A modular deep learning pipeline for enhanced plane-wave beamforming and B-mode image quality.

We present a new spatial compounding method to improve the contrast of ultrasonic images for non-delayed sequential beamforming (NDSB). Sequential beamforming adopts more than one beamformer to reconstruct B-mode images which has the advantage of simple front-end electronics and fast data transfer rate. Via field pattern analysis, we propose a compounding method where two more sub-images can be reconstructed along with the NDSB sub-image. These sub-images can be seen as being produced with different transmit origins; thus, their summation enhances image contrast. Image quality was analyzed in terms of spatial resolution, contrast ratio (CR), and contrast-to-noise ratio (CNR). The proposed compounding method improves the lateral resolution up to 41%. In vitro results confirm a 13.0-dB CR and 4.0-dB CNR improvement. In vivo results reveal 10.9-dB and 6.0-dB improvement in CR and CNR for cross-section jugular vein and 8.0-dB and 4.5-dB improvement in CR and CNR for the longitudinal-section carotid artery.

To compare the image quality and acquisition time of thoracic great vessels using a novel, 3D isotropic non-contrast-enhanced MR angiography that combines respiratory compensation, electrocardiogram-triggering, compressed sensing, and Dixon water-fat separation at 3T (CS-MRA) to a conventional MRA employing parallel imaging and spectral attenuated inversion recovery (cMRA). This study included 20 healthy volunteers and 19 patients with thoracic vascular diseases. All scans were performed on a 3T MRI scanner using a T2-prepared 3D fast low angle shot (FLASH) sequence. Visual image quality of thoracic great vessels was assessed using a 4-point scale. Quantitative parameters including SNR, contrast-to-noise ratio (CNR), contrast ratio (CR), and coefficient of variation (CV) were evaluated. Aortic diameters were measured, and reproducibility was assessed using Bland-Altman analysis and the intraclass correlation coefficient (ICC). Image acquisition times were also compared. Visual assessment demonstrated significantly higher scores with CS-MRA compared to cMRA across all evaluated thoracic vessels in both volunteers and patients (median score: 3 vs. 2, P < 0.05). Quantitative analysis showed that CS-MRA yielded significantly higher SNR, CNR, and CR (P < 0.05 for all), along with better signal homogeneity (lower CV). CS-MRA showed excellent intra- and inter-observer agreement for aortic diameter measurements, with ICCs ≥ 0.91 in volunteers and ≥ 0.95 in patients. CS-MRA also achieved significantly shorter acquisition times (volunteers: 5.3 ± 1.4 min vs. 6.0 ± 1.7 min, P < 0.01; patients: 5.1 [4.6-6.3] min vs. 6.9 [5.0-9.6] min, P < 0.01). CS-MRA provides superior image quality, comparable measurement reproducibility, and significantly reduced acquisition times compared to cMRA. These findings support the clinical utility of CS-MRA as an efficient and reliable non-contrast technique for thoracic vascular evaluation.

In mammography, thick or dense breast regions persistently suffer from reduced contrast-to-noise ratio (CNR) because of degraded contrast from large scatter intensities and relatively high noise. Area x-ray beam equalization can improve image quality by increasing the x-ray exposure to under-penetrated regions without increasing the exposure to other breast regions. Optimal equalization parameters with respect to image quality and patient dose were determined through computer simulations and validated with experimental observations on a step phantom and an anthropomorphic breast phantom. Three parameters important in equalization digital mammography were considered: attenuator material (Z = 13–92), beam energy (22–34 kVp) and equalization level. A Mo/Mo digital mammography system was used for image acquisition. A prototype 16 × 16 piston driven equalization system was used for preparing patient-specific equalization masks. Simulation studies showed that a molybdenum attenuator and an equalization level of 20 were optimal for improving contrast, CNR and figure of merit (FOM = CNR2/dose). Experimental measurements using these parameters showed significant improvements in contrast, CNR and FOM. Moreover, equalized images of a breast phantom showed improved image quality. These results indicate that area beam equalization can improve image quality in digital mammography.

Computed tomography angiography (CTA) provides significant information on image quality in vascular imaging, thus offering high-resolution images despite having the disadvantages of increased radiation doses and contrast agent-related side effects. The deep-learning image reconstruction strategies were used to quantitatively evaluate the enhanced contrast-to-noise ratio (CNR) and the dose reduction effect of subtracted images. This study aimed to elucidate a comprehensive understanding of the quantitative image quality features of the conventional filtered back projection (FBP) and the advanced intelligent clear-IQ engine (AiCE), a deep learning reconstruction technique. The comparison was made in subtracted images with variable concentrations of contrast agents at variable tube currents and voltages, enhancing our knowledge of these two techniques. Data were obtained using a state-of-the-art 320-detector CT scanner. Image reconstruction was performed using FBP and AiCE with various intensities. The image quality evaluation was based on eight iodine concentrations in the phantom setup. The efficiency of AiCE relative to FBP was assessed by computing parameters including the root mean square error (RMSE), dose-dependent CNR, and potential dose reduction. The results showed that elevated concentrations of iodine and increased tube currents improved AiCE performance regarding CNR enhancement compared to FBP. AiCE also demonstrated a potential dose reduction ranging from 13.7 to 81.9% compared to FBP, suggesting a significant reduction in radiation exposure while maintaining image quality. The employment of deep learning image reconstruction with AiCE presented a significant improvement in CNR and potential dose reduction in CT angiography. This study highlights the potential of AiCE to improve vascular image quality and decrease radiation exposure risk, thereby improving diagnostic precision and patient care in vascular imaging practices.

By using plane wave for high frame rate imaging, the image quality is relatively poor due to the lack of transmit focusing. This paper proposes a deep-learning based approach to achieve two-way dynamic focusing in plane wave imaging, which improves the image quality of plane wave, particularly in image contrast. This method takes advantage of the complete dataset of synthetic transmit aperture (STA) imaging, by building the linear relationship between the channel data of plane waves with apodizations and the complete dataset of STA. A customized network is proposed to train a set of optimized plane wave apodization coefficients, and to recover the complete dataset of STA from the channel data of apodized plane waves. Compared with a similar method using fixed apodizations and compressed-sensing (CS) based dataset recovery in in-vivo experiments, the proposed method significantly improves the contrast ratio (CR) by around 9 dB and contrast-to-noise ratio (CNR) by around 6 dB. Meanwhile, the computational time of dataset recovery is reduced from thousands of seconds to less than 2 seconds.

Pulse-inversion-based tissue harmonic imaging has been utilized for many years because it can effectively eliminate the harmonic leakage and produce low side-lobe. However, the pulse inversion method is sensitive to imaging object movements, which may result in motion artifacts. Spatial resolution and contrast were limited. To improve ultrasound image quality by a new pulse-inversion-based tissue harmonic imaging technique. Continuous wavelet transform is applied to investigate the correlation between mother wavelet and the received echoes from two opposite pulses. To get a better correlation, a novel mother wavelet named 'tissue wavelet' is designed based on the Khokhlov-Zabolotskaya- Kuznetsov (KZK) wave equation. Radio frequency data were obtained from open Ultrasonix SonixTouch imaging system. Experiments were carried on ultrasonic tissue phantom, human carotid artery and human liver. The average improvement of lateral spatial resolution is 49.52% compared to pulse-inversion-based tissue second-harmonic Imaging (PIHI). Contrast ratio (CR) and contrast-to-noise ratio (CNR) increased by 5.55 dB and 1.40 dB over PIHI. Tissue wavelet performs better than Mexh and Morl wavelet in lateral spatial resolution, CR, and CNR. The proposed technique effectively improves the imaging quality in lateral spatial resolution, CR, and CNR.

PurposeTo evaluate 18F-FDG myocardial metabolism imaging (MMI) using a total-body PET/CT scanner and explore the feasible scan duration to guide the clinical practice.MethodsA retrospective analysis was conducted on 41 patients who underwent myocardial perfusion-metabolism imaging to assess myocardial viability. The patients underwent 18F-FDG MMI with a total-body PET/CT scanner using a list-mode for 600 s. PET data were trimmed and reconstructed to simulate images of 600-s, 300-s, 120-s, 60-s, and 30-s acquisition time (G600-G30). Images among different groups were subjectively evaluated using a 5-point Likert scale. Semi-quantitative evaluation was performed using standardized uptake value (SUV), myocardial to background activity ratio (M/B), signal to noise ratio (SNR), contrast to noise ratio (CNR), contrast ratio (CR), and coefficient of variation (CV). Myocardial viability analysis included indexes of Mismatch and Scar. G600 served as the reference.ResultsSubjective visual evaluation indicated a decline in the scores of image quality with shortening scan duration. All the G600, G300, and G120 images were clinically acceptable (score ≥ 3), and their image quality scores were 4.9 ± 0.3, 4.8 ± 0.4, and 4.5 ± 0.8, respectively (P > 0.05). Moreover, as the scan duration reduced, the semi-quantitative parameters M/B, SNR, CNR, and CR decreased, while SUV and CV increased, and significant difference was observed in G300-G30 groups when comparing to G600 group (P < 0.05). For myocardial viability analysis of left ventricular and coronary segments, the Mismatch and Scar values of G300-G30 groups were almost identical to G600 group (ICC: 0.968-1.0, P < 0.001).ConclusionSufficient image quality for clinical diagnosis could be achieved at G120 for MMI using a total-body PET/CT scanner, while the image quality of G30 was acceptable for myocardial viability analysis.

Improving the ultrasound image contrast ratio (CR) and contrast to noise ratio (CNR) has many clinical advantages. Breast cancer detection is one example. Anechoic cysts which fill with clutter noise can be easily misinterpreted and classified as malignant lesions instead of benign. Beamforming techniques contribute to off-axis side lobes and clutter. These two side effects inherent in beamforming are undesirable since they will degrade the image quality by lowering the image CR and CNR. To overcome this issue a new post-processing technique known as contrast enhanced delay and sum (CEDAS) is proposed. Here the energy of every envelope signals are calculated, mapped, and clustered in order to identify the cyst and clutter location. CEDAS reduce clutter inside the cyst by attenuating it from envelope signals before the new B-Mode image is formed. With CEDAS, the image CR and CNR improved by average 12 dB and 1.1 dB respectively for cysts size 2 mm to 6 mm and imaging depth from 40 mm to 80 mm.

Synthetic transmit aperture (STA) has been widely investigated in ultrasound system recently with characteristics of high frame rate and low hardware cost. Since the high-resolution image (HRI) of STA is formed by summation of low-resolution images (LRIs), it is susceptible to inter-firing motions. In this paper, we propose a low-complexity global motion compensation algorithm. We use the common region of interest (ROIcom) between different transmissions of STA imaging to beamform backward and forward beam vectors. Then, the magnitude and direction of motion can be evaluated by cross-correlations between specific beam vectors in STA imaging. Compared with the uncompensated image in two-dimentional (2D) motion environment, the proposed motion compensation algorithm can improve the contrast ratio (CR) and contrast-to-noise ratio (CNR) by 13.73 and 2.04 dB, respectively. Also, the proposed algorithm improves the CR and CNR about 7.84 and 1.36 dB comparing with the reference work, respectively. In the Field II breath model, the proposed method also improves the CR and CNR about 6.65 and 1.04 dB than the reference method, respectively. Moreover, we propose a low-complexity delay generator in the architecture design to further reduce the computational complexity of the whole beamforming system. Finally, we verify the proposed low-complexity motion compensation beamforming engine by using the VLSI implementation with CMOS 90 nm technology. In the post-layout result, the core size is 2.39 mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> at 125 MHz operating frequency and the frame rate of the beamforming system is 42.23 frames per second.

BackgroundThis study investigates the integration of Artificial Intelligence (AI) in compensating the lack of time-of-flight (TOF) of the GE Omni Legend PET/CT, which utilizes BGO scintillation crystals.MethodsThe current study evaluates the image quality of the GE Omni Legend PET/CT using a NEMA IQ phantom. It investigates the impact on imaging performance of various deep learning precision levels (low, medium, high) across different data acquisition durations. Quantitative analysis was performed using metrics such as contrast recovery coefficient (CRC), background variability (BV), and contrast to noise Ratio (CNR). Additionally, patient images reconstructed with various deep learning precision levels are presented to illustrate the impact on image quality.ResultsThe deep learning approach significantly reduced background variability, particularly for the smallest region of interest. We observed improvements in background variability of 11.8%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document}, 17.2%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document}, and 14.3%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document} for low, medium, and high precision deep learning, respectively. The results also indicate a significant improvement in larger spheres when considering both background variability and contrast recovery coefficient. The high precision deep learning approach proved advantageous for short scans and exhibited potential in improving detectability of small lesions. The exemplary patient study shows that the noise was suppressed for all deep learning cases, but low precision deep learning also reduced the lesion contrast (about −30%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document}), while high precision deep learning increased the contrast (about 10%\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\%$$\\end{document}).ConclusionThis study conducted a thorough evaluation of deep learning algorithms in the GE Omni Legend PET/CT scanner, demonstrating that these methods enhance image quality, with notable improvements in CRC and CNR, thereby optimizing lesion detectability and offering opportunities to reduce image acquisition time.

Radiation dose and image quality estimation of three X-ray volume imaging (XVI) systems. A total of 126 patients were examined using three XVI systems (groups 1-3) and their data were retrospectively analysed from 2007 to 2012. Each group consisted of 42 patients and each patient was examined using cone-beam computed tomography (CBCT), digital subtraction angiography (DSA) and digital fluoroscopy (DF). Dose parameters such as dose-area product (DAP), skin entry dose (SED) and image quality parameters such as Hounsfield unit (HU), noise, signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were estimated and compared using appropriate statistical tests. Mean DAP and SED were lower in recent XVI than its previous counterparts in CBCT, DSA and DF. HU of all measured locations was non-significant between the groups except the hepatic artery. Noise showed significant difference among groups (P < 0.05). Regarding CNR and SNR, the recent XVI showed a higher and significant difference compared to its previous versions. Qualitatively, CBCT showed significance between versions unlike the DSA and DF which showed non-significance. A reduction of radiation dose was obtained for the recent-generation XVI system in CBCT, DSA and DF. Image noise was significantly lower; SNR and CNR were higher than in previous versions. The technological advancements and the reduction in the number of frames led to a significant dose reduction and improved image quality with the recent-generation XVI system. • X-ray volume imaging (XVI) systems are increasingly used for interventional radiological procedures. • More modern XVI systems use lower radiation doses compared with earlier counterparts. • Furthermore more modern XVI systems provide higher image quality. • Technological advances reduce radiation dose and improve image quality.

Funding Acknowledgements Type of funding sources: None. Background Dynamic myocardial CT perfusion (CTP) is a non-invasive imaging technique able to depict cardiac ischemia. It is expected that the application of a new dynamic perfusion image noise reducing filter, the similarity filter (4DSF) would improve the image quality of CTP, resulting in clear visualization of perfusion deficits and hypo-perfused tissue and allowing for a substantial reduction in the radiation dose in dynamic myocardial CTP. Purpose To evaluate image quality before and after the application of a novel noise reducing filter on dynamic myocardial CT perfusion images reconstructed with deep learning-based algorithm. Methods The image datasets of patients with stable chest pain at intermediate risk of cardiovascular disease, who underwent stress dynamic CTP scanning as clinically indicated were retrospectively collected. The images were acquired with a 320-slice CT system, reconstructed using a deep learning-based algorithm (8 mm slice thickness) and filtered with the 4DSF. For each case, the signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) in six regions-of-interest (ROIs, 33-38 mm2) were computed, before and after filtering. The signal ROIs were located midventricular, in both four-chamber and short-axis views in the left ventricle blood pool, and the background ROIs in the cardiac septum and lateral wall. Differences in the SNR and CNR values were compared using paired t-tests. Results As a part of an ongoing study, twenty five cases were analysed to date. The following results (means for AiCE and AiCE + 4DSF, and corresponding p-values) were obtained in short axis view: septum SNR: 3.46 vs 6.25 (p = 0.007), lateral wall SNR: 3.66 vs 6.52 (p = 0.000), septum CNR: 13.35 vs 23.85 (p = 0.005), lateral wall CNR: 13.16 vs 23.57 (p = 0.010); and in four-chamber view: septum SNR: 3.67 vs 7.44 (p = 0.003), lateral wall SNR: 3.97 vs 7.65 (p = 0.001) and septum CNR: 13.92 vs 29.97 (p = 0.016), lateral wall CNR: 13.63 vs 29.75 (p = 0.018). All results are statistically significant. Conclusions We confirmed that the similarity filter decreases noise and therefore improves the objective image quality in both short axis and four chamber views. Therefore, the improved tissue texture depiction and improved visualization of perfusion defects may facilitate detection of ischemia and lead to reduction in radiation dose necessary for dynamic CTP, maintaining image quality. Abstract Figure. Region-of-interests (ROIs) location

To evaluate a new iterative reconstruction (IR) algorithm for radiation dose, image quality (IQ), signal-to-noise-ratio (SNR), and contrast-to-noise-ratio (CNR) in multidetector computed tomography (MDCT) dynamic myocardial perfusion imaging (MPI). ECG-gated 256-slice MDCT dynamic MPI was performed in six pigs after subtotal balloon occlusion of one artery. Two 100 kVp protocols were compared: high dose (HD): 150 mAs; low dose (LD): 100 mAs. HD images were reconstructed with filtered back projection (FBP), LD images with FBP and different strengths of IR (L1, L4, and L7). IQ (5-point scale), SNR, and CNR (ischemic vs. normal myocardium) values derived from the HD (FBP) images and the different LD images were compared. Mean SNR values for myocardium were 16.3, 11.3, 13.1, 17.1, and 28.9 for the HD, LD (FBP), LD (L1), LD (L4), and LD (L7) reconstructions, respectively. Mean CNR values were 8.9, 6.3, 7.8, 9.3, and 12.8. IQ was scored as 4.6, 3.3, 4.4, 4.7, and 3.4, respectively. A significant loss of IQ was observed for the LD (L7) images compared to the HD (FBP) images (P < 0.05). Appropriate levels of iterative reconstruction can improve SNR and CNR, facilitating radiation dose savings in CT-MPI without influencing diagnostic quality. Iterative reconstruction (IR) can reduce radiation dose in myocardial perfusion CT. Our study also demonstrated improvements in image quality (noise, SNR, and CNR). Dynamic CT-MPI could help determine the hemodynamic significance of coronary artery disease. With dynamic CT MPI, myocardial blood flow can be determined quantitatively.

The purpose of this study was to compare the image quality of coronary computed tomography angiography (CTA) subjected to deep learning-based image restoration (DLR) method with images subjected to hybrid iterative reconstruction (IR). We enrolled 30 patients (22 men, 8 women) who underwent coronary CTA on a 320-slice CT scanner. The images were reconstructed with hybrid IR and with DLR. The image noise in the ascending aorta, left atrium, and septal wall of the ventricle was measured on all images and the contrast-to-noise ratio (CNR) in the proximal coronary arteries was calculated. We also generated CT attenuation profiles across the proximal coronary arteries and measured the width of the edge rise distance (ERD) and the edge rise slope (ERS). Two observers visually evaluated the overall image quality using a 4-point scale (1 = poor, 4 = excellent). On DLR images, the mean image noise was lower than that on hybrid IR images (18.5 ± 2.8 HU vs. 23.0 ± 4.6 HU, p < 0.01) and the CNR was significantly higher (p < 0.01). The mean ERD was significantly shorter on DLR than on hybrid IR images, whereas the mean ERS was steeper on DLR than on hybrid IR images. The mean image quality score for hybrid IR and DLR images was 2.96 and 3.58, respectively (p < 0.01). DLR reduces the image noise and improves the image quality at coronary CTA. • Deep learning-based image restoration is a new technique that employs the deep convolutional neural network for image quality improvement. • Deep learning-based restoration reduces the image noise and improves image quality at coronary CT angiography. • This method may allow for a reduction in radiation exposure.

Diffusion-weighted imaging (DWI) using single-shot echo-planar imaging (ss-EPI) is prone to artifacts, geometric distortion, and T2* blurring. Readout-segmented echo-planar imaging (rs-EPI) may improve image quality in the DWI of cervical cancer (CC). This study aimed to compare the image quality between rs-EPI and ss-EPI DWI in CC and to evaluate whether the apparent diffusion coefficient (ADC) values of ss-EPI (ssADC) and rs-EPI (rsADC) can differentiate the status of lymphovascular space invasion (LVSI) and lymph node metastasis (LNM). This prospective study included 69 patients with CC who underwent ss-EPI and rs-EPI DWI before surgery. Qualitative reader scores, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and ADC values derived from ss-EPI and rs-EPI were compared. The differences in ADC values were analyzed in patients who were (a) LNM-positive (LNM+, n = 17) and LNM-negative (LNM-, n = 52); (b) LVSI-positive (LVSI+, n = 33) and LVSI-negative (LVSI-, n = 36). The rs-EPIs of CC had higher subjective image quality scores and a lower SNR than ss-EPI (all P < 0.001); no significant differences existed between rs-EPI and ss-EPI for either CNR or ADC (CNR, P = 0.313; ADC, P = 0.949; P > 0.05 for all). The rsADC and ssADC of the LNM+ group were substantially lower than those of the LNM- group (rsADC, P = 0.000; ssADC, P = 0.000; P < 0.001 for all); the areas under the receiver operating characteristic curve were 0.855 and 0.851, respectively. However, there were no differences in ADC values between the LVSI+ and LVSI- groups (rsADC, P = 0.271; ssADC, P = 0.200; P > 0.05 for all). Over a similar scan time, rs-EPI improves the qualitative image quality of DWI significantly more than ss-EPI and has good diagnostic accuracy for LNM status in CC. However, neither could predict the LVSI status. Readout-segmented EPI improves the qualitative image quality of DWI and has good diagnostic accuracy for LNM status in CC, compared with conventional ss-EPI. It is more inclined to qualitative analysis of CC foci and provides a better scheme when choosing the DWI sequence scanning strategy for CC.