Radiologists In Ultrasound Research Articles

Ensemble Deep Learning Model for Multicenter Classification of Thyroid Nodules on Ultrasound Images.

BackgroundThyroid nodules are extremely common and typically diagnosed with ultrasound whether benign or malignant. Imaging diagnosis assisted by Artificial Intelligence has attracted much attention in recent years. The aim of our study was to build an ensemble deep learning classification model to accurately differentiate benign and malignant thyroid nodules.Material/MethodsBased on current advanced methods of image segmentation and classification algorithms, we proposed an ensemble deep learning classification model for thyroid nodules (EDLC-TN) after precise localization. We compared diagnostic performance with four other state-of-the-art deep learning algorithms and three ultrasound radiologists according to ACR TI-RADS criteria. Finally, we demonstrated the general applicability of EDLC-TN for diagnosing thyroid cancer using ultrasound images from multi medical centers.ResultsThe method proposed in this paper has been trained and tested on a thyroid ultrasound image dataset containing 26 541 images and the accuracy of this method could reach 98.51%. EDLC-TN demonstrated the highest value for area under the curve, sensitivity, specificity, and accuracy among five state-of-the-art algorithms. Combining EDLC-TN with models and radiologists could improve diagnostic accuracy. EDLC-TN achieved excellent diagnostic performance when applied to ultrasound images from another independent hospital.ConclusionsBased on ensemble deep learning, the proposed approach in this paper is superior to other similar existing methods of thyroid classification, as well as ultrasound radiologists. Moreover, our network represents a generalized platform that potentially can be applied to medical images from multiple medical centers.

Evaluation of the Quadri-Planes Method in Computer-Aided Diagnosis of Breast Lesions by Ultrasonography: Prospective Single-Center Study.

BackgroundComputer-aided diagnosis (CAD) is a tool that can help radiologists diagnose breast lesions by ultrasonography. Previous studies have demonstrated that CAD can help reduce the incidence of missed diagnoses by radiologists. However, the optimal method to apply CAD to breast lesions using diagnostic planes has not been assessed.ObjectiveThe aim of this study was to compare the performance of radiologists with different levels of experience when using CAD with the quadri-planes method to detect breast tumors.MethodsFrom November 2018 to October 2019, we enrolled patients in the study who had a breast mass as their most prominent symptom. We assigned 2 ultrasound radiologists (with 1 and 5 years of experience, respectively) to read breast ultrasonography images without CAD and then to perform a second reading while applying CAD with the quadri-planes method. We then compared the diagnostic performance of the readers for the 2 readings (without and with CAD). The McNemar test for paired data was used for statistical analysis.ResultsA total of 331 patients were included in this study (mean age 43.88 years, range 17-70, SD 12.10), including 512 lesions (mean diameter 1.85 centimeters, SD 1.19; range 0.26-9.5); 200/512 (39.1%) were malignant, and 312/512 (60.9%) were benign. For CAD, the area under the receiver operating characteristic curve (AUC) improved significantly from 0.76 (95% CI 0.71-0.79) with the cross-planes method to 0.84 (95% CI 0.80-0.88; P<.001) with the quadri-planes method. For the novice reader, the AUC significantly improved from 0.73 (95% CI 0.69-0.78) for the without-CAD mode to 0.83 (95% CI 0.80-0.87; P<.001) for the combined-CAD mode with the quadri-planes method. For the experienced reader, the AUC improved from 0.85 (95% CI 0.81-0.88) to 0.87 (95% CI 0.84-0.91; P=.15). The kappa indicating consistency between the experienced reader and the novice reader for the combined-CAD mode was 0.63. For the novice reader, the sensitivity significantly improved from 60.0% for the without-CAD mode to 79.0% for the combined-CAD mode (P=.004). The specificity, negative predictive value, positive predictive value, and accuracy improved from 84.9% to 87.8% (P=.53), 76.8% to 86.7% (P=.07), 71.9% to 80.6% (P=.13), and 75.2% to 84.4% (P=.12), respectively. For the experienced reader, the sensitivity improved significantly from 76.0% for the without-CAD mode to 87.0% for the combined-CAD mode (P=.045). The NPV and accuracy moderately improved from 85.8% and 86.3% to 91.0% (P=.27) and 87.0% (P=.84), respectively. The specificity and positive predictive value decreased from 87.4% to 81.3% (P=.25) and from 87.2% to 93.0% (P=.16), respectively.ConclusionsS-Detect is a feasible diagnostic tool that can improve the sensitivity, accuracy, and AUC of the quadri-planes method for both novice and experienced readers while also improving the specificity for the novice reader. It demonstrates important application value in the clinical diagnosis of breast cancer.Trial RegistrationChiCTR.org.cn 1800019649; http://www.chictr.org.cn/showproj.aspx?proj=33094

Differential Diagnosis of Benign and Malignant Thyroid Nodules Using Deep Learning Radiomics of Thyroid Ultrasound Images

PurposeWe aimed to propose a highly automatic and objective model named deep learning Radiomics of thyroid (DLRT) for the differential diagnosis of benign and malignant thyroid nodules from ultrasound (US) images. MethodsWe retrospectively enrolled and finally include US images and fine-needle aspiration biopsies from 1734 patients with 1750 thyroid nodules. A basic convolutional neural network (CNN) model, a transfer learning (TL) model, and a newly designed model named deep learning Radiomics of thyroid (DLRT) were used for the investigation. Their diagnostic accuracy was further compared with human observers (one senior and one junior US radiologist). Moreover, the robustness of DLRT over different US instruments was also validated. Analysis of receiver operating characteristic (ROC) curves were performed to calculate optimal area under it (AUC) for benign and malignant nodules. One observer helped to delineate the nodules. ResultsAUCs of DLRT were 0.96 (95% confidence interval [CI]: 0.94-0.98), 0.95 (95% confidence interval [CI]: 0.93-0.97) and 0.97 (95% confidence interval [CI]: 0.95-0.99) in the training, internal and external validation cohort, respectively, which were significantly better than other deep learning models (P < 0.01) and human observers (P < 0.001). No significant difference was found when applying DLRT on thyroid US images acquired from different US instruments. ConclusionsDLRT shows the best overall performance comparing with other deep learning models and human observers. It holds great promise for improving the differential diagnosis of benign and malignant thyroid nodules.

P10 Rapid-access temporal artery ultrasound by vascular sonographers in suspected GCA: an audit during and after a training phase

Abstract Background Fast, same-day access ultrasound (US) of temporal (TA) and axillary (AxA) arteries has emerged as part of a gold standard of assessment for suspected giant cell arteritis (GCA). GCA is a rheumatological emergency but many rheumatology units lack expertise and capacity to provide such a service themselves. We audited US scanning for GCA by experienced vascular sonographers, but new to TA scanning, during and after a training phase. Methods Before the training audit a meeting between three vascular sonographers (LD, AO and MT), the radiology US Lead (KR) and an experienced rheumatologist sonographer (RK) took place to share the rationale, technique and evidence-base of TA/AXA scanning in suspected GCA. Early during the training period ad hoc co-scanning between technicians and RK/KR was performed and one technician attended the Vascular Scanning Unit at Nuffield Orthopaedic Centre in Oxford. Rheumatology medical staff was encouraged to refer all patients with suspected GCA to the Vascular US unit for scanning, with patients and clinicians being aware that the scan was for training purposes only. At the end of the training period, a sample review of images and reports was undertaken and diagnostic criteria and terminology rehearsed before service scanning commenced. Results Twenty-six subjects were scanned during the training period between July 2018 and May 2019: mean [range] 73 [53-89] years, 77% female. Of these 9 (35%) showed US changes reported as diagnostic of GCA in TAs, one equivocal and the remainder negative for GCA. Three of the 9 US-positive subjects underwent TA biopsy, which confirmed GCA in all cases. Of the 6 cases not undergoing biopsy, all were treated as GCA except one (3rd US training case, ESR 2). A total of 10 (38%) of all subjects underwent TA biopsy of which 7 showed a negative biopsy and negative TA US for GCA, thus resulting in a 100 % concordance of US and TA biopsy. In the first 4 months following the training phase, 23 subjects have been scanned for suspected GCA (mean [range] 69 [53-87] years, 65 % female): diagnostic, equivocal and negative changes for GCA were seen in 4, 2 and 17 subjects, respectively. No US-positive and 3 US-negative subjects underwent TA biopsy which showed negative histology for GCA in all three. Duration of steroid treatment before US and TA biopsy was a median 1 (IQR 0.75-4.0) and 6.5 (IQR 5.0-9.5) days, respectively, across both cohorts. Conclusion This audit suggests that experienced general vascular sonographers can achieve good proficiency in TA/AxA US for suspected GCA with excellent concordance with TA biopsy results to support a rapid-access clinical pathway for patients with suspected GCA. Furthermore, the audit suggests a reduction in biopsy rates in subjects whose US is positive for GCA. Disclosures L. Devonshire None. C. Koutsianas None. A. Ostrowski None. M. Trumper None. K. Randhawa None. R. Klocke None.

FEASIBILITY AND RELIABILITY OF MUSCULOSKELETAL ULTRASOUND MEASUREMENT OF THE MEDIAL PATELLOFEMORAL LIGAMENT

BackgroundThe medial patellofemoral ligament (MPFL) is an important soft tissue constraint to preventing patellar dislocations in young athletes. The anatomy of the MPFL has been investigated in cadaveric studies and magnetic resonance studies. No studies to date have provided anatomical data of the MPFL on ultrasonography.PurposeTo investigate the feasibility of musculoskeletal ultrasonography for the evaluation of the MPFL, and to determine interrater and intrarater reliability for MPFL ultrasound measures.MethodsTen control participants (20 knees) 20 to 50 years underwent ultrasonography performed by 3 researchers (musculoskeletal ultrasound radiologist, athletic trainer/biomechanist, primary care sports medicine physician) from 3 different institutions for interrater reliability testing. Intrarater reliability testing was performed at 2 separate institutions by 4 physicians, each performing the same knee ultrasound protocol on 20 knees in 10 study participants 2 to 3 weeks apart. In total, 180 images were created for interrater reliability, and 480 images for intrarater reliability. Examinations were performed with linear high-frequency transducers (10-18 MHz) with the participant in the supine position and the extremity flexed at 45°. Measurements included ligament length (long axis to ligament) from the patellar to the femoral attachment sites, ligament width (short axis to ligament) at the patellar attachment, and ligament thickness (long axis to ligament) midway between the patella and femur. Mean and SD were calculated for all measurements. Intraclass correlation coefficient (ICC) analysis was used to assess intrarater and interrater reliability. ICC values < 0.40 indicated poor reliability, whereas those between 0.40 and 0.75 indicated fair to good reliability, and those > 0.75 indicated excellent reliability.ResultsThe mean US value for MPFL length was 44.83mm (SD 6.68), mean thickness 2.66mm (SD 0.85), and mean width 11.76mm (SD 2.99). The overall ICC values for interrater reliability testing indicated fair to good reliability for length measures (0.7) and poor reliability for thickness (–0.1) and width (0.3; Table 1.1). Overall ICC values for intrarater reliability indicated fair to good reliability for length (0.5), excellent for thickness (0.9), and poor reliability for width (–0.3; Table 1.2).ConclusionsMusculoskeletal ultrasonography is a feasible and reliable office-based method of measuring MPFL length and thickness. These quantitative measures set the groundwork for establishing normative anatomical measures of the MPFL in athletes and establish a protocol for testing and measuring the MPFL using musculoskeletal ultrasonography.Table 1.1.Intraclass Correlation Coefficients (ICC) for testing reliability between physicians (inter-rater). ICC Length Thickness Width Overall0.708(0.291, 0.898)-0.047(-0.452, 0.442)0.253(-0.179, 0.652)xx vs xx0.833(0.562, 0.937)-0.055(-0.312, 0.309)0.062(-0.389, 0.498)xx vs xx0.297(-1.246, 0.776)-0.014(-0.385, 0.447)0.253(-0.130, 0.672)xx vs xx0.707(0.053, 0.907)0.032(-2.554, 0.704)0.286(-0.874, 0.756)*Data analyzed using two-way mixed, absolute agreement analysis. Expected value is shown, along with lower and upper bounds of 95% confidence intervals.Table 1.2.Intraclass Correlation Coefficients (ICC) for test-retest (intra-rater) reliability among individual physicians. Length Thickness Width Overall 0.492*(0.190, 0.681) 0.858(0.772, 0.912) -0.319(-0.056, 0.566)Physician 10.589(-0.024, 0.842)0.400(-0.300, 0.754)0.386(-0.241, 0.756)Physician 20.014(-0.393, 0.447)0.260(-0.965, 0.722)-0.097(-0.683, 0.432)Physician 3-0.085(-2.104, 0.604)0.463(-0.228, 0.786)-1.910(-7.683, -0.027)Physician 40.792(0.467, 0.918)0.634(0.053, 0.857)-0.224(-1.272, 0.435)*Data analyzed using two-way mixed, absolute agreement analysis. Expected value is shown, along with lower and upper bounds of 95% confidence intervals.

4:21 PM Abstract No. 136 Use of contrast-enhanced ultrasound after computed tomography–guided microwave ablation for hepatocellular carcinoma in our practice with magnetic resonance imaging and pathologic correlation

Contrast-enhanced ultrasound (CEUS) is a low cost and effective imaging modality that is accessible immediately post procedure which allows the clinician to instantly perform additional ablation of the tumor if a residual tumor is detected. CEUS can be an alternative to contrast enhanced CT in patients with underlying renal disease or iodine allergy. The purpose of this abstract is to share our experience with use of contrast enhanced ultrasound after microwave ablation for HCC with MRI correlation and posttransplant pathology. This is a retrospective analysis of 13 patients in an IRB-approved study conducted at our institution from 2017-2019. The patients were identified during liver tumor board with diagnosis of hepatocellular carcinoma (HCC) amenable to microwave ablation. These patients underwent CT-guided microwave ablation of HCC. Immediately post procedure patients underwent CEUS to evaluate for residual tumor, which was reported by an expert ultrasound radiologist providing us with instant feedback regarding residual tumor. Patients underwent MRI one month post procedure to evaluate for residual tumor within the treatment bed. Pathology reports were obtained for the patients that received liver transplants and radiologic and pathologic correlation was made to the treatment site. 7 postablation CEUS examinations demonstrated complete response to therapy with concordant results on follow-up MRI. 3 went on to have a liver transplant with complete or near complete response. Four post ablation CEUS examinations were read as suboptimal. Follow-up MRI for 3 showed complete response and 1 showed viable tumor in the treatment cavity. 1 patient had transplant before MRI. Transplant pathology showed complete response. Out of 13 patients 1 study was discordant with no enhancement seen on CEUS at the time of ablation with viable tumor on MRI. See Table 1. Preliminary data demonstrates congruence between post ablation CEUS and follow-up MRI or CEUS with no residual tumor in the treatment bed in 7 of 13 of our patients. Artifact and operator dependence can limit evaluation however CEUS is still an inexpensive and useful tool post ablation.Table 1Summary of Patients with Postablation CEUS and MRI13 Patients7 patients: concordant CEUS and MRI complete response3 patients transplanted2 pathology reports with complete response; 1 with partial response4 patients: postablation CEUS suboptimal because of artifact3 compete response; 1 viable tumor1 transplant with complete response1 patient: CEUS no MRI because of transplantPathology report shows complete response1 patient: discordant CEUS and MRI Open table in a new tab

Diagnosis of Thyroid Nodules: Performance of a Deep Learning Convolutional Neural Network Model vs. Radiologists

Computer-aided diagnosis (CAD) systems hold potential to improve the diagnostic accuracy of thyroid ultrasound (US). We aimed to develop a deep learning-based US CAD system (dCAD) for the diagnosis of thyroid nodules and compare its performance with those of a support vector machine (SVM)-based US CAD system (sCAD) and radiologists. dCAD was developed by using US images of 4919 thyroid nodules from three institutions. Its diagnostic performance was prospectively evaluated between June 2016 and February 2017 in 286 nodules, and was compared with those of sCAD and radiologists, using logistic regression with the generalized estimating equation. Subgroup analyses were performed according to experience level and separately for small thyroid nodules 1–2 cm. There was no difference in overall sensitivity, specificity, positive predictive value (PPV), negative predictive value and accuracy (all p > 0.05) between radiologists and dCAD. Radiologists and dCAD showed higher specificity, PPV, and accuracy than sCAD (all p < 0.001). In small nodules, experienced radiologists showed higher specificity, PPV and accuracy than sCAD (all p < 0.05). In conclusion, dCAD showed overall comparable diagnostic performance with radiologists and assessed thyroid nodules more effectively than sCAD, without loss of sensitivity.

348 Training and Competency of Emergency Department Advanced Practice Providers in First Trimester Pregnancy Ultrasound

Assess advanced practice provider (APP) competency in ultrasound (US) for symptomatic first trimester pregnancy after a targeted training protocol. Describe confidence of APPs in US image acquisition and interpretation. As part of a quality assurance (QA) initiative, APPs were trained in point-of-care (POC) first trimester pregnancy transabdominal ultrasound (TAU) via a didactic and supervised examinations. Prior to training, APPs did not utilize POCUS. This is a retrospective chart review of the 6-month period encompassing the training period and subsequent US. All TAU was performed on stable patients in an academic urban tertiary care center emergency department. APPs utilized a brief, systematic acquisition protocol. If there was an intrauterine pregnancy (IUP), large free fluid, or large adnexal mass, a credentialed emergency physician over-read the US for safety. If no IUP was visualized, a radiology ultrasound (RUS) was ordered. A 10-question QA form including interpretations and confidence was completed for all US. Confidence was based on a 5-point Likert scale with the anchors of “not at all confident” (1) to “completely confident” (5). Of the 12 APPs who completed the didactic, 8 (67%) performed TAU. A total of 52 exams were performed identifying 36 (70%) IUPs. Fetal cardiac activity was noted in 94% (34), of which M-mode was obtained in 78% (28) versus visual inspection in 22% (5). Crown rump length (CRL) measurements were obtained in 67% (24) of IUPs. Of those with a calculated CRL, the average gestation was 8 weeks 6 days (SD 15 days). There were no missed ectopic pregnancies. For patients with a TAU confirmed IUP, confidence in identifying IUP, determining CRL, and scanning the adnexa was rated fairly to completely confident (4 or 5) in 86.1%, 79.2%, and 66.7% of exams respectively. There were 4 RUS for patients with an APP-identified IUP. The reasons for RUS were indeterminate if free fluid in the abdomen; no fetal cardiac activity; abnormal appearing cervix; and abnormal tissue in the gestational sac. RUS noted multiple uterine fibroids and small subchorionic hemorrhage; early embryonic demise; Nabothian cysts and cervical calcifications; and choriodecidual hematoma or chorionic bump, respectively. Of the 16 (30%) TAU that did not identify an IUP, 13 of the RUS found an IUP with average gestation of 6 weeks 4 days (SD 15 days). All RUS utilized transvaginal US. In 7 of the 13 RUS confirmed IUP, only a yolk sac and no fetal pole was identified. For TAU with no identified IUP, confidence in scanning the adnexa was 50%. Providing APPs with training and a protocol increased their POCUS utilization. In patients with an IUP, APPs were most confident in the assessment of the location of the pregnancy and least confident in scanning the adnexa. The average gestational age for radiology determined IUP was 2 weeks younger than APP determined IUP. In this small sample, APPs competently identified IUP or appropriately referred to a credentialed emergency physician or RUS. There were no missed ectopic pregnancies. The study of APP-performed first trimester pregnancy US is ongoing.

Abstract 1394: Diagnosis of thyroid cancer using deep convolutional neural network models applied to sonographic images from clinical ultrasound exams

Abstract Background Incidence rate of thyroid cancer is steadily increasing due to overdiagnosis and overtreatment. Thyroid ultrasound is commonly used to diagnose thyroid cancer. The aim of this study is to examine the accuracy of using deep convolutional neural network (DCNN) models to improve diagnosis of thyroid cancer by analyzing sonographic imaging data from clinical thyroid ultrasound. Methods A total of 131,731 sonographic images from 17,627 thyroid cancer patients and 180,668 sonographic images from 25,325 controls used as training set were obtained from Tianjin Cancer Hospital. Images from anatomical sites that did not have cancer according to location sign on the image were not included. All thyroid cancer patients and 13·2% of controls (51,255 images) were confirmed by pathological reports. DCNN is a specific type of neural network optimized for image recognition. We trained two DCNN models on the training set and subsequently evaluated the performance on one independent internal (Tianjin, 1,118 individuals) and two external (Jilin,154 individuals; Weihai, 1,420 individuals) validation sets. Individuals in the validation sets all have pathological examinations. We compared the specificity/sensitivity of DCNN models with the performance of six thyroid ultrasound radiologists on these three validation sets. Findings DCNN model achieved high performance in identifying thyroid cancer patients versus six experience radiologists: for Tianjin validation set, sensitivity was 92·2% versus 96·9% (95% CI 89·7% - 94·3% vs. 93·9% - 98·6%; p = 0·003), and specificity was 85·6% versus 59·4% (95% CI 82·4% - 88·4% vs. 53% - 65·6%; p &amp;lt; 0·0001); for Jilin validation set, sensitivity was 84·3% versus 92·9% (95% CI 73·6% - 91·9% vs. 84·1% - 97·6%; p = 0·05), and specificity was 86·9% versus 57·1% (95% CI 77·8% - 93·3% vs. 45·9% - 67·9%; p &amp;lt; 0·0001); for Weihai validation set, sensitivity was 84·5% versus 89% (95% CI 81·2% - 87·4% vs. 81·9% - 94%; p = 0·2), and specificity was 87·5% versus 68·6% (95% CI 85·1% - 89·6% vs. 60·7% - 75·8%; p &amp;lt; 0·0001). Interpretation DCNN models exhibited high accuracy, sensitivity, and specificity in identifying thyroid cancer patients at levels comparable to or higher than six experienced radiologists. Conferred by the high specificity of DCNN models, the rate of overdiagnosis and overtreatment of patients with thyroid cancer is expected to decrease. This supports future application of the deep learning models to clinical practice for thyroid cancer diagnosis. However, further validation of these DCNN models in prospective clinical trials is warranted. Funding The Program for Changjiang Scholars and Innovative Research Team in University in China (IRT_14R40), National Natural Science Foundation of China (31801117). Citation Format: Xiangchun Li, Sheng Zhang, Qiang Zhang, Xi Wei, Yi Pan, Jing Zhao, Xiaojie Xin, Xiaoqing Wang, Fan Yang, Jianxin Li, Meng Yang, Qinghua Wang, Xiangqian Zheng, Yanhui Zhao, Lun Zhang, Xudong Wang, Zhimin Zheng, Christopher T. Whitlow, Metin N. Gurcan, Boris Pasche, Ming Gao, Wei Zhang, Kexin Chen. Diagnosis of thyroid cancer using deep convolutional neural network models applied to sonographic images from clinical ultrasound exams [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1394.

Systematic radiologic detection of kidney stones in Canadian children: a new era of asymptomatic stones?

Asymptomatic, or incidental, pediatric kidney stones detected on abdominal imaging pose a clinical challenge as their significance and expected outcome are not well described. Our primary objective was to estimate the incidence of nephrolithiasis in pediatric patients undergoing abdominal ultrasound (US) or computerized tomography (CT), for all indications, in a pediatric tertiary care hospital. Our secondary objective was to determine the clinical outcome of these radiographically detected stones. All abdominal US or CT radiology reports for patients younger than 18 years between January 1, 2011, and December 31, 2016, were retrieved. Reports were automatically flagged using textual analysis if they contained one of the 32 keywords potentially indicating nephrolithiasis. Flagged reports, as well as 10% of unflagged reports, were reviewed to confirm the radiologistimpression of presence or absence of stones. Patient and stone-related clinical data were extracted. Two thousand four hundred forty-nine (5%) of 53235 imaging reports cited at least one of the keywords. Manual review of flagged reports identified 498 studies having a radiologist impression of stones (244 unique patients). Stone incidence in children undergoing abdominal imaging more than doubled between 2011 and 2016. Medical record review found that 140 patients (57%) were symptomatic, while the other 104 patients (43%) were asymptomatic. Spontaneous resolution was greater (57% vs 34%) in asymptomatic patients compared to symptomatic patients. Asymptomatic patients were younger with a median age of 0.6 compared to 12.3 years. Asymptomatic and symptomatic stones were followed up for similar lengths of time (2.3 vs 2.0 years, p>0.05) and had a similar number of follow-up USs (3.9 vs 4.6 studies, p>0.05). The incidence of radiologically identified stones in children undergoing abdominal imaging at our center increased over time. Asymptomatic stones follow a benign course with minimal need for intervention; however, they were detected almost as frequently as symptomatic stones. Asymptomatic stones may represent increased healthcare resource utilization due to similar follow-up compared to symptomatic stones. Further research regarding the optimal imaging intervals and long-term follow-up for asymptomatic stones is warranted.

Diagnosis of thyroid cancer using deep convolutional neural network models applied to sonographic images: a retrospective, multicohort, diagnostic study.

The incidence of thyroid cancer is rising steadily because of overdiagnosis and overtreatment conferred by widespread use of sensitive imaging techniques for screening. This overall incidence growth is especially driven by increased diagnosis of indolent and well-differentiated papillary subtype and early-stage thyroid cancer, whereas the incidence of advanced-stage thyroid cancer has increased marginally. Thyroid ultrasound is frequently used to diagnose thyroid cancer. The aim of this study was to use deep convolutional neural network (DCNN) models to improve the diagnostic accuracy of thyroid cancer by analysing sonographic imaging data from clinical ultrasounds. We did a retrospective, multicohort, diagnostic study using ultrasound images sets from three hospitals in China. We developed and trained the DCNN model on the training set, 131 731 ultrasound images from 17 627 patients with thyroid cancer and 180 668 images from 25 325 controls from the thyroid imaging database at Tianjin Cancer Hospital. Clinical diagnosis of the training set was made by 16 radiologists from Tianjin Cancer Hospital. Images from anatomical sites that were judged as not having cancer were excluded from the training set and only individuals with suspected thyroid cancer underwent pathological examination to confirm diagnosis. The model's diagnostic performance was validated in an internal validation set from Tianjin Cancer Hospital (8606 images from 1118 patients) and two external datasets in China (the Integrated Traditional Chinese and Western Medicine Hospital, Jilin, 741 images from 154 patients; and the Weihai Municipal Hospital, Shandong, 11 039 images from 1420 patients). All individuals with suspected thyroid cancer after clinical examination in the validation sets had pathological examination. We also compared the specificity and sensitivity of the DCNN model with the performance of six skilled thyroid ultrasound radiologists on the three validation sets. Between Jan 1, 2012, and March 28, 2018, ultrasound images for the four study cohorts were obtained. The model achieved high performance in identifying thyroid cancer patients in the validation sets tested, with area under the curve values of 0·947 (95% CI 0·935-0·959) for the Tianjin internal validation set, 0·912 (95% CI 0·865-0·958) for the Jilin external validation set, and 0·908 (95% CI 0·891-0·925) for the Weihai external validation set. The DCNN model also showed improved performance in identifying thyroid cancer patients versus skilled radiologists. For the Tianjin internal validation set, sensitivity was 93·4% (95% CI 89·6-96·1) versus 96·9% (93·9-98·6; p=0·003) and specificity was 86·1% (81·1-90·2) versus 59·4% (53·0-65·6; p<0·0001). For the Jilin external validation set, sensitivity was 84·3% (95% CI 73·6-91·9) versus 92·9% (84·1-97·6; p=0·048) and specificity was 86·9% (95% CI 77·8-93·3) versus 57·1% (45·9-67·9; p<0·0001). For the Weihai external validation set, sensitivity was 84·7% (95% CI 77·0-90·7) versus 89·0% (81·9-94·0; p=0·25) and specificity was 87·8% (95% CI 81·6-92·5) versus 68·6% (60·7-75·8; p<0·0001). The DCNN model showed similar sensitivity and improved specificity in identifying patients with thyroid cancer compared with a group of skilled radiologists. The improved technical performance of the DCNN model warrants further investigation as part of randomised clinical trials. The Program for Changjiang Scholars and Innovative Research Team in University in China, and National Natural Science Foundation of China.

Elderly persons with acute pancreatitis - specifics of the clinical course of the disease.

BackgroundAs may readily be inferred from clinical observations, there is an increasing prevalence of severe acute pancreatitis (SAP) in the elderly, even though the precise nature of the relationship between a patient’s age and severity of the disease has not been unequivocally determined to date. This study aimed therefore to investigate the incidence, clinical course, and mortality rate among the elderly patients suffering from acute pancreatitis (AP), as compared to younger ones.MethodsA prospective study, lasting a single calendar year, covered patients with AP successively admitted to hospitals. The final assessment comprised 963 patients. The patients were subsequently divided into three groups corresponding to the three grades of disease severity, based on the Revised Atlanta Classification for Acute Pancreatitis. The actual cause of the disease, its clinical course, results of radiological ultrasonography, computed tomography, and laboratory tests, as well as the duration of hospital stay were assessed in due consideration of patients’ age (groups aged 65–79 years and ≥80 years vs <65 years).ResultsCholelithiasis was determined as the main cause of AP among the older patients (54.08% and 58.12% vs 22.46%; P<0.000). Among the oldest patients (≥80 years), its course was often significantly more severe, in comparison with the ones under 65 years of age: 14.53% vs 6.31% (P<0.00); a phenomenon not observed in the age range 65–79 years, nor among the younger patients (7.69% vs 6.31%; P>0.05). Moderate AP occurred significantly more often in the younger patients compared to those aged ≥80 years (16% vs 8.55%; P<0.00), although without any significant differences observed between the group aged 65–79 years and the younger patients (13.27% vs 8.55%). SAP more frequently ended in death among the oldest patients – 11.97% vs 2.31% (P<0.000) than in the group aged 65–79 years (4.59%), as compared to the younger groups (P>0.05).ConclusionAn appreciably higher susceptibility of older patients aged ≥80 years to AP, with cholelithiasis being the main cause that results in high mortality rate, is presently acknowledged a serious diagnostic and therapeutic management challenge to a national healthcare system.

Evaluation of Ventricle Size Measurements in Infants by Pediatric Emergency Medicine Physicians.

The identification of hydrocephalus in infants by pediatric emergency medicine (PEM) physicians using cranial point-of-care ultrasound (POCUS) has not been evaluated. We aimed to conduct a pilot/proof-of-concept study to evaluate whether PEM physicians can identify hydrocephalus (anterior horn width >5 mm) in 15 infants (mean 69 ± 42 days old) from the neonatal intensive care unit using POCUS. Our exploratory aims were to determine the test characteristics of cranial POCUS performed by PEM physicians for diagnosing hydrocephalus and the interrater reliability between measurements made by the PEM physicians and the radiologist. Depending on the availability, 1 or 2 PEM physicians performed a cranial POCUS through the open anterior fontanel for each infant after a 30-minute didactic lecture to determine the size of the left and right ventricles by measuring the anterior horn width at the foramen of Monroe in coronal view. Within 1 week, an ultrasound (US) technologist performed a cranial US and a radiologist determined the ventricle sizes from the US images; these measurements were the criterion standard. A radiologist determined 12 of the 30 ventricles as hydrocephalic. The sensitivity and specificity of the PEM physicians performed cranial POCUS was 66.7% (95% confidence interval [CI], 34.9%-90.1%) and 94.4% (95% CI, 72.7%-99.9%), whereas the positive and negative predictive values were 88.9% (95% CI, 53.3%-98.2%) and 81.0% (95% CI, 65.5%-90.5%), respectively. The interrater reliability between the PEM physician's and radiologist's measurements was r = 0.91. The entire POCUS examinations performed by the PEM physicians took an average of 1.5 minutes. The time between the cranial POCUS and the radiology US was, on average, 4 days. While the PEM physicians in our study were able to determine the absence of hydrocephalus in infants with high specificity using cranial POCUS, there was insufficient evidence to support the use of this modality for identifying hydrocephalus. Future studies with more participants are warranted to accurately determine test characteristics.

Gallbladder Scalloping, Mammillated Caudate Lobe, and Inferior Vena Cava Scalloping: Three Novel Ultrasound Signs of Cirrhosis

We aimed to present three new ultrasound signs-gallbladder scalloping, mammillated caudate lobe, and inferior vena cava scalloping-and determine their accuracy in diagnosing liver cirrhosis. A total of 201 consecutive patients with a history of chronic liver disease who had undergone ultrasound imaging and liver biopsy were identified. A senior ultrasound radiologist blindly reviewed the ultrasound examinations. Specificity, sensitivity, positive predictive value, and negative predictive value of diagnosing cirrhosis were calculated for all evaluated ultrasound signs and selected combinations of signs, using the liver biopsy results as the reference standard. Of the 201 patients, 152 (76%) had either pathology-proven cirrhosis or significant fibrosis. Caudate lobe hypertrophy was the most specific (88%) and most positive predictor (90%) for cirrhosis, whereas mammillated caudate lobe was the most sensitive (78%). Inferior vena cava scalloping was the most specific (78%) of the three proposed ultrasound signs. When signs were combined, the presence of either gallbladder scalloping or liver surface nodularity was highly sensitive for cirrhosis (87%), whereas the presence of either gallbladder scalloping or inferior vena cava scalloping with caudate lobe hypertrophy was highly specific (93%). Gallbladder scalloping, mammillated caudate lobe, and inferior vena cava scalloping are three novel signs that improve the accuracy of ultrasound in diagnosing cirrhosis.

Effect of Emergency Physician-Performed Point-of-Care Ultrasound and Radiology Department-Performed Ultrasound Examinations on the Emergency Department Length of Stay Among Pregnant Women at Less Than 20 Weeks' Gestation.

We sought to confirm retrospective studies that measured an approximately 20% reduction in emergency department (ED) length of stay (LOS) in early-gestation pregnant women who receive emergency physician-performed point-of-care ultrasound (US) examinations rather than radiology department-performed US examinations for evaluation of intrauterine pregnancy (IUP). A randomized controlled clinical trial was performed at an urban academic safety net hospital and 2 Naval medical centers in the United States. The allocation was concealed before enrollment. Clinically stable adult pregnant women at less than 20 weeks' gestation who presented to the ED with abdominal pain or vaginal bleeding were randomized to receive a point-of-care or radiology US to assess for IUP. The primary outcome measure was the ED LOS. A total of 224 patients (point-of-care US, n = 118; radiology US, n = 106) were included for the analysis. The ED LOS was 20 minutes shorter in the point-of-care US arm (95% confidence interval [CI], -54 to 7 minutes). Adjusting for variability due to the location, the ED LOS was calculated to be 31 minutes shorter (95% CI, -64 to 1 minute) than for patients in the radiology US arm. Excluding patients in the point-of-care US arm who crossed over to radiology US after an inconclusive point-of-care US examination, the ED LOS was 75 minutes shorter than in the radiology US arm (95% CI, -97 to -53 minutes). Early-gestation pregnant ED patients requiring pelvic US were discharged earlier when point-of-care US was used rather than radiology US; however, this trial did not achieve our target of 30 minutes. Nevertheless, our data support the routine use of ED point-of-care US for IUP, saving the most time if a conclusive IUP is identified.

Sonographically Occult Abscesses of the Buttock and Perineum in Children.

Ultrasound (US) is used to differentiate abscess from cellulitis. At our institution, we observed children who had purulent fluid obtained after a negative abscess US. We sought to determine the incidence of sonographically occult abscess (SOA) of the buttock and perineum, and identify associated clinical and demographic characteristics. Retrospective chart review including children younger than 18 years old presenting to pediatric emergency department with soft tissue infection of the buttock or perineum and diagnostic radiology US read as negative for abscess. We defined SOA as wound culture growing pathogenic organism obtained within 48 hours of the US. Clinical and demographic characteristics included age, sex, race, ethnicity, fever, history of spontaneous drainage, duration of symptoms, previous methicillin resistant Staphylococcus aureus (MRSA) infection, or previous abscess. We used univariate and multivariate logistic regression to assess correlation between these characteristics and SOA. A total of 217 children were included. Sixty-one (28%) children had SOA; 33 of 61 (54%) had incision and drainage within 4 hours of the US. Of children with SOA, 49 (80%) grew MRSA and 12 (20%) grew methicillin-sensitive S. aureus. In univariate analysis, a history of MRSA, symptom duration 4 days or less, age of younger than 4 years, and Hispanic ethnicity increased the odds of having SOA. In multivariate analysis, history of MRSA and duration of 4 days or less were associated with SOA. Twenty-eight percent of children in our institution with US of the buttock and perineum negative for abscess had clinical abscess within 48 hours, most within 4 hours. History of MRSA and shorter symptom duration increased the odds of SOA.

Hip ultrasonography in clinical practice.

The current practice of hip ultrasonography for developmental dysplasia of the hip (DDH) is based on more than three decades of experience. During this time, professional medical organizations in the United States have addressed both the clinical practice and imaging aspects of early detection of DDH. Current evidence-based clinical practice recommendations come from the American Academy of Pediatrics and the American Academy of Orthopaedic Surgeons. Consensus imaging recommendations come from the American Institute of Ultrasound in Medicine, the American College of Radiology, Society for Pediatric Radiology and the Society of Radiologists in Ultrasound. In this review the author compares and contrasts the current guidelines and offers additional practice tips for those providing comprehensive hip ultrasound services to infants.

Searching the perfect ultrasonic classification in assessing carotid artery stenosis: comparison and remarks upon the existing ultrasound criteria.

Doppler ultrasound scanning is the first line investigation for quantifying the internal carotid artery stenosis. Nevertheless, the lack of internationally accepted ultrasound criteria for describing the degree of stenosis has contributed to the different and confusing measurements ranges. The use of two different angiographic methods, the North American Symptomatic Carotid Endoarterectomy Study and the European Carotid Surgery Trial was probably the major initial source of confusion in deriving valid and reliable duplex ultrasound criteria worldwide. The consensus proposed in 2003 by the Society of Radiologists in Ultrasound has been a great attempt to create a conformity document, establishing grey scale and Doppler criteria in considering the different degrees of stenosis. According to this attempt, in 2010, the multi-parametric Deutsche Gesellschaft für Ultraschall in der Medizin ultrasound criteria have been proposed with a precise differentiation between main and additional criteria and depicted a different peak systolic velocity (PSV) threshold. In 2012, these criteria have been implemented, focusing on the multi-parametric approach, re-defining the PSV values and clearly introducing the concept of PSV average. Despite these attempts, a wide range of practice patterns still exists, with consistent disparities in patients' care. This paper collects these previous experiences and summarizes their strengths and weaknesses, to give a contribution in the carotid artery stenosis grading standardization using ultrasonic methods. Carotid ultrasound as the only diagnostic tool for the selection of patients for carotid surgery or stenting will be possible only with internationally accepted criteria.

The utility of transvaginal ultrasound in the ED evaluation of complications of first trimester pregnancy

BackgroundFor patients with early intrauterine pregnancy (IUP), the sonographic signs of the gestation may be below the resolution of transabdominal ultrasound (TAU); however, it may be identified by transvaginal ultrasound (TVU). We sought to determine how often TVU performed in the emergency department (ED) reveals a viable IUP after a nondiagnostic ED TAU and the impact of ED TVU on patient length of stay (LOS). MethodsThis was a retrospective cohort study of women presenting to our ED with complications of early pregnancy from January 1, 2007 to February 28, 2009 in a single urban adult ED. Abstractors recorded clinical and imaging data in a database. Patient imaging modality and results were recorded and compared with respect to ultrasound (US) findings and LOS. ResultsOf 2429 subjects identified, 795 required TVU as part of their care. Emergency department TVU was performed in 528 patients, and 267 went to radiology (RAD). Emergency department TVU identified a viable IUP in 261 patients (49.6%). Patients having initial ED US had shorter LOS than patients with initial RAD US (median 4.0 vs 6.0 hours; P < .001). Emergency department LOS was shorter for women who had ED TVU performed compared with those sent for RAD TVU regardless of the findings of the US (median 4.9 vs 6.7 hours; P < .001). There was no increased LOS for patients who needed further RAD US after an indeterminate ED TVU (7.0 vs 7.1 hours; P = .43). There was no difference in LOS for those who had a viable IUP confirmed on ED TAU vs ED (median 3.1 vs 3.2 hours, respectively; P < .32). ConclusionWhen an ED TVU was performed, a viable IUP was detected 49.6% of the time. Emergency department LOS was significantly shorter for women who received ED TVU after indeterminate ED TAU compared with those sent to RAD for TVU, with more marked time savings among those with live IUP diagnosed on ED TVU. For patients who do not receive a definitive diagnosis of IUP on ED TVU, this approach does not result in increased LOS.

Ultrasonography versus Computed Tomography for Suspected Nephrolithiasis

BackgroundThere is a lack of consensus about whether the initial imaging method for patients with suspected nephrolithiasis should be computed tomography (CT) or ultrasonography.MethodsIn this multicenter, pragmatic, comparative effectiveness trial, we randomly assigned patients 18 to 76 years of age who presented to the emergency department with suspected nephrolithiasis to undergo initial diagnostic ultrasonography performed by an emergency physician (point-of-care ultrasonography), ultrasonography performed by a radiologist (radiology ultrasonography), or abdominal CT. Subsequent management, including additional imaging, was at the discretion of the physician. We compared the three groups with respect to the 30-day incidence of high-risk diagnoses with complications that could be related to missed or delayed diagnosis and the 6-month cumulative radiation exposure. Secondary outcomes were serious adverse events, related serious adverse events (deemed attributable to study participation), pain (assessed on an 11-point visual-analogue scale, with higher scores indicating more severe pain), return emergency department visits, hospitalizations, and diagnostic accuracy.ResultsA total of 2759 patients underwent randomization: 908 to point-of-care ultrasonography, 893 to radiology ultrasonography, and 958 to CT. The incidence of high-risk diagnoses with complications in the first 30 days was low (0.4%) and did not vary according to imaging method. The mean 6-month cumulative radiation exposure was significantly lower in the ultrasonography groups than in the CT group (P<0.001). Serious adverse events occurred in 12.4% of the patients assigned to point-of-care ultrasonography, 10.8% of those assigned to radiology ultrasonography, and 11.2% of those assigned to CT (P=0.50). Related adverse events were infrequent (incidence, 0.4%) and similar across groups. By 7 days, the average pain score was 2.0 in each group (P=0.84). Return emergency department visits, hospitalizations, and diagnostic accuracy did not differ significantly among the groups.ConclusionsInitial ultrasonography was associated with lower cumulative radiation exposure than initial CT, without significant differences in high-risk diagnoses with complications, serious adverse events, pain scores, return emergency department visits, or hospitalizations. (Funded by the Agency for Healthcare Research and Quality; ClinicalTrials.gov number, NCT01451931.)