Chest radiography is a frequently used imaging modality in children. However, only fair to moderate inter-observer agreement has been reported between chest radiograph interpreters. Most studies were not performed in real-world clinical settings. Our aims were to examine the agreement between emergency department pediatricians and board-certified radiologists in a pediatric real-life setting and to identify clinical risk factors for the discrepancies. Included were children aged 3 months to 18 years who underwent chest radiography in the emergency department not during the regular hours of radiologist interpretation. Every case was reviewed by an expert panel. Inter-observer agreement between emergency department pediatricians and board-certified radiologists was assessed by Cohen\'s kappa; risk factors for disagreement were analyzed. Among 1373 cases, the level of agreement between emergency department pediatricians and board-certified radiologists was \"moderate\" (k = 0.505). For radiographs performed after midnight, agreement was only \"fair\" (k = 0.391). The expert panel identified clinically relevant disagreements in 260 (18.9%) of the radiographs. Over-treatment of antibiotics was identified in 121 (8.9%) of the cases and under-treatment in 79 (5.8%). In a multivariable logistic regression, the following parameters were found to be significantly associated with disagreements: neurological background (p = 0.046), fever (p = 0.001), dyspnea (p = 0.014), and radiographs performed after midnight (p = 0.007).
CONCLUSIONS: Moderate agreement was found between emergency department pediatricians and board-certified radiologists in interpreting chest radiographs. Neurological background, fever, dyspnea, and radiographs performed after midnight were identified as risk factors for disagreement. Implementing these findings could facilitate the use of radiologist expertise, save time and resources, and potentially improve patient care.
BACKGROUND: • Only fair to moderate inter-observer agreement has been reported between chest radiograph interpreters. • Most studies were not performed in real-world clinical settings. Clinical risk factors for disagreements have not been reported.
BACKGROUND: • In this study, which included 1373 cases at the emergency department, the level of agreement between interpreters was only \"moderate.\" • The major clinical parameters associated with interpretation discrepancies were neurological background, fever, dyspnea, and interpretations conducted during the night shift.

BACKGROUND: Chest X-rays (CXR) are essential for diagnosing a variety of conditions, but when used on new populations, model generalizability issues limit their efficacy. Generative AI, particularly denoising diffusion probabilistic models (DDPMs), offers a promising approach to generating synthetic images, enhancing dataset diversity. This study investigates the impact of synthetic data supplementation on the performance and generalizability of medical imaging research.
METHODS: The study employed DDPMs to create synthetic CXRs conditioned on demographic and pathological characteristics from the CheXpert dataset. These synthetic images were used to supplement training datasets for pathology classifiers, with the aim of improving their performance. The evaluation involved three datasets (CheXpert, MIMIC-CXR, and Emory Chest X-ray) and various experiments, including supplementing real data with synthetic data, training with purely synthetic data, and mixing synthetic data with external datasets. Performance was assessed using the area under the receiver operating curve (AUROC).
RESULTS: Adding synthetic data to real datasets resulted in a notable increase in AUROC values (up to 0.02 in internal and external test sets with 1000% supplementation, p-value <0.01 in all instances). When classifiers were trained exclusively on synthetic data, they achieved performance levels comparable to those trained on real data with 200%-300% data supplementation. The combination of real and synthetic data from different sources demonstrated enhanced model generalizability, increasing model AUROC from 0.76 to 0.80 on the internal test set (p-value <0.01).
CONCLUSIONS: Synthetic data supplementation significantly improves the performance and generalizability of pathology classifiers in medical imaging.
BACKGROUND: Dr. Gichoya is a 2022 Robert Wood Johnson Foundation Harold Amos Medical Faculty Development Program and declares support from RSNA Health Disparities grant (#EIHD2204), Lacuna Fund (#67), Gordon and Betty Moore Foundation, NIH (NIBIB) MIDRC grant under contracts 75N92020C00008 and 75N92020C00021, and NHLBI Award Number R01HL167811.

UNASSIGNED: Scoliosis is an abnormal side-to-side spinal curve of greater than or equal to 100 Cobb angle. It is the most common spinal deformity in children and adolescents. Epidemiological evidence about scoliosis is scarce in Africa, including Ethiopia. This study was aimed at determining the prevalence of scoliosis among adolescents and analysing its association with age and sex using plain chest radiographs obtained for non-spinal reasons in Tikur Anbessa Specialized Hospital.
UNASSIGNED: All non-tilted, non-rotated, and non-poorly penetrated digital plain chest radiographs of adolescents aged 10 to 19 years obtained at Tikur Anbessa Specialized Hospital between January 1 and December 31, 2019, were measured for the coronal Cobb angle. The data were cleaned, coded, and entered into SPSS version 26 for analysis. Chi-square, and linear regression, and logistic regression analyses were also carried out to evaluate the effect of sex and age on scoliosis.
UNASSIGNED: The Cobb angles of 1,369 posteroanterior chest radiographs of adolescents were measured. Thirty (2.2%, 95% CI: 1.4%, 3.0%) of these were found to have scoliosis. The mean coronal Cobb angle was 2.270±6.320. There was no statistically significant difference between the prevalence of scoliosis in boys (2.21%) and girls (2.17%) (X2=0.003, P=0.954). Likewise, age did not show any statistically significant difference in the prevalence of scoliosis (X2=2.655, P=0.265).
UNASSIGNED: This study revealed that incidental finding of adolescent scoliosis in plain chest radiographs is common. Further study using whole spine radiography should be carried out to determine the true general population prevalence of scoliosis in Ethiopia.

UNASSIGNED: Central venous catheters are increasingly inserted using point-of-care ultrasound (POCUS) guidance. Following insertion, it is still common to request a confirmatory chest radiograph for subclavian and internal jugular lines, at least outside of the operating theater. This scoping review addresses: (i) the justification for routine post-insertion radiographs, (ii) whether it would better to use post-insertion POCUS instead, and (iii) the perceived barriers to change.
UNASSIGNED: We searched the electronic databases, Ovid MEDLINE (1946-) and Ovid EMBASE (1974-), using the MESH terms (\"Echography\" OR \"Ultrasonography\" OR \"Ultrasound\") AND \"Central Venous Catheter\" up until February 2023. We also searched clinical practice guidelines, and targeted literature, including cited and citing articles. We included adults (⩾18 years) and English and French language publications. We included randomized control trials, prospective and retrospective cohort studies, systematic reviews, and surveys.
UNASSIGNED: Four thousand seventy-one articles were screened, 117 full-text articles accessed, and 41 retained. Thirteen examined cardiac/vascular methods; 5 examined isolated contrast-enhanced ultrasonography; 7 examined isolated rapid atrial swirl sign; and 13 examined combined/integrated methods. In addition, three systematic reviews/meta-analyses and one survey addressed barriers to POCUS adoption.
UNASSIGNED: We believe that the literature supports retiring the routine post-central line chest radiograph. This is not only because POCUS has made line insertion safer, but because POCUS performs at least as well, and is associated with less radiation, lower cost, time savings, and greater accuracy. There has been less written about perceived barriers to change, but the literature shows that these concerns- which include upfront costs, time-to-train, medicolegal concerns and habit- can be challenged and hence overcome.

BACKGROUND: The study aimed to develop and validate a deep learning-based Computer Aided Triage (CADt) algorithm for detecting pleural effusion in chest radiographs using an active learning (AL) framework. This is aimed at addressing the critical need for a clinical grade algorithm that can timely diagnose pleural effusion, which affects approximately 1.5 million people annually in the United States.
METHODS: In this multisite study, 10,599 chest radiographs from 2006 to 2018 were retrospectively collected from an institution in Taiwan to train the deep learning algorithm. The AL framework utilized significantly reduced the need for expert annotations. For external validation, the algorithm was tested on a multisite dataset of 600 chest radiographs from 22 clinical sites in the United States and Taiwan, which were annotated by three U.S. board-certified radiologists.
RESULTS: The CADt algorithm demonstrated high effectiveness in identifying pleural effusion, achieving a sensitivity of 0.95 (95% CI: [0.92, 0.97]) and a specificity of 0.97 (95% CI: [0.95, 0.99]). The area under the receiver operating characteristic curve (AUC) was 0.97 (95% DeLong\'s CI: [0.95, 0.99]). Subgroup analyses showed that the algorithm maintained robust performance across various demographics and clinical settings.
CONCLUSIONS: This study presents a novel approach in developing clinical grade CADt solutions for the diagnosis of pleural effusion. The AL-based CADt algorithm not only achieved high accuracy in detecting pleural effusion but also significantly reduced the workload required for clinical experts in annotating medical data. This method enhances the feasibility of employing advanced technological solutions for prompt and accurate diagnosis in medical settings.

OBJECTIVE: Infants with anterior abdominal wall defects (AWD) can suffer from pulmonary complications. Our aims were to determine if the chest radiographic thoracic areas (CRTAs) on day one differed between infants with exomphalos or gastroschisis, whether this related to differing severity of outcomes and if they were lower than those of controls indicating abnormal antenatal lung growth.
METHODS: A review of infants with exomphalos or gastroschisis born between January 2004 and January 2023 was conducted. The control group was term, newborn infants ventilated for poor respiratory drive at birth. Chest radiographs on day one were analysed and the highest CRTA in the first 24 h after birth for each infant included in the analysis.
RESULTS: The 127 infants with gastroschisis had a lower gestational age and birthweight than the 62 exomphalos infants and 130 controls (all p<0.001) The CRTAs of the controls were greater than the CRTAs of the exomphalos and the gastroschisis infants (p = 0.001). The median CRTA corrected for birthweight was lower in the exomphalos infants [688, IQR 568-875 mm2/kg] than the gastroschisis infants [813, IQE 695-915 mm2/kg] No gastroschisis infant developed bronchopulmonary dysplasia (BPD). A CRTA of 1759 mm2 had a sensitivity of 81 % and specificity of 71 % in predicting BPD in infants with exomphalos.
CONCLUSIONS: Infants with gastroschisis or exomphalos had lower CRTAs than controls suggesting both groups had abnormal antenatal lung development. The CRTA was lower in the exomphalos infants who also had worse respiratory outcomes, hence CRTA assessment may a useful prognostic aid.

Background and Objectives: Our quality management project aims to decrease by 20% the number of neonates with respiratory distress undergoing chest radiographs as part of their diagnosis and monitoring. Materials and Methods: This quality management project was developed at Life Memorial Hospital, Bucharest, between 2021 and 2023. Overall, 125 patients were included in the study. The project consisted of a training phase, then an implementation phase, and the final results were measured one year after the end of the implementation phase. The imaging protocol consisted of the performance of lung ultrasounds in all the patients on CPAP (continuous positive airway pressure) or mechanical ventilation (first ultrasound at about 90 min after delivery) and the performance of chest radiographs after endotracheal intubation in any case of deterioration of the status of the patient or if such a decision was taken by the clinician. The baseline characteristics of the population were noted and compared between years 2021, 2022, and 2023. The primary outcome measures were represented by the number of X-rays performed in ventilated patients per year (including the patients on CPAP, SIMV (synchronized intermittent mandatory ventilation), IPPV (intermittent positive pressure ventilation), HFOV (high-frequency oscillatory ventilation), the number of X-rays performed per patient on CPAP/year, the number of chest X-rays performed per mechanically ventilated patient/year and the mean radiation dose/patient/year. There was no randomization of the patients for the intervention. The results were compared between the year before the project was introduced and the 2 years across which the project was implemented. Results: The frequency of cases in which no chest X-ray was performed was significantly higher in 2023 compared to 2022 (58.1% vs. 35.8%; p = 0.03) or 2021 (58.1% vs. 34.5%; p = 0.05) (a decrease of 22.3% in 2023 compared with 2022 and of 23.6% in 2023 compared with 2021). The frequency of cases with one chest X-ray was significantly lower in 2023 compared to 2022 (16.3% vs. 35.8%; p = 0.032) or 2021 (16.3% vs. 44.8%; p = 0.008). The mean radiation dose decreased from 5.89 Gy × cm2 in 2021 to 3.76 Gy × cm2 in 2023 (36% reduction). However, there was an increase in the number of ventilated patients with more than one X-ray (11 in 2023 versus 6 in 2021). We also noted a slight annual increase in the mean number of X-rays per patient receiving CPAP followed by mechanical ventilation (from 1.80 in 2021 to 2.33 in 2022 and then 2.50 in 2023), and there was a similar trend in the patients that received only mechanical ventilation without a statistically significant difference in these cases. Conclusions: The quality management project accomplished its goal by obtaining a statistically significant increase in the number of ventilated patients in which chest radiographs were not performed and also resulted in a more than 30% decrease in the radiation dose per ventilated patient. This task was accomplished mainly by increasing the number of patients on CPAP and the use only of lung ultrasound in the patients on CPAP and simple cases.

OBJECTIVE: While computed tomography (CT) is widely acknowledged as superior to chest radiographs for acute diagnostics, its efficacy in diagnosing acute heart failure (AHF) remains unexplored. This prospective study included consecutive patients with dyspnoea undergoing simultaneous low-dose chest CT (LDCT) and chest radiographs. Here, we aimed to determine if LDCT is superior to chest radiographs to confirm pulmonary congestion in dyspnoeic patients with suspected AHF.
RESULTS: An observational, prospective study, including dyspnoeic patients from the emergency department. All patients underwent concurrent clinical examination, laboratory tests, echocardiogram, chest radiographs, and LDCT. The primary efficacy measure to compare the two radiological methods was conditional odds ratio (cOR). The primary outcome was adjudicated AHF, ascertained by comprehensive expert consensus. The secondary outcome, echo-bnp AHF, was an objective AHF diagnosis based on echocardiographic cardiac dysfunction, elevated cardiac filling pressure, loop diuretic administration, and NT-pro brain natriuretic peptide > 300 pg/mL. Of 228 dyspnoeic patients, 64 patients (28%) had adjudicated AHF, and 79 patients (35%) had echo-bnp AHF. Patients with AHF were older (78 years vs. 73 years), had lower left ventricular ejection fraction (36% vs. 55%), had higher elevated left ventricular filling pressures (98% vs. 18%), and had higher NT-pro brain natriuretic peptide levels (3628 pg/mL vs. 470 pg/mL). The odds to diagnose adjudicated AHF and echo-bnp AHF were up to four times greater using LDCT (cOR: 3.89 [2.15, 7.06] and cOR: 2.52 [1.45, 4.38], respectively). For each radiologic sign of pulmonary congestion, the LDCT provided superior or equivalent results as the chest radiographs, and the interrater agreement was higher using LDCT (kappa 0.88 [95% CI: 0.81, 0.95] vs. 0.73 [95% CI: 0.63, 0.82]). As first-line imaging modality, LDCT will find one additional adjudicated AHF in 12.5 patients and prevent one false-positive in 20 patients. Similar results were demonstrated for echo-bnp AHF.
CONCLUSIONS: In consecutive dyspnoeic patients admitted to the emergency department, LDCT is significantly better than chest radiographs in detecting pulmonary congestion.

Background: Defining risk factors for long-term comorbidities in patients after neonatal repair of congenital diaphragmatic hernia (CDH) is an important cornerstone of the implementation of targeted longitudinal follow-up programs. Methods: This study systematically assessed serial chest radiographs of 89 patients with left-sided CDH throughout a mean follow-up of 8.2 years. These geometrical variables for the left and right side were recorded: diaphragmatic angle (LDA, RDA), diaphragmatic diameter (LDD, RDD), diaphragmatic height (LDH, RDH), diaphragmatic curvature index (LDCI, RDCI), lower lung diameter (LLLD, RLLD) and thoracic area (LTA, RTA). Results: It was demonstrated that the shape of the diaphragm in patients with large defects systematically differs from that of patients with small defects. Characteristically, patients with large defects present with a smaller LDCI (5.1 vs. 8.4, p < 0.001) at 6 months of age, which increases over time (11.4 vs. 7.0 at the age of 15.5 years, p = 0.727), representing a flattening of the patch and the attached rudimentary diaphragm as the child grows. Conclusions: Multiple variables during early follow-up were significantly associated with comorbidities such as recurrence, scoliotic curves of the spine and a reduced thoracic area. Some geometrical variables may serve as surrogate parameters for disease severity, which is associated with long-term comorbidities.

Artificial Intelligence (AI) methods, particularly Deep Neural Networks (DNNs), have shown great promise in a range of medical imaging tasks. However, the susceptibility of DNNs to producing erroneous outputs under the presence of input noise and variations is of great concern and one of the largest challenges to their adoption in medical settings. Towards addressing this challenge, we explore the robustness of DNNs trained for chest radiograph classification under a range of perturbations reflective of clinical settings. We propose RoMIA, a framework for the creation of Robust Medical Imaging AI models. RoMIA adds three key steps to the model training and deployment flow: (i) Noise-added training, wherein a part of the training data is synthetically transformed to represent common noise sources, (ii) Fine-tuning with input mixing, in which the model is refined with inputs formed by mixing data from the original training set with a small number of images from a different source, and (iii) DCT-based denoising, which removes a fraction of high-frequency components of each image before applying the model to classify it. We applied RoMIA to create six different robust models for classifying chest radiographs using the CheXpert dataset. We evaluated the models on the CheXphoto dataset, which consists of naturally and synthetically perturbed images intended to evaluate robustness. Models produced by RoMIA show 3%-5% improvement in robust accuracy, which corresponds to an average reduction of 22.6% in misclassifications. These results suggest that RoMIA can be a useful step towards enabling the adoption of AI models in medical imaging applications.