Visual classification of pressure injury stages for nurses: A deep learning model applying modern convolutional neural networks.

AIMS: To develop a deep learning model for pressure injury stages classification based on real-world photographs and compare its performance with that of clinical nurses to seek the opportunity of its application in clinical settings. DESIGN: This was a retrospective observational study using a deep learning model. REVIEW METHODS: A plastic surgeon and two wound care nurses labelled a set of pressure injury images. We applied several modern Convolutional Neural Networks architectures and compared the performances with those of clinical nurses. DATA SOURCES: We retrospectively analysed the electronic medical records of hospitalized patients between January 2019 and April 2021. RESULTS: A set of 2464 pressure injury images were compiled and analysed. Using EfficientNet, in classifying pressure injury images, the macro F1-score was calculated to be 0.8941, and the average performance of two experienced nurses was reported as 0.8781. CONCLUSION: A deep learning model for classifying pressure injury images by stages was successfully developed, and the performance of the model was compared with that of experienced nurses. The classification model developed in this study is expected to help less-experienced nurses or those working in under-resourced healthcare settings determine the stages of pressure injury. IMPACT: Our deep learning model can minimize discrepancies in nurses' assessment of classifying pressure injury stages. Follow-up studies on improving the performance of deep learning models using modern techniques and clinical usability will lead to improved quality of care among patients with pressure injury. NO PATIENT OR PUBLIC CONTRIBUTION: Patients or the public were not involved in our research's design, conduct, reporting or dissemination plans because this was a retrospective study that used electronic medical records.

Multi-center validation of machine learning model for preoperative prediction of postoperative mortality.

Accurate prediction of postoperative mortality is important for not only successful postoperative patient care but also for information-based shared decision-making with patients and efficient allocation of medical resources. This study aimed to create a machine-learning prediction model for 30-day mortality after a non-cardiac surgery that adapts to the manageable amount of clinical information as input features and is validated against multi-centered rather than single-centered data. Data were collected from 454,404 patients over 18 years of age who underwent non-cardiac surgeries from four independent institutions. We performed a retrospective analysis of the retrieved data. Only 12-18 clinical variables were used for model training. Logistic regression, random forest classifier, extreme gradient boosting (XGBoost), and deep neural network methods were applied to compare the prediction performances. To reduce overfitting and create a robust model, bootstrapping and grid search with tenfold cross-validation were performed. The XGBoost method in Seoul National University Hospital (SNUH) data delivers the best performance in terms of the area under receiver operating characteristic curve (AUROC) (0.9376) and the area under the precision-recall curve (0.1593). The predictive performance was the best when the SNUH model was validated with Ewha Womans University Medical Center data (AUROC, 0.941). Preoperative albumin, prothrombin time, and age were the most important features in the model for each hospital. It is possible to create a robust artificial intelligence prediction model applicable to multiple institutions through a light predictive model using only minimal preoperative information that can be automatically extracted from each hospital.