Aim Pneumomediastinum (PM) is associated with several etiologies and mechanisms. Although it has been described more than 100 years ago, the literature is limited to small retrospective studies. This study aimed to follow patients with coronavirus disease (COVID-19) that developed PM during hospitalization and describe their clinical and radiological evolution. Methods A prospective cohort was developed with patients with PM, excluding those with aerodigestive trauma, inside a hospital COVID-19 dedicated hospital. Clinical variables including onset of symptoms, hemodynamic instability, associated complications, the need of interventions, and disease course were all recorded. Also, radiological findings such as the presence of the Macklin effect, extension of lung involvement by COVID-19, and characteristics of the PM were analyzed. Results Twenty-one patients with non-traumatic PM were followed, resulting in an overall incidence of 0.5% during the study period. Seven (33%) patients had associated pneumothorax and malignant/tension PM was observed in three (14%) cases. The Macklin effect could be found in 11 patients (52%) and the majority of them had more than 50% of lung involvement due to COVID-19. The mortality rate was 49%; however, no deaths were directly related to the PM. Conclusions PM incidence is probably increased in the severe acute respiratory syndrome caused by COVID-19, especially in those with greater involvement of the lungs, and the Macklin effect may be an important underlying mechanism of this complication. Usually, PM has a benign course, but complications like tension/malignant PM may occur requiring prompt detection and intervention.

Hepatocellular carcinoma (HCC) has become the 4th leading cause of cancer-related deaths, with high social, economical and health implications. Imaging techniques such as multiphase computed tomography (CT) have been successfully used for diagnosis of liver tumors such as HCC in a feasible and accurate way and its interpretation relies mainly on comparing the appearance of the lesions in the different contrast phases of the exam. Recently, some researchers have been dedicated to the development of tools based on machine learning (ML) algorithms, especially by deep learning techniques, to improve the diagnosis of liver lesions in imaging exams. However, the lack of standardization in the naming of the CT contrast phases in the DICOM metadata is a problem for real-life deployment of machine learning tools. Therefore, it is important to correctly identify the exam phase based only on the image and not on the exam metadata, which is unreliable. Motivated by this problem, we successfully created an annotation platform and implemented a convolutional neural network (CNN) to automatically identify the CT scan phases in the HCFMUSP database in the city of São Paulo, Brazil. We improved this algorithm with hyperparameter tuning and evaluated it with cross validation methods. Comparing its predictions with the radiologists annotation, it achieved an accuracy of 94.6%, 98% and 100% in the testing dataset for the slice, volume and exam evaluation, respectively.

OBJECTIVE: To determine the correlation between the two tomographic classifications for coronavirus disease (COVID-19), COVID-19 Reporting and Data System (CORADS) and Radiological Society of North America Expert Consensus Statement on Reporting Chest Computed Tomography (CT) Findings Related to COVID-19 (RSNA), in the Brazilian population and to assess the agreement between reviewers with different experience levels. METHODS: Chest CT images of patients with reverse transcriptase-polymerase chain reaction (RT-PCR)-positive COVID-19 were categorized according to the CORADS and RSNA classifications by radiologists with different levels of experience and who were initially unaware of the RT-PCR results. The inter- and intra-observer concordances for each of the classifications were calculated, as were the concordances between classifications. RESULTS: A total of 100 patients were included in this study. The RSNA classification showed an almost perfect inter-observer agreement between reviewers with similar experience levels, with a kappa coefficient of 0.892 (95% confidence interval [CI], 0.788-0.995). CORADS showed substantial agreement among reviewers with similar experience levels, with a kappa coefficient of 0.642 (95% CI, 0.491-0.793). There was inter-observer variation when comparing less experienced reviewers with more experienced reviewers, with the highest kappa coefficient of 0.396 (95% CI, 0.255-0.588). There was a significant correlation between both classifications, with a Kendall coefficient of 0.899 ( p <0.001) and substantial intra-observer agreement for both classifications. CONCLUSION: The RSNA and CORADS classifications showed excellent inter-observer agreement for reviewers with the same level of experience, although the agreement between less experience reviewers and the reviewer with the most experience was only reasonable. Combined analysis of both classifications with the first RT-PCR results did not reveal any false-negative results for detecting COVID-19 in patients.

Editorial.

To determine the correlation between the two tomographic classifications for coronavirus disease (COVID-19), COVID-19 Reporting and Data System (CORADS) and Radiological Society of North America Expert Consensus Statement on Reporting Chest Computed Tomography (CT) Findings Related to COVID-19 (RSNA), in the Brazilian population and to assess the agreement between reviewers with different experience levels. METHODS: Chest CT images of patients with reverse transcriptase-polymerase chain reaction (RT-PCR)-positive COVID-19 were categorized according to the CORADS and RSNA classifications by radiologists with different levels of experience and who were initially unaware of the RT-PCR results. The inter-and intra-observer concordances for each of the classifications were calculated, as were the concordances between classifications. RESULTS: A total of 100 patients were included in this study. The RSNA classification showed an almost perfect inter-observer agreement between reviewers with similar experience levels, with a kappa coefficient of 0.892 (95% confidence interval [CI], 0.788-0.995). CORADS showed substantial agreement among reviewers with similar experience levels, with a kappa coefficient of 0.642 (95% CI, 0.491-0.793). There was inter-observer variation when comparing less experienced reviewers with more experienced reviewers, with the highest kappa coefficient of 0.396 (95% CI, 0.255-0.588). There was a significant correlation between both classifications, with a Kendall coefficient of 0.899 (po0.001) and substantial intra-observer agreement for both classifications. CONCLUSION: The RSNA and CORADS classifications showed excellent inter-observer agreement for reviewers with the same level of experience, although the agreement between less experience reviewers and the reviewer with the most experience was only reasonable. Combined analysis of both classifications with the first RT-PCR results did not reveal any false-negative results for detecting COVID-19 in patients.