Zhou, Wei

Objectives Periodic quality control (QC) testing of ultrasound (US) imaging systems is essential to ensure and maintain image quality and safety. The study aims to analyze QC findings from medical physics annual surveys of modern clinical US systems in a multi‐institutional survey. Methods QC results from annual surveys between 2018 and 2021 were retrospectively collected from 12 medical physicists from 11 institutions or consulting companies (hereafter referred to as sites). QC tests were classified as scanner‐ and transducer‐related tests, where scanner‐related tests also included evaluation of diagnostic workstation display monitors and routine QC programs if applicable. Test methodology and pass/fail criteria were established by each site. QC findings were defined as results requiring follow‐up action by service engineers, vendors, or other personnel. The percentage of annual findings was calculated for each site and averaged across the years of data contributed. Furthermore, findings across all sites were aggregated for an overall analysis. Results QC data from a total of 1069 scanner tests and 4542 transducer tests were collected. The average annual percentage of scanner‐related findings varied from 0 to 40.8% among all sites (first quartile 8.7%, median 18.2%, and third quartile 34.1%), while that of transducer‐related findings ranged between 0.5% and 19.9% (first quartile 3.9%, median 7.2%, and third quartile 14.5%). For scanner‐related tests, the top 3 categories of findings were associated with physical and mechanical integrity (171 findings, 74.7%), quantitative testing of the scanner display monitor (40 findings, 17.5%), and scanner port creating artifacts (10 findings, 4.4%). The top 3 findings for transducers related to uniformity and artifact (227 findings, 79.9%), physical and mechanical integrity (44 findings, 15.5%), and sensitivity in fundamental mode (9 findings, 3.2%). Conclusions Medical physics annual surveys revealed considerable actionable findings of modern US systems, which inform the status of current US QC practices in the United States and may guide future standardization and recommendation of QC tests.

PurposeTo develop a convolutional neural network (CNN) that can directly estimate material density distribution from multi‐energy computed tomography (CT) images without performing conventional material decomposition.MethodsThe proposed CNN (denoted as Incept‐net) followed the general framework of encoder–decoder network, with an assumption that local image information was sufficient for modeling the nonlinear physical process of multi‐energy CT. Incept‐net was implemented with a customized loss function, including an in‐house‐designed image‐gradient‐correlation (IGC) regularizer to improve edge preservation. The network consisted of two types of customized multibranch modules exploiting multiscale feature representation to improve the robustness over local image noise and artifacts. Inserts with various densities of different materials [hydroxyapatite (HA), iodine, a blood–iodine mixture, and fat] were scanned using a research photon‐counting detector (PCD) CT with two energy thresholds and multiple radiation dose levels. The network was trained using phantom image patches only, and tested with different‐configurations of full field‐of‐view phantom and in vivo porcine images. Furthermore, the nominal mass densities of insert materials were used as the labels in CNN training, which potentially provided an implicit mass conservation constraint. The Incept‐net performance was evaluated in terms of image noise, detail preservation, and quantitative accuracy. Its performance was also compared to common material decomposition algorithms including least‐square‐based material decomposition (LS‐MD), total‐variation regularized material decomposition (TV‐MD), and U‐net‐based method.ResultsIncept‐net improved accuracy of the predicted mass density of basis materials compared with the U‐net, TV‐MD, and LS‐MD: the mean absolute error (MAE) of iodine was 0.66, 1.0, 1.33, and 1.57 mgI/cc for Incept‐net, U‐net, TV‐MD, and LS‐MD, respectively, across all iodine‐present inserts (2.0–24.0 mgI/cc). With the LS‐MD as the baseline, Incept‐net and U‐net achieved comparable noise reduction (both around 95%), both higher than TV‐MD (85%). The proposed IGC regularizer effectively helped both Incept‐net and U‐net to reduce image artifact. Incept‐net closely conserved the total mass densities (i.e., mass conservation constraint) in porcine images, which heuristically validated the quantitative accuracy of its outputs in anatomical background. In general, Incept‐net performance was less dependent on radiation dose levels than the two conventional methods; with approximately 40% less parameters, the Incept‐net achieved relatively improved performance than the comparator U‐net, indicating that performance gain by Incept‐net was not achieved by simply increasing network learning capacity.ConclusionIncept‐net demonstrated superior qualitative image appearance, quantitative accuracy, and lower noise than the conventional methods and less sensitive to dose change. Incept‐net generalized and performed well with unseen image structures and different material mass densities. This study provided preliminary evidence that the proposed CNN may be used to improve the material decomposition quality in multi‐energy CT.

AbstractPurposeUltrasound grayscale imaging preset optimization has often been qualitative and dependent upon vendor application specialists. This study aimed to propose a systematic approach for grayscale imaging preset optimization and apply the approach in a clinical abdominal scan setting.MethodsA six‐step approach was detailed including identification of clinical task, adjustment of basic parameters, fine‐tuning of advanced parameters, image performance metrics confirmation, clinical evaluation and data analysis, and implementation of new presets and monitoring of clinical usage. Its application in an abdominal scanning task was described for each step with phantoms, volunteers, and software tools.ResultsClinical image data analytics facilitated the understanding of the imaging task, relevant transducers, and target characteristics, in addition to specific requests from radiologists. Quantitative measurements were made on global image contrast and gray map function. In addition, clinically relevant phantoms and volunteer scans without and with acoustic distortion layers were involved to determine the new presets. Furthermore, phantom signal to noise ratio study and clinical evaluation using volunteers with different body habitus were utilized to confirm the superiority of the new presets. Quantitative clinical usage monitoring demonstrated successful implementation of the new presets.ConclusionsA systematic approach for grayscale imaging preset optimization has been proposed and successfully applied for a specific clinical task. This approach was designed to be generalizable and relatively flexible, which would facilitate movement away from previous qualitative and subjective approaches.

Objective The aim of this study was to quantitatively demonstrate radiation dose reduction for sinus and temporal bone examinations using high-resolution photon-counting detector (PCD) computed tomography (CT) with an additional tin (Sn) filter. Materials and Methods A multienergy CT phantom, an anthropomorphic head phantom, and a cadaver head were scanned on a research PCD-CT scanner using ultra-high-resolution mode at 100-kV tube potential with an additional tin filter (Sn-100 kV) and volume CT dose index of 10 mGy. They were also scanned on a commercial CT scanner with an energy-integrating detector (EID) following standard clinical protocols. Thirty patients referred to clinically indicated sinus examinations, and two patients referred to temporal bone examinations were scanned on the PCD-CT system after their clinical scans on an EID-CT. For the sinus cohort, PCD-CT scans were performed using Sn-100 kV at 4 dose levels at 10 mGy (n = 9), 8 mGy (n = 7), 7 mGy (n = 7), and 6 mGy (n = 7), and the clinical EID-CT was performed at 120 kV and 13.7 mGy (mean CT volume dose index). For the temporal bone scans, PCD-CT was performed using Sn-100 kV (10.1 mGy), and EID-CT was performed at 120 kV and routine clinical dose (52.6 and 66 mGy). For both PCD-CT and EID-CT, sinus images were reconstructed using H70 kernel at 0.75-mm slice thickness, and temporal bone images were reconstructed using a U70 kernel at 0.6-mm slice thickness. In addition, iterative reconstruction with a dedicated sharp kernel (V80) was used to obtain high-resolution PCD-CT images from a sinus patient scan to demonstrate improved anatomic delineation. Improvements in spatial resolution from the dedicated sharp kernel was quantified using modulation transfer function measured with a wire phantom. A neuroradiologist assessed the H70 sinus images for visualization of critical anatomical structures in low-dose PCD-CT images and routine-dose EID-CT images using a 5-point Likert scale (structural detection obscured and poor diagnostic confidence, score = 1; improved anatomic delineation and diagnostic confidence, score = 5). Image contrast and noise were measured in representative regions of interest and compared between PCD-CT and EID-CT, and the noise difference between the 2 acquisitions was used to estimate the dose reduction in the sinus and temporal bone patient cohorts. Results The multienergy phantom experiment showed a noise reduction of 26% in the Sn-100 kV PCD-CT image, corresponding to a total dose reduction of 56% compared with EID-CT (clinical dose) without compromising image contrast. The PCD-CT images from the head phantom and the cadaver scans demonstrated a dose reduction of 67% and 83%, for sinus and temporal bone examinations, respectively, compared with EID-CT. In the sinus cohort, PCD-CT demonstrated a mean dose reduction of 67%. The 10- and 8-mGy sinus patient images from PCD-CT were significantly superior to clinical EID-CT for visualization of critical sinus structures (median score = 5 ± 0.82 and P = 0.01 for lesser palatine foramina, median score = 4 ± 0.68 and P = 0.039 for nasomaxillary sutures, and median score = 4 ± 0.96 and P = 0.01 for anterior ethmoid artery canal). The 6- and 7-mGy sinus patient images did not show any significant difference between PCD-CT and EID-CT. In addition, V80 (sharp kernel, 10% modulation transfer function = 18.6 cm −1 ) PCD-CT images from a sinus patient scan increased the conspicuity of nasomaxillary sutures compared with the clinical EID-CT images. The temporal bone patient images demonstrated a dose reduction of up to 85% compared with clinical EID-CT images, whereas visualization of inner ear structures such as the incudomalleolar joint were similar between EID-CT and PCD-CT. Conclusions Phantom and cadaver studies demonstrated dose reduction using Sn-100 kV PCD-CT compared with current clinical EID-CT while maintaining the desired image contrast. Dose reduction was further validated in sinus and temporal bone patient studies. The ultra-high resolution capability from PCD-CT allowed improved anatomical delineation for sinus imaging compared with current clinical standard.

Abstract Objective The quality of all clinical MRI is dependent on B0 homogeneity, which is optimized during the shimming part of a prescan or preparatory phase before image acquisition. The purpose of this study was to assess shimming techniques clinically employed for breast MRI across our practice, and to determine factors that correlate with higher image quality for contrast-enhanced breast MRI at 1.5T. Methods One hundred consecutive female patients were retrospectively collected with Institutional Review Board approval. Shimming-related parameters, including shim-box placement and shimming gradient offsets were extracted from prior contrast-enhanced 3D fat-suppressed T1-weighted gradient echo image acquisitions. Three breast radiologists evaluated these images for fat saturation, breast density, overall image quality, and artifacts. Technologist experience was also evaluated for variability of shimming. Generalized linear mixed models were used to compare acquisition parameters between fat saturation. P < 0.05 was considered as statistical significance. Results The percentage of soft tissue inside the field of view (FOV) (ie, Tissue/FOV) in the good fat-saturation group (0.37 ± 0.06) was significantly lower (P < 0.01) than that in the poor fat-saturation group (0.39 ± 0.06). Other shimming-related parameters were found not significantly affecting the fat-saturation outcomes. Technologists with more experience tended to have less variable shimming performance than junior technologists did. Conclusions The quality of clinical MRI and especially breast MRI is highly dependent on shimming. Decreasing Tissue/FOV was associated with good image quality (good fat saturation). Optimization of shimming may require manual shimming or higher-order field-correction strategies.

AbstractPurposeIt is unclear if a 3D transducer with the special design of mechanical swing or 2D array could provide acceptable 2D grayscale image quality for the general diagnosis purpose. The aim of this study is to compare the 2D image quality of a 3D intracavitary transducer with a conventional 2D intracavitary transducer using clinically relevant phantom experiments.MethodsAll measurements were performed on a GE Logiq E9 scanner with both a 2D (IC5‐9‐D) and a 3D (RIC5‐9‐D) transducer used in 2D mode. Selection of phantom targets and acquisition parameters were determined from analysis of 33 clinical pelvic exams. Depth of penetration (DOP), contrast response, contrast of anechoic cylinders (diameter: 6.7 mm) at 1.5 and 4.5 cm depths in transverse planes, and in‐plane resolution represented by full‐width half‐maximum of pin targets at multiple depths were measured with transmit frequencies of 7 and 8 MHz. Spherical signal‐noise‐ratio (SNR) (diameter: 4 and 2 mm) at multiple depths were measured at 8 MHz.ResultsRIC5‐9‐D demonstrated <8% decrease in DOP for both transmit frequencies (7 MHz: 69.7 ± 8.2 mm; 8 MHz: 64.3 ± 7.8 mm) compared with those from IC5‐9‐D (7 MHz: 73.9 ± 4.4 mm; 8 MHz: 69.4 ± 7.8 mm). A decreased anechoic contrast was observed with a 4.5 cm depth for RIC5‐9‐D (7 MHz: 23.2 ± 1.8 dB, P > 0.05; 8 MHz: 17.7 ± 0.9 dB, P < 0.01) compared with IC5‐9‐D (7 MHz: 25.9 ± 1.2 dB; 8 MHz: 21.5 ± 0.8 dB). The contrast response and spatial resolution performance were comparable between the two transducers. RIC5‐9‐D showed comparable SNR of anechoic spheres compared to IC5‐9‐D.Conclusions2D images from a 3D probe exhibited comparable overall image quality for routine clinical pelvic imaging.

PurposeThis work aims to develop a new framework of image quality assessment using deep learning‐based model observer (DL‐MO) and to validate it in a low‐contrast lesion detection task that involves CT images with patient anatomical background.MethodsThe DL‐MO was developed using the transfer learning strategy to incorporate a pretrained deep convolutional neural network (CNN), a partial least square regression discriminant analysis (PLS‐DA) model and an internal noise component. The CNN was previously trained to achieve the state‐of‐the‐art classification accuracy over a natural image database. The earlier layers of the CNN were used as a deep feature extractor, with the assumption that similarity exists between the CNN and the human visual system. The PLSR model was used to further engineer the deep feature for the lesion detection task in CT images. The internal noise component was incorporated to model the inefficiency and variability of human observer (HO) performance, and to generate the ultimate DL‐MO test statistics. Seven abdominal CT exams were retrospectively collected from the same type of CT scanners. To compare DL‐MO with HO, 12 experimental conditions with varying lesion size, lesion contrast, radiation dose, and reconstruction types were generated, each condition with 154 trials. CT images of a real liver metastatic lesion were numerically modified to generate lesion models with four lesion sizes (5, 7, 9, and 11 mm) and three contrast levels (15, 20, and 25 HU). The lesions were inserted into patient liver images using a projection‐based method. A validated noise insertion tool was used to synthesize CT exams with 50% and 25% of routine radiation dose level. CT images were reconstructed using the weighted filtered back projection algorithm and an iterative reconstruction algorithm. Four medical physicists performed a two‐alternative forced choice (2AFC) detection task (with multislice scrolling viewing mode) on patient images across the 12 experimental conditions. DL‐MO was operated on the same datasets. Statistical analyses were performed to evaluate the correlation and agreement between DL‐MO and HO.ResultsA statistically significant positive correlation was observed between DL‐MO and HO for the 2AFC low‐contrast detection task that involves patient liver background. The corresponding Pearson product moment correlation coefficient was 0.986 [95% confidence interval (0.950, 0.996)]. Bland–Altman agreement analysis did not indicate statistically significant differences.ConclusionsThe proposed DL‐MO is highly correlated with HO in a low‐contrast detection task that involves realistic patient liver background. This study demonstrated the potential of the proposed DL‐MO to assess image quality directly based on patient images in realistic, clinically relevant CT tasks.

Objectives The aim of this study was to investigate the impact on metal artifacts and dose efficiency of using a tin filter in combination with high-energy threshold (TH) images of a photon-counting detector (PCD) computed tomography (CT) system. Materials and Methods A 3D-printed spine with pedicle screws was scanned on a PCD-CT system with and without tin filtration. Image noise and severity of artifacts were measured for low-energy threshold (TL) and TH images. In a prospective, institutional review board–approved, Health Insurance Portability and Accountability Act-compliant study, 20 patients having a clinical energy-integrating detector (EID) CT were scanned on a PCD-CT system using tin filtration. Images were reviewed by 3 radiologists to evaluate visualization of anatomic structures, diagnostic confidence, and image preference. Artifact severity and image noise were measured. Wilcoxon signed rank was used to test differences between PCD-CT TH and EID-CT images. Results Phantom TH images with tin filtration reduced metal artifacts and had comparable noise (32 HU) to TL images (29 HU) acquired without tin filtration. Visualization scores for the cortex, trabeculae, and implant-trabecular interface from PCD-CT TH images (4.4 ± 0.9, 4.4 ± 1.0, and 4.4 ± 1.0) were significantly higher (P < 0.0001) than EID-CT images (3.3 ± 1.3, 3.3 ± 1.2, and 3.3 ± 1.6). A strong preference was shown for PCD-CT TH images due to improved diagnostic confidence and decreased artifact severity. Noise in PCD-CT TH images (93 ± 41 HU) was significantly lower than that in EID-CT images (133 ± 92 HU, P < 0.05). Conclusions Threshold high images acquired with tin filtration on PCD-CT demonstrated a substantial decrease in metal artifacts and an increase in dose efficiency compared with EID-CT.

Objective The aims of this study were to quantitatively assess two new scan modes on a photon-counting detector computed tomography system, each designed to maximize spatial resolution, and to qualitatively demonstrate potential clinical impact using patient data. Materials and Methods This Health Insurance Portability Act–compliant study was approved by our institutional review board. Two high–spatial-resolution scan modes (Sharp and UHR) were evaluated using phantoms to quantify spatial resolution and image noise, and results were compared with the standard mode (Macro). Patients were scanned using a conventional energy-integrating detector scanner and the photon-counting detector scanner using the same radiation dose. In first patient images, anatomic details were qualitatively evaluated to demonstrate potential clinical impact. Results Sharp and UHR modes had a 69% and 87% improvement in in-plane spatial resolution, respectively, compared with Macro mode (10% modulation-translation-function values of 16.05, 17.69, and 9.48 lp/cm, respectively). The cutoff spatial frequency of the UHR mode (32.4 lp/cm) corresponded to a limiting spatial resolution of 150 μm. The full-width-at-half-maximum values of the section sensitivity profiles were 0.41, 0.44, and 0.67 mm for the thinnest image thickness for each mode (0.25, 0.25, and 0.5 mm, respectively). At the same in-plane spatial resolution, Sharp and UHR images had up to 15% lower noise than Macro images. Patient images acquired in Sharp mode demonstrated better delineation of fine anatomic structures compared with Macro mode images. Conclusions Phantom studies demonstrated superior resolution and noise properties for the Sharp and UHR modes relative to the standard Macro mode and patient images demonstrated the potential benefit of these scan modes for clinical practice.