Search

Radiologists face high burnout rates, partially due to the increasing volume of Chest X-rays (CXRs) requiring interpretation and reporting. Automated CXR report generation holds promise for reducing this burden and improving patient care. While current models show potential, their diagnostic accuracy is limited. Our proposed CXR report generator integrates elements of the radiologist workflow and introduces a novel reward for reinforcement learning. Our approach leverages longitudinal data from a patient’s prior CXR study and effectively handles cases where no prior study exists, thus mirroring the radiologist’s workflow. In contrast, existing models typically lack this flexibility, often requiring prior studies for the model to function optimally. Our approach also incorporates all CXRs from a patient’s study and distinguishes between report sections through section embeddings. Our reward for reinforcement learning leverages CXR-BERT, which forces our model to learn the clinical semantics of radiology reporting. We conduct experiments on publicly available datasets — MIMIC-CXR and Open-i IU X-ray — with metrics shown to more closely correlate with radiologists’ assessment of reporting. Results from our study demonstrate that the proposed model generates reports that are more aligned with radiologists’ reports than state-of-the-art models, such as those utilising large language models, reinforcement learning, and multi-task learning. The proposed model improves the diagnostic accuracy of CXR report generation, which could one day reduce radiologists’ workload and enhance patient care. Our Hugging Face checkpoint (https://huggingface.co/aehrc/cxrmate) and code (https://github.com/aehrc/cxrmate) are publicly available.

Abstract Background There is an increasing amount of literature describing the pathogenesis of coronavirus disease 2019 (COVID-19) pneumonia and its associated complications. Historically, a small pneumothorax has been shown to be successfully treated without chest tube insertion, but this management has yet to be proven in COVID-19 pneumonia patients. In addition, pneumothorax in an intubated patient with high positive end-expiratory pressure (PEEP) provides additional uncertainty with pursuing non-operative management. Case presentation In this series we report four cases of patients with respiratory distress who tested positive for COVID-19 via nasopharyngeal swab and developed ventilator-induced pneumothoraces which were successfully managed with observation alone. Conclusions Management of patients with COVID-19 pneumonia on positive pressure ventilation who develop small stable pneumothoraces can be safely observed without chest tube insertion.

The standard methods of applying iterative techniques do not apply when the nonlinear term is neither monotonic (corresponding to an increasing or decreasing operator) nor Lipshitz (corresponding to a condensing operator). However, by applying the Layered Compression–Expansion Theorem in conjunction with an alternative inversion technique, we show how one can apply monotonicity techniques to a right focal boundary value problem.

Limited attention has been afforded to mapping the ‘landscape’ of South African computing research. Prior studies have considered singular sub-disciplines, publications, or publication types. Given the growing prominence of computing disciplines, it is necessary to identify the patterns of research production, publication, collaboration, and impact of South African computing research. This study presents a scientometric investigation in this regard. Through the analysis of data accessed from the Scopus citation enhanced bibliographic database, the investigation presents findings in relation to annual research production, institutional differences in outputs, topics, collaboration, and citation impact. While characterised by institutional differences, over the period considered, South African computing research output has increased at a greater rate than that of South African research at large. Additionally, despite accounting for a greater proportion of all outputs, conference papers account for a smaller proportion of citations relative to journal articles or book chapters. Corresponding to previous investigations, there exists a tendency towards applied computing topics in contrast to more theoretical topics. Finally, the collaboration network was shown to be particularly de-centralised with many researchers clustered around institutions. The findings are of interest to all researchers conducting computing or related research in South Africa.