4 consistently demonstrated higher quality results in all quantitative parameters, making it a simple and easy option when imaging obese populations. The major advantage of probes, like Transducer-4, can help the operator to acquire high quality ultrasound images and reduce the scanning time in patient care.
      There were several limitations in the study. First, the sample size of the study was small. Second, we performed B-mode and color Doppler on participants with class II and III obesity based on classification of BMI. Although BMI has been useful in classifying the degree of obesity in clinic, information related to patient waist circumference/absolute weight and distribution of adipose tissue may help the operator to select the proper transducer for obtaining high quality ultrasound images.
      • Uppot R.N.
      Impact of obesity on radiology.
      Third, the same operator performed all scans in the study to reduce the variability in scanning skills between operators. Further studies to test inter-observer reproducibility in ultrasound scanning are warranted. Finally, the variation in transducer design, image processing, and hardware development related to ultrasound imaging made by different vendors may contribute to results of significant differences in image quality among the four transducers, even with standardized machine settings, scanning protocol, and offline image processing in the study. Future assessments of differences in image quality in relations to clinical outcome, scanning time, imaging cost, technical pitfalls, and/or incident findings among variable transducers are warranted.

      5. Conclusion

      It is important to select a transducer that produces high quality B-mode and color Doppler images in clinical patient care. It is particularly challenging when performing ultrasound on high BMI individuals. The study results suggest that Transducer-4, which is designed with high element density allowing dynamic elevation aperture control for the obese population, shows the highest B-mode image quality and color Doppler sensitivity in imaging deep organs and depicting microvasculature compared to the other three transducers in high BMI participants. With the increasing incidence of obesity, the improved image quality of Transducer-4 demonstrates the benefit for the general use of specialized transducers to overcome the limitation of the conventional abdominal transducer in imaging deep organs and vasculatures in obese patients. Our results also demonstrate that the quality of ultrasound images significantly varies from one transducer to another. Therefore, it is encouraged the operator to try different ultrasound transducers and/or machines to acquire higher quality ultrasound images in clinical patient care, particularly in the context of challenging obese populations and imaging deep organs and vasculatures.

      Acknowledgment

      Authors thank Siemens Healthineers for research grant and equipment to support the study.

      References

        • Huang D.
        • Yusuf G.
        • Daneshi M.
        • Ramnarine R.
        • Deganello A.
        • Sellers M.
        • Sidhu P.
        Contrast-enhanced ultrasound (CEUS) in abdominal intervention.
        Abdom Radiol (NY). 2018; 43: 960-976
        • Jeeji A.
        • Ekstein S.
        • Ifelayo O.
        • Oyemade K.
        • Tawfic S.
        • Hyde R.
        • et al.
        Increased body mass index is associated with decreased imaging quality of point-of-care abdominal aortic ultrasonography.
        J Clin Ultrasound. 2020; 49: 328-333
        • Machann J.
        • Horstmann A.
        • Born M.
        • Hesse S.
        • Wolfgang Hirsch F.
        Diagnostic imaging in obesity.
        Best Pract Res Clin Endocrinol Metab. 2013; 27: 261-277
        • Cwik G.
        • Gierblinski I.
        Errors and mistakes in the ultrasound diagnosis of the pancreas.
        J Ultrason. 2013; 13: 178-191
      1. Browne, J., Watson, A., Hoskins, P. Investigation of the effect of subcutaneous fat on image quality performance of 2D conventional imaging and tissue harmonic imaging.

        • Ihnatsenka B.
        • Boezaart A.P.
        Ultrasound: basic understanding and learning the language.
        Int J Should Surg. 2010; 4: 55-62
        • Uppot R.N.
        Technical challenges of imaging and image-guided interventions in obese patients.
        Br J Radiol. 2018; 91: 20170931
        • Gao J.
        • Thai A.
        • Erpelding T.
        Comparison of superb microvascular imaging to conventional color Doppler ultrasonography in depicting renal cortical microvasculature.
        Clin Imaging. 2019; 58: 90-95
        • Gress V.S.
        • Glawion E.N.
        • Schmidberger J.
        • Kratzer W.
        Comparison of liver shear wave elastography measurements using Siemens acuson S300, GE logiq E9, Philips EPIQ7 and Toshiba aplio 500 (software versions 5.0 and 6.0) in healthy volunteers.
        Ultraschall in Med. 2019; 40: 504-512
        • Cruz-Lemini M.
        • Valenzuela-Alcaraz B.
        • Figueras F.
        • Sitges M.
        • Gomez O.
        • Martinez J.M.
        • et al.
        Comparison of two different ultrasound systems for the evaluation of tissue Doppler velocities in fetuses.
        Fetal Diagn Ther. 2016; 40: 35-40
        • Brattain L.
        • Telfer B.
        • Dhyani M.
        • Grajo J.
        • Samir A.
        Machine learning for medical ultrasound: status methods, and future opportunities.
        Abdom Radiol. 2018; 43: 786-799
        • Bahner D.P.
        • Adkins E.J.
        • Nagel R.
        • Way D.
        • Werman H.A.
        • Royall N.A.
        Brightness mode quality ultrasound imaging examination technique.
        J Ultrasound Med. 2011; 30: 1649-1655
        • Gao J.
        • Mennitt K.
        • Belfi L.
        • Zheng Y.Y.
        • Chen Z.
        • Rubin J.M.
        Green tagging in displaying color doppler aliasing: a comparison to standard color mapping in renal artery stenosis.
        Ultrasound Med Biol. 2013; 39: 1976-1982
        • Farsalinos K.
        • Daraban A.
        • Unlu S.
        • Thomas J.
        • Badano L.
        • Voigt J.U.
        Head-to-head comparison of global longitudinal strain measurements among nine different vendors.
        J Am Soc Echocardiography. 2015; 28: 1171-1181
        • Uppot R.N.
        Impact of obesity on radiology.
        Radiol Clin North Am. 2007; 45: 231-246
        • Uppot R.N.
        • Dushyant S.V.
        • Hahn P.F.
        • Gervais D.
        Impact of obesity on medical imaging and image-guided intervention.
        AJR. 2007; 188: 433-440
        • Browne J.E.
        • Watson A.J.
        • Hoskins P.R.
        • Elliott A.T.
        Investigation of the effect of subcutaneous fat on image quality performance of 2D conventional imaging and tissue harmonic imaging.
        Ultrasound Med Biol. 2005; 31: 957-964