Proposal and definition of an intelligent decision-support system based on deep learning techniques for the management of possible COVID-19 cases in patients attending emergency departments
The COVID-19 pandemic drastically transformed the integration of technology into medicine, testing the ability of health systems to make quick and effective decisions. This has been especially noticeable in emergency departments, which were overwhelmed by the massive influx of patients. In this cont...
| Autores: | , , , , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Fecha de publicación: | 2024 |
| País: | España |
| Institución: | Universidad Autónoma de Madrid |
| Repositorio: | Biblos-e Archivo. Repositorio Institucional de la UAM |
| Idioma: | inglés |
| OAI Identifier: | oai:repositorio.uam.es:10486/721308 |
| Acceso en línea: | http://hdl.handle.net/10486/721308 https://dx.doi.org/10.1109/ACCESS.2024.3424907 |
| Access Level: | acceso abierto |
| Palabra clave: | Artificial intelligence convolutional neural networks decision making decision support systems deep learning Informática Medicina |
| Sumario: | The COVID-19 pandemic drastically transformed the integration of technology into medicine, testing the ability of health systems to make quick and effective decisions. This has been especially noticeable in emergency departments, which were overwhelmed by the massive influx of patients. In this context, this article presents the design, development, and proof of concept of a new intelligent decision support system applied to the management of patients suspected of having COVID-19 upon their arrival at an emergency department. To achieve this, starting from our proprietary database of chest X-rays (CXRs) collected at the Ribera Povisa Hospital, two modules based on the use of convolutional neural networks (CNNs) were sequentially run. The first was based on the DenseNet-121 model to identify whether a pneumonia condition was presented in the CXR, while the second was based on the COVID-Net CXR-S model and aimed to quantify the severity of airspace opacity in the CXR on a scale 0-24. Thus, based on this architecture, it will be possible to make predictions based on the CXR of new patients that, after interpretation, might allow physicians to determine whether cases are high-risk and, for example, should be admitted to the intensive care unit. Although the results we obtained were encouraging, it is important to note that this proposal is still at a conceptual stage of development and so future work will be required to validate it in real environments and develop techniques that can help explain its results |
|---|