Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room

Objective@#Recent studies have suggested that deep-learning models can satisfactorily assist in fracture diagnosis. We aimed to evaluate the performance of two of such models in wrist fracture detection. @*Methods@#We collected image data of patients who visited with wrist trauma at the emergency department. A dataset extracted from January 2018 to May 2020 was split into training (90%) and test (10%) datasets, and two types of convolutional neural networks (i.e., DenseNet-161 and ResNet-152) were trained to detect wrist fractures. Gradient-weighted class activation mapping was used to highlight the regions of radiograph scans that contributed to the decision of the model. Performance of the convolutional neural network models was evaluated using the area under the receiver operating characteristic curve. @*Results@#For model training, we used 4,551 radiographs from 798 patients and 4,443 radiographs from 1,481 patients with and without fractures, respectively. The remaining 10% (300 radiographs from 100 patients with fractures and 690 radiographs from 230 patients without fractures) was used as a test dataset. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of DenseNet-161 and ResNet-152 in the test dataset were 90.3%, 90.3%, 80.3%, 95.6%, and 90.3% and 88.6%, 88.4%, 76.9%, 94.7%, and 88.5%, respectively. The area under the receiver operating characteristic curves of DenseNet-161 and ResNet-152 for wrist fracture detection were 0.962 and 0.947, respectively. @*Conclusion@#We demonstrated that DenseNet-161 and ResNet-152 models could help detect wrist fractures in the emergency room with satisfactory performance.

Application of convolutional neural networks for distal radio-ulnar fracture detection on plain radiographs in the emergency room

Objective@#Recent studies have suggested that deep-learning models can satisfactorily assist in fracture diagnosis. We aimed to evaluate the performance of two of such models in wrist fracture detection. @*Methods@#We collected image data of patients who visited with wrist trauma at the emergency department. A dataset extracted from January 2018 to May 2020 was split into training (90%) and test (10%) datasets, and two types of convolutional neural networks (i.e., DenseNet-161 and ResNet-152) were trained to detect wrist fractures. Gradient-weighted class activation mapping was used to highlight the regions of radiograph scans that contributed to the decision of the model. Performance of the convolutional neural network models was evaluated using the area under the receiver operating characteristic curve. @*Results@#For model training, we used 4,551 radiographs from 798 patients and 4,443 radiographs from 1,481 patients with and without fractures, respectively. The remaining 10% (300 radiographs from 100 patients with fractures and 690 radiographs from 230 patients without fractures) was used as a test dataset. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of DenseNet-161 and ResNet-152 in the test dataset were 90.3%, 90.3%, 80.3%, 95.6%, and 90.3% and 88.6%, 88.4%, 76.9%, 94.7%, and 88.5%, respectively. The area under the receiver operating characteristic curves of DenseNet-161 and ResNet-152 for wrist fracture detection were 0.962 and 0.947, respectively. @*Conclusion@#We demonstrated that DenseNet-161 and ResNet-152 models could help detect wrist fractures in the emergency room with satisfactory performance.

Validation and modification of HEART score components for patients with chest pain in the emergency department

Objective@#This study aimed to clarify the relative prognostic value of each History, Electrocardiography, Age, Risk Factors, and Troponin (HEART) score component for major adverse cardiac events (MACE) within 3 months and validate the modified HEART (mHEART) score. @*Methods@#This study evaluated the HEART score components for patients with chest symptoms visiting the emergency department from November 19, 2018 to November 19, 2019. All components were evaluated using logistic regression analysis and the scores for HEART, mHEART, and Thrombolysis in Myocardial Infarction (TIMI) were determined using the receiver operating characteristics curve. @*Results@#The patients were divided into a derivation (809 patients) and a validation group (298 patients). In multivariate analysis, age did not show statistical significance in the detection of MACE within 3 months and the mHEART score was calculated after omitting the age component. The areas under the receiver operating characteristics curves for HEART, mHEART and TIMI scores in the prediction of MACE within 3 months were 0.88, 0.91, and 0.83, respectively, in the derivation group; and 0.88, 0.91, and 0.81, respectively, in the validation group. When the cutoff value for each scoring system was determined for the maintenance of a negative predictive value for a MACE rate >99%, the mHEART score showed the highest sensitivity, specificity, positive predictive value, and negative predictive value (97.4%, 54.2%, 23.7%, and 99.3%, respectively). @*Conclusion@#Our study showed that the mHEART score better detects short-term MACE in high-risk patients and ensures the safe disposition of low-risk patients than the HEART and TIMI scores.

Central Venous Catheterization before Versus after Computed Tomography in Hemodynamically Unstable Patients with Major Blunt Trauma: Clinical Characteristics and Factors for Decision Making

PURPOSE@#When hemodynamically unstable patients with blunt major trauma arrive at the emergency department (ED), the safety of performing early whole-body computed tomography (WBCT) is concerning. Some clinicians perform central venous catheterization (CVC) before WBCT (pre-computed tomography [CT] group) for hemodynamic stabilization. However, as no study has reported the factors affecting this decision, we compared clinical characteristics and outcomes of the pre- and post-CT groups and determined factors affecting this decision.@*METHODS@#This retrospective study included 70 hemodynamically unstable patients with chest or/and abdominal blunt injury who underwent WBCT and CVC between March 2013 and November 2017.@*RESULTS@#Univariate analysis revealed that the injury severity score, intubation, pulse pressure, focused assessment with sonography in trauma positivity score, and pH were different between the pre-CT (34 patients, 48.6%) and post-CT (all, p < 0.05) groups. Multivariate analysis revealed that injury severity score (ISS) and intubation were factors affecting the decision to perform CVC before CT (p=0.003 and p=0.043). Regarding clinical outcomes, the interval from ED arrival to CT (p=0.011) and definite bleeding control (p=0.038), and hospital and intensive care unit lengths of stay (p=0.018 and p=0.053) were longer in the pre-CT group than in the post-CT group. Although not significant, the pre-CT group had lower survival rates at 24 hours and 28 days than the post-CT group (p=0.168 and p=0.226).@*CONCLUSIONS@#Clinicians have a tendency to perform CVC before CT in patients with blunt major trauma and high ISS and intubation.

The Clinical Value of TnI and BNP for the Evaluation of Sepsis-related Cardiac Dysfunction in the Emergency Department

PURPOSE: Acute cardiac dysfunction is a well recognized manifestation of organ failure in severe sepsis and septic shock. Although echocardiography is the golden standard for the evaluation of cardiac dysfunction, it is difficult to use in the emergency department (ED). The purpose of this study was to determine the availability of cardiac biomarkers for the estimation of cardiac dysfunction in septic shock patients. METHODS: All study subjects included consecutive patients with septic shock diagnosed in the ED and treated with an algorithm of early goal-directed therapy between January 2011 and June 2012. We enrolled patients measured for cardiac biomarkers and performed echocardiography within 24 hours. We divided patients into two groups based on the occurrence of left ventricular dysfunction (defined as an ejection fraction< or =40%) and compared serum levels of troponin-I (TnI) and B-type natriuretic peptide (BNP) between the two groups. The area under the receiver operating characteristic (ROC) curve was used to compare the diagnostic ability of TnI and BNP. RESULTS: A total of 127 patients with septic shock and evaluated for cardiac dysfunction were enrolled in this study. TnI and BNP were significantly higher in the left ventricular dysfunction group group (4.2+/-9.0 vs. 0.6+/-1.8 ng/mL, respectively, p<0.05) compared with the non-dysfunction group (1087.6+/-680.1 vs. 633.2+/-859.1 pg/mL, respectively, p<0.05). However, in the ROC curve for predicting left ventricular dysfunction, the area under the curves of TnI and BNP, respectively, were 0.631(95% CI 0.473-0.788, p=0.103) and 0.704 (95% CI 0.552-0.856, p=0.011). TnI and BNP showed a 84.6% negative predictive value. CONCLUSION: Although TnI and BNP were significantly higher in septic shock patients with cardiac dysfunction but demonstrated limited accuracy compared to echocardiography. However, TnI and BNP have high negative predictive value in septic shock patients for the evaluation of cardiac dysfunction. Therefore they could serve as a valuable supplement for the detection of cardiac dysfunction.