Cardiac Computed Tomography in Congenital Heart Disease.

Computed tomography (CT) has emerged as a leading imaging modality in the evaluation of congenital heart disease (CHD). With ever-faster acquisition speed, decreasing radiation exposure, impeccable anatomic detail, optional functional data, and numerous post-processing tools, CT offers broad utility in CHD diagnosis, preoperative planning, and postoperative assessment. In this article, the far-reaching role of CT in CHD is reviewed, focusing on technical imaging considerations and key clinical applications.

Abbreviated cardiac magnetic resonance imaging versus echocardiography for interval assessment of systolic function in Duchenne muscular dystrophy: patient satisfaction, clinical utility, and image quality.

PURPOSE: Poor acoustic windows make interval assessment of systolic function in patients with (Duchenne Muscular Dystrophy) DMD by echocardiography (echo) difficult. Cardiac magnetic resonance imaging (CMR) can be challenging in DMD patients due to study duration and patient discomfort. We developed an abbreviated CMR (aCMR) protocol and hypothesized that aCMR would compare favorably to echo in image quality and clinical utility without significant differences in exam duration, patient satisfaction, and functional measurements. METHODS: DMD patients were recruited prospectively to undergo echo and aCMR. Modalities were compared with a global quality assessment score (GQAS), clinical utility score (CUS), and patient satisfaction score (PSS). Results were compared using Wilcoxon signed-rank tests, Spearman correlations, intraclass correlations, and Bland-Altman analyses. RESULTS: Nineteen DMD patients were included. PSS scores and exam duration were equivalent between modalities, while CUS and GQAS scores favored aCMR. ACMR scored markedly higher than echo in RV visualization and assessment of atrial size. Older age was negatively correlated with echo GQAS and CUS scores, as well as aCMR PSS scores. Higher BMI was positively correlated with aCMR GQAS scores. Nighttime PPV requirement and non-ambulatory status were correlated with worse echo CUS scores. Poor image quality precluding quantification existed in five (26%) echo and zero (0%) aCMR studies. There was moderate correlation between aCMR and echo for global circumferential strain and left ventricular four chamber global longitudinal strain. CONCLUSION: The aCMR protocol resulted in improved clinical relevance and quality scores relative to echo, without significant detriment to patient satisfaction or exam duration.

Compact pediatric cardiac magnetic resonance imaging protocols.

Cardiac MRI is in many respects an ideal modality for pediatric cardiovascular imaging, enabling a complete noninvasive assessment of anatomy, morphology, function and flow in one radiation-free and potentially non-contrast exam. Nonetheless, traditionally lengthy and complex imaging acquisition strategies have often limited its broader use beyond specialized centers. In this review, the author presents practical cardiac MRI imaging protocols to facilitate the performance of succinct yet successful exams that provide the most salient clinical data for the majority of congenital and acquired pediatric cardiac disease. In addition, the author reviews newer and evolving techniques that permit more rapid but similarly diagnostic MRI, including compressed sensing and artificial intelligence/machine learning reconstruction, four-dimensional flow acquisition and blood pool contrast agents. With the modern armamentarium of cardiac MRI methods, the goal of compact yet comprehensive exams in children can now be realized.

Pediatric cardiothoracic vasculitis: multimodality imaging review.

The pediatric vasculitides are a relatively uncommon and heterogeneous group of disorders characterized by vessel inflammation, often with cardiothoracic involvement. Diagnosis and monitoring are often clinically challenging because of the nonspecific symptoms and laboratory markers. Thus, imaging has assumed increasing importance for early detection of disease activity, extent and complications as well as long-term monitoring pre- and post-treatment. Herein, we review the major pediatric vasculitides with frequent chest manifestations, including Takayasu arteritis, Kawasaki disease, granulomatosis with polyangiitis, eosinophilic granulomatosis with polyangiitis, microscopic polyangiitis, Behçet disease and potential mimics. We highlight key clinical features and management considerations, emphasizing the central role of imaging.

Computed tomography in tetralogy of Fallot: pre- and postoperative imaging evaluation.

Tetralogy of Fallot (TOF) is the most common cause of cyanotic congenital heart disease (CHD) and the most frequent complex CHD encountered in adulthood. Although children with TOF share four characteristic features (subaortic ventricular septal defect, overriding aorta, right ventricular hypertrophy, pulmonary stenosis), the clinical spectrum and course are in fact greatly heterogeneous. Echocardiography remains the mainstay for diagnosis, presurgical planning and postoperative follow-up. However, with continued technological advances, CT now plays an increasing role in TOF evaluation and management, helping to minimize routine invasive catheter angiography. Preoperatively, CT is uniquely suited to assess associated pulmonary arterial, aortic and coronary anomalies as well as extra-cardiovascular structures and is particularly helpful for delineating complex anatomy in the TOF subtypes of absent pulmonary valve and pulmonary atresia with major aortopulmonary collaterals. Postoperatively, CT is useful for identifying surgical complications and for long-term monitoring including volumetry quantification, especially in children for whom MRI is contraindicated or limited by implanted devices such as pacemakers and stents. In this article, we review key clinical features and considerations in the pre- and postoperative TOF patient and the burgeoning role of CT for facilitating accurate diagnosis and personalized intervention.

Increasing the Utilization of Moderate Sedation Services for Pediatric Imaging.

Performing motion-free imaging is frequently challenging in children. To bridge the gap between examinations performed in children who are awake and in those under general anesthesia, a moderate sedation program was implemented at our institution but was seldom used despite substantial eligibility. In conjunction with a 5-month quality improvement (QI) course, a multidisciplinary team was assembled and, by using an A3 approach, sought to address the most important key drivers of low utilization, namely the need for clear moderate sedation eligibility criteria, reliable protocol routing order, consistent moderate sedation screening performed by registered nurses (RNs), and enhanced visibility of moderate sedation services to ordering providers. Initial steps focused on developing better-defined criteria and protocoling standard work for technologists and RNs, with coaching and audits. Modality-specific forecasting was then implemented to reroute profiles of patients who were awaiting scheduling or already scheduled for an examination with general anesthesia to the moderate sedation queue to identify more eligible patients. These manual efforts were coupled with higher reliability but more protracted electronic health record changes, facilitating automated protocol routing on the basis of moderate sedation eligibility and order entry constraints. As a result, scheduled imaging examinations requiring moderate sedation increased from a mean of 1.2 examinations per week to a sustained 6.1 examinations per week (range, 4-8) over the 5-month period, exceeding the team SMART (specific, measurable, achievable, relevant, and time bound) goal to achieve an average of five examinations per week by the QI course end. By targeting the most high-impact yet modifiable process deficiencies through a multifaceted team approach and initially investing in manual efforts to gain cultural buy-in while awaiting higher-reliability interventions, the project achieved success and may serve as a more general model for workflow change when there is organizational resistance. ©RSNA, 2021.

Free-breathing Accelerated Cardiac MRI Using Deep Learning: Validation in Children and Young Adults.

Background Obtaining ventricular volumetry and mass is key to most cardiac MRI but challenged by long multibreath-hold acquisitions. Purpose To assess the image quality and performance of a highly accelerated, free-breathing, two-dimensional cine cardiac MRI sequence incorporating deep learning (DL) reconstruction compared with reference standard balanced steady-state free precession (bSSFP). Materials and Methods A DL algorithm was developed to reconstruct custom 12-fold accelerated bSSFP cardiac MRI cine images from coil sensitivity maps using 15 iterations of separable three-dimensional convolutions and data consistency steps. The model was trained, validated, and internally tested in 10, two, and 10 adult human volunteers, respectively, based on vendor partner-supplied fully sampled bSSFP acquisitions. For prospective external clinical validation, consecutive children and young adults undergoing cardiac MRI from September through December 2019 at a single children's hospital underwent both conventional and highly accelerated short-axis bSSFP cine acquisitions in one MRI examination. Two radiologists scored overall and volumetric three-dimensional mesh image quality of all short-axis stacks on a five-point Likert scale and manually segmented endocardial and epicardial contours. Scan times and image quality were compared using the Wilcoxon rank sum test. Measurement agreement was assessed with intraclass correlation coefficient and Bland-Altman analysis. Results Fifty participants (mean age, 16 years ± 4 [standard deviation]; range, 5-30 years; 29 men) were evaluated. The mean prescribed acquisition times of accelerated scans (non-breath-held) and bSSFP (excluding breath-hold time) were 0.9 minute ± 0.3 versus 3.0 minutes ± 1.9 (P < .001). Overall and three-dimensional mesh image quality scores were, respectively, 3.8 ± 0.6 versus 4.3 ± 0.6 (P < .001) and 4.0 ± 1.0 versus 4.4 ± 0.8 (P < .001). Raters had strong agreement between all bSSFP and DL measurements, with intraclass correlation coefficients of 0.76 to 0.97, near-zero mean differences, and narrow limits of agreement. Conclusion With slightly lower image quality yet much faster speed, deep learning reconstruction may allow substantially shorter acquisition times of cardiac MRI compared with conventional balanced steady-state free precession MRI performed for ventricular volumetry. © RSNA, 2021 Online supplemental material is available for this article.

Clinical use of shear-wave elastography for detecting liver fibrosis in children and adolescents with cystic fibrosis.

BACKGROUND: Complications from liver cirrhosis are a leading cause of death in children with cystic fibrosis. Identifying children at risk for developing liver cirrhosis and halting its progression are critical to reducing liver-associated mortality. OBJECTIVE: Quantitative US imaging, such as shear-wave elastography (SWE), might improve the detection of liver fibrosis in children with cystic fibrosis (CF) over gray-scale US alone. We incorporated SWE in our pediatric CF liver disease screening program and evaluated its performance using magnetic resonance (MR) elastography. MATERIALS AND METHODS: Ninety-four children and adolescents with CF underwent 178 SWE exams, aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-glutamyl transferase (GGT) and platelet measurements. Of these, 27 children underwent 34 MR elastography exams. We evaluated SWE performance using 6-MHz and 9-MHZ point SWE, and 9-MHz two-dimensional (2-D) SWE. RESULTS: The 6-MHz point SWE was the only method that correlated with MR elastography (r=0.52; 95% confidence interval [CI] 0.20-0.74; P=0.003). SWE of 1.45 m/s distinguished normal from abnormal MR elastography (79% sensitivity, 100% specificity, 100% positive predictive value [PPV], 55% negative predictive value [NPV], area under the receiver operating characteristic [AUROC] curve 0.94). SWE of 1.84 m/s separated mild-moderate (3.00-4.77 kPa) from severe (>4.77 kPa) MR elastography (88% sensitivity, 86% specificity, 78% PPV, 93% NPV, AUROC 0.79). Elevations of AST, ALT, GGT and thrombocytopenia were associated with higher SWE. AST-to-platelet ratio index of 0.42, fibrosis-4 of 0.29, and GGT-to-platelet ratio of 1.43 all had >95% NPV for SWE >1.84 m/s. CONCLUSION: Given its correlation with MR elastography, SWE might be a clinically useful predictor of liver fibrosis. We identified imaging criteria delineating the use of SWE to identify increased liver stiffness in children with CF. With multicenter validation, these data might be used to improve the detection and monitoring of liver fibrosis in children with CF.

Image-based scaling laws for somatic growth and pulmonary artery morphometry from infancy to adulthood.

Pulmonary artery (PA) morphometry has been extensively explored in adults, with particular focus on intra-acinar arteries. However, scaling law relationships for length and diameter of extensive preacinar PAs by age have not been previously reported for in vivo human data. To understand preacinar PA growth spanning children to adults, we performed morphometric analyses of all PAs visible in the computed tomography (CT) and magnetic resonance (MR) images from a healthy subject cohort [n = 16; age: 1-51 yr; body surface area (BSA): 0.49-2.01 m2]. Subject-specific anatomic PA models were constructed from CT and MR images, and morphometric information-diameter, length, tortuosity, bifurcation angle, and connectivity-was extracted and sorted into diameter-defined Strahler orders. Validation of Murray's law, describing optimal scaling exponents of radii for branching vessels, was performed to determine how closely PAs conform to this classical relationship. Using regression analyses of vessel diameters and lengths against orders and patient metrics (BSA, age, height), we found that diameters increased exponentially with order and allometrically with patient metrics. Length increased allometrically with patient metrics, albeit weakly. The average tortuosity index of all vessels was 0.026 ± 0.024, average bifurcation angle was 28.2 ± 15.1°, and average Murray's law exponent was 2.92 ± 1.07. We report a set of scaling laws for vessel diameter and length, along with other morphometric information. These provide an initial understanding of healthy structural preacinar PA development with age, which can be used for computational modeling studies and comparison with diseased PA anatomy.NEW & NOTEWORTHY Pulmonary artery (PA) morphometry studies to date have focused primarily on large arteries and intra-acinar arteries in either adults or children, neglecting preacinar arteries in both populations. Our study is the first to quantify in vivo preacinar PA morphometry changes spanning infants to adults. For preacinar arteries > 1 mm in diameter, we identify scaling laws for vessel diameters and lengths with patient metrics of growth and establish a healthy PA morphometry baseline for most preacinar PAs.

Rosette Trajectories Enable Ungated, Motion-Robust, Simultaneous Cardiac and Liver T2 * Iron Assessment.

BACKGROUND: Quantitative T2 * MRI is the standard of care for the assessment of iron overload. However, patient motion corrupts T2 * estimates. PURPOSE: To develop and evaluate a motion-robust, simultaneous cardiac and liver T2 * imaging approach using non-Cartesian, rosette sampling and a model-based reconstruction as compared to clinical-standard Cartesian MRI. STUDY TYPE: Prospective. PHANTOM/POPULATION: Six ferumoxytol-containing phantoms (26-288 µg/mL). Eight healthy subjects and 18 patients referred for clinically indicated iron overload assessment. FIELD STRENGTH/SEQUENCE: 1.5T, 2D Cartesian and rosette gradient echo (GRE) ASSESSMENT: GRE T2 * values were validated in ferumoxytol phantoms. In healthy subjects, test-retest and spatial coefficient of variation (CoV) analysis was performed during three breathing conditions. Cartesian and rosette T2 * were compared using correlation and Bland-Altman analysis. Images were rated by three experienced radiologists on a 5-point scale. STATISTICAL TESTS: Linear regression, analysis of variance (ANOVA), and paired Student's t-testing were used to compare reproducibility and variability metrics in Cartesian and rosette scans. The Wilcoxon rank test was used to assess reader score comparisons and reader reliability was measured using intraclass correlation analysis. RESULTS: Rosette R2* (1/T2 *) was linearly correlated with ferumoxytol concentration (r2 = 1.00) and not significantly different than Cartesian values (P = 0.16). During breath-holding, ungated rosette liver and heart T2 * had lower spatial CoV (liver: 18.4 ± 9.3% Cartesian, 8.8% ± 3.4% rosette, P = 0.02, heart: 37.7% ± 14.3% Cartesian, 13.4% ± 1.7% rosette, P = 0.001) and higher-quality scores (liver: 3.3 [3.0-3.6] Cartesian, 4.7 [4.1-4.9] rosette, P = 0.005, heart: 3.0 [2.3-3] Cartesian, 4.5 [3.8-5.0] rosette, P = 0.005) compared to Cartesian values. During free-breathing and failed breath-holding, Cartesian images had very poor to average image quality with significant artifacts, whereas rosette remained very good, with minimal artifacts (P = 0.001). DATA CONCLUSION: Rosette k-sampling with a model-based reconstruction offers a clinically useful motion-robust T2 * mapping approach for iron quantification. J. MAGN. RESON. IMAGING 2020;52:1688-1698.

Deep learning to automate Brasfield chest radiographic scoring for cystic fibrosis.

BACKGROUND: The aim of this study was to evaluate the hypothesis that a deep convolutional neural network (DCNN) model could facilitate automated Brasfield scoring of chest radiographs (CXRs) for patients with cystic fibrosis (CF), performing similarly to a pediatric radiologist. METHODS: All frontal/lateral chest radiographs (2058 exams) performed in CF patients at a single institution from January 2008-2018 were retrospectively identified, and ground-truth Brasfield scoring performed by a board-certified pediatric radiologist. 1858 exams (90.3%) were used to train and validate the DCNN model, while 200 exams (9.7%) were reserved for a test set. Five board-certified pediatric radiologists independently scored the test set according to the Brasfield method. DCNN model vs. radiologist performance was compared using Spearman correlation (ρ) as well as mean difference (MD), mean absolute difference (MAD), and root mean squared error (RMSE) estimation. RESULTS: For the total Brasfield score, ρ for the model-derived results computed pairwise with each radiologist's scores ranged from 0.79-0.83, compared to 0.85-0.90 for radiologist vs. radiologist scores. The MD between model estimates of the total Brasfield score and the average score of radiologists was -0.09. Based on MD, MAD, and RMSE, the model matched or exceeded radiologist performance for all subfeatures except air-trapping and large lesions. CONCLUSIONS: A DCNN model is promising for predicting CF Brasfield scores with accuracy similar to that of a pediatric radiologist.

Lumbar Spine MRI: Missed Opportunities for Abdominal Aortic Aneurysm Detection.

PURPOSE: The US Preventive Services Task Force (USPSTF) recommends 1-time sonographic screening for abdominal aortic aneurysms (AAAs) in male smokers ages 65-75 and other selected individuals in this age group based on risk factors. Patients in this age range are frequent utilizers of lumbar spine MRI, in which the abdominal aorta is typically fully imaged. The purpose of this study was to assess the potential detection rate of AAAs on lumbar spine MRI performed in the USPSTF screening age range with systematic aortic measurement and the frequency with which AAAs are currently reported in practice. MATERIALS AND METHODS: All consecutive lumbar spine MRI exams performed without contrast at a single academic tertiary care center over a 1-year period (4/1/2016-3/31/2017) in patients ages 65-75 were retrospectively reviewed. Maximal anteroposterior, and transverse dimensions of the abdominal aorta were measured using axial T2-weighted images, supplemented with sagittal T2-weighted images if assessment was limited by field-of-view or artifact. The detection rate of AAA, defined as dilation of the aorta to a diameter of ≥3 cm, size of AAAs detected, and frequency with which AAAs were reported, were assessed. Differences in aortic diameters and aneurysm detection rates between genders were compared with the unpaired 2-sample t test. RESULTS: Three hundred and ninety-five lumbar spine MRIs were reviewed, 240 (60.8%) in women and 155 (39.2%) in men, with mean ± standard deviation (SD) age of 70.2 ± 3.2 years. AAAs were detected in 38/395 (9.6%) cases, most (33/38, 86.8%) of which were <4 cm. Of these, only 4 (10.5%) were reported by the interpreting radiologist; 3/4 (75%) corresponded to aneurysms ≥4 cm. CONCLUSION: Lumbar spine MRI performed in the USPSTF AAA screening age range, especially in men, facilitates frequent detection of AAA when the aorta is systematically measured. However, in typical lumbar spine assessment, AAAs are often underreported, particularly for smaller aneurysms.

Cardiac MRI-guided interventional occlusion of ventricular septal rupture in a patient with cobalt alloy stent.

The case of a 68-year-old man with chest pain for 3 days is presented. Coronary angiography demonstrated subtotal occlusion of the mid-left anterior descending artery. A drug-eluting cobalt alloy stent was implanted after balloon dilation. On the 3rd postoperative day, echocardiography showed a ventricular septal rupture (VSR) (7 mm diameter) near the cardiac apex and ventricular aneurysm. On cardiac magnetic resonance imaging (MRI), the VSR was shown to be 11 mm in diameter. The membranous septum was 32 and 27.8 mm along the anteroposterior and superoinferior axes, respectively. The left-to-right shunt was apparent. Four weeks later, interventional therapy was performed to occlude the VSR according to the result of the MRI. The symptoms improved rapidly, and the patient was discharged. At the 4-month follow up visit, cardiac MRI revealed no shunt at the occlusion site, and the edge of the occluder was secured in the adjacent normal cardiac tissues. In conclusion, cardiac MRI could be considered for patients with a newly implanted cobalt alloy stent to provide an accurate assessment of VSR.

Cross-sectional imaging of congenital pulmonary artery anomalies.

Congenital pulmonary artery (PA) anomalies comprise a rare and heterogeneous spectrum of disease, ranging from abnormal origins to complete atresia. They may present in early infancy or more insidiously in adulthood, often in association with congenital heart disease such as tetralogy of Fallot or other syndromes. In recent years, cross-sectional imaging, including computed tomography (CT) and magnetic resonance imaging (MRI), has become widely utilized for the noninvasive assessment of congenital PA diseases, supplementing echocardiography and at times supplanting invasive angiography. In this article, modern CT and MRI techniques for imaging congenital PA disorders are summarized. The key clinical features, cross-sectional imaging findings, and treatment options for the most commonly encountered entities are then reviewed. Emphasis is placed on the ever-growing role of cross-sectional imaging options in facilitating early and accurate diagnosis and tailored treatment.

Feasibility and utility of dual-energy chest CTA for preoperative planning in pediatric pulmonary artery reconstruction.

The purpose of this study was to assess in pediatric pulmonary artery (PA) reconstruction candidates the feasibility and added utility of preoperative chest computed tomography angiography (CTA) using dual-energy technique, from which perfused blood volume (PBV)/iodine maps can be generated as a surrogate of pulmonary perfusion. Pediatric PA reconstruction patients were prospectively recruited for a new dose-neutral dual-energy CTA protocol. For each case, the severity of anatomic PA obstruction was graded by two pediatric cardiovascular radiologists in consensus using a modified Qanadli index. PBV maps were qualitatively reviewed and auto-segmented using Siemens syngo.via software. Associations between Qanadli scores and PBV were assessed with Spearman correlation (r) and ROC analysis. Effective radiation doses were estimated from dose-length product and ICRP 103 k-factors, using cubic Hermite spline interpolation. 19 patients were recruited with mean (SD) age of 6.0 (5.1), 11 (57.9%) female, 11 (73.7%) anesthetized. Higher QS correlated with lower PBV, both on a whole lung (r = - 0.54, p < 0.001) and lobar (r = - 0.50, p < 0.001) basis. The lung with lowest absolute PBV was predictive of the lung with highest Qanadli score, with AUC of 0.70 (95% CI 0.47-0.93). Qualitatively, PBV maps were heterogeneous, corresponding to multifocal PA stenoses, with decreased iodine content in areas of most severe obstruction. In conclusion, dual-energy chest CTA is feasible for pediatric PA reconstruction candidates. PBV maps show deficits in regions of more severe anatomic obstruction and may serve as a novel biomarker in this population.

Cross-sectional imaging of thoracic traumatic aortic injury.

Aortic injury remains a major contributor to morbidity and mortality from acute thoracic trauma. While such injuries were once nearly uniformly fatal, the advent of cross-sectional imaging in recent years has facilitated rapid diagnosis and triage, greatly improving outcomes. In fact, cross-sectional imaging is now the diagnostic test of choice for traumatic aortic injury (TAI), specifically computed tomography angiography (CTA) in the acute setting and CTA or magnetic resonance angiography (MRA) in follow-up. In this review, we present an up-to-date discussion of acute traumatic thoracic aortic injury with a focus on optimal and emerging CT/MR techniques, imaging findings of TAI, and potential pitfalls.

Ultrafast pediatric chest computed tomography: comparison of free-breathing vs. breath-hold imaging with and without anesthesia in young children.

BACKGROUND: General anesthesia (GA) or sedation has been used to obtain good-quality motion-free breath-hold chest CT scans in young children; however pulmonary atelectasis is a common and problematic accompaniment that can confound diagnostic utility. Dual-source multidetector CT permits ultrafast high-pitch sub-second examinations, minimizing motion artifact and potentially eliminating the need for a breath-hold. OBJECTIVE: The purpose of this study was to evaluate the feasibility of free-breathing ultrafast pediatric chest CT without GA and to compare it with breath-hold and non-breath-hold CT with GA. MATERIALS AND METHODS: Young (≤3 years old) pediatric outpatients scheduled for chest CT under GA were recruited into the study and scanned using one of three protocols: GA with intubation, lung recruitment and breath-hold; GA without breath-hold; and free-breathing CT without anesthesia. In all three protocols an ultrafast high-pitch CT technique was used. We evaluated CT images for overall image quality, presence of atelectasis and motion artifacts. RESULTS: We included 101 scans in the study. However the GA non-breath-hold technique was discontinued after 15 scans, when it became clear that atelectasis was a major issue despite diligent attempts to mitigate it. This technique was therefore not included in statistical evaluation (86 remaining patients). Overall image quality was higher (P=0.001) and motion artifacts were fewer (P<.001) for scans using the GA with intubation and recruitment technique compared to scans in the non-GA free-breathing group. However no significant differences were observed regarding the presence of atelectasis between these groups. CONCLUSION: We demonstrated that although overall image quality was best and motion artifact least with a GA-breath-hold intubation and recruitment technique, free-breathing ultrafast pediatric chest CT without anesthesia provides sufficient image quality for diagnostic purposes and can be successfully performed both without and with contrast agent in young infants.

Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists.

BACKGROUND: Chest radiograph interpretation is critical for the detection of thoracic diseases, including tuberculosis and lung cancer, which affect millions of people worldwide each year. This time-consuming task typically requires expert radiologists to read the images, leading to fatigue-based diagnostic error and lack of diagnostic expertise in areas of the world where radiologists are not available. Recently, deep learning approaches have been able to achieve expert-level performance in medical image interpretation tasks, powered by large network architectures and fueled by the emergence of large labeled datasets. The purpose of this study is to investigate the performance of a deep learning algorithm on the detection of pathologies in chest radiographs compared with practicing radiologists. METHODS AND FINDINGS: We developed CheXNeXt, a convolutional neural network to concurrently detect the presence of 14 different pathologies, including pneumonia, pleural effusion, pulmonary masses, and nodules in frontal-view chest radiographs. CheXNeXt was trained and internally validated on the ChestX-ray8 dataset, with a held-out validation set consisting of 420 images, sampled to contain at least 50 cases of each of the original pathology labels. On this validation set, the majority vote of a panel of 3 board-certified cardiothoracic specialist radiologists served as reference standard. We compared CheXNeXt's discriminative performance on the validation set to the performance of 9 radiologists using the area under the receiver operating characteristic curve (AUC). The radiologists included 6 board-certified radiologists (average experience 12 years, range 4-28 years) and 3 senior radiology residents, from 3 academic institutions. We found that CheXNeXt achieved radiologist-level performance on 11 pathologies and did not achieve radiologist-level performance on 3 pathologies. The radiologists achieved statistically significantly higher AUC performance on cardiomegaly, emphysema, and hiatal hernia, with AUCs of 0.888 (95% confidence interval [CI] 0.863-0.910), 0.911 (95% CI 0.866-0.947), and 0.985 (95% CI 0.974-0.991), respectively, whereas CheXNeXt's AUCs were 0.831 (95% CI 0.790-0.870), 0.704 (95% CI 0.567-0.833), and 0.851 (95% CI 0.785-0.909), respectively. CheXNeXt performed better than radiologists in detecting atelectasis, with an AUC of 0.862 (95% CI 0.825-0.895), statistically significantly higher than radiologists' AUC of 0.808 (95% CI 0.777-0.838); there were no statistically significant differences in AUCs for the other 10 pathologies. The average time to interpret the 420 images in the validation set was substantially longer for the radiologists (240 minutes) than for CheXNeXt (1.5 minutes). The main limitations of our study are that neither CheXNeXt nor the radiologists were permitted to use patient history or review prior examinations and that evaluation was limited to a dataset from a single institution. CONCLUSIONS: In this study, we developed and validated a deep learning algorithm that classified clinically important abnormalities in chest radiographs at a performance level comparable to practicing radiologists. Once tested prospectively in clinical settings, the algorithm could have the potential to expand patient access to chest radiograph diagnostics.