State-of-the-art sleep arousal detection evaluated on a comprehensive clinical dataset.

Aiming to apply automatic arousal detection to support sleep laboratories, we evaluated an optimized, state-of-the-art approach using data from daily work in our university hospital sleep laboratory. Therefore, a machine learning algorithm was trained and evaluated on 3423 polysomnograms of people with various sleep disorders. The model architecture is a U-net that accepts 50 Hz signals as input. We compared this algorithm with models trained on publicly available datasets, and evaluated these models using our clinical dataset, particularly with regard to the effects of different sleep disorders. In an effort to evaluate clinical relevance, we designed a metric based on the error of the predicted arousal index. Our models achieve an area under the precision recall curve (AUPRC) of up to 0.83 and F1 scores of up to 0.81. The model trained on our data showed no age or gender bias and no significant negative effect regarding sleep disorders on model performance compared to healthy sleep. In contrast, models trained on public datasets showed a small to moderate negative effect (calculated using Cohen's d) of sleep disorders on model performance. Therefore, we conclude that state-of-the-art arousal detection on our clinical data is possible with our model architecture. Thus, our results support the general recommendation to use a clinical dataset for training if the model is to be applied to clinical data.

Challenges in Retrieving Patterns from Generic Data Structures in Clinical Systems - A Technical Case Report.

INTRODUCTION: The secondary use of data in clinical environments offers significant opportunities to enhance medical research and practices. However, extracting data from generic data structures, particularly the Entity-Attribute-Value (EAV) model, remains challenging. This study addresses these challenges by developing a methodological approach to convert EAV-based data into a format more suitable for analysis. BACKGROUND: The EAV model is widely used in clinical information systems due to its adaptability, but it often complicates data retrieval for research purposes due to its vertical data structure and dynamic schema. OBJECTIVE: The objective of this study is to develop a methodological approach to address the handling of these generic data structures, Methods: We introduce a five-step methodological approach: 1) understanding the specific clinical processes to determine data collection points and involved roles; 2) analysing the data source to understand the data structure and metadata; 3) reversing a use-case-specific data structure to map the front-end data input to its storage format; 4) analysing the content to identify medical information and establish connections; and 5) managing schema changes to maintain data integrity. RESULTS: Applying this method to the hospital information system has shown that EAV-based data can be converted into a structured format, suitable for research. This conversion reduced data sparsity and improved the manageability of schema changes without affecting other classes of data. CONCLUSION: The developed approach provides a systematic method for handling complex data relationships and maintaining data integrity in clinical systems using EAV models. This approach facilitates the secondary use of clinical data, enhancing its utility for medical research and practice.

Nucleated red blood cells as a prognostic marker for mortality in patients with SARS-CoV-2-induced ARDS: an observational study.

BACKGROUND: The presence of nucleated red blood cells (NRBCs) in the peripheral blood of critically ill patients is associated with poor outcome. Evidence regarding the predictive value of NRBCs in patients with SARS-CoV-2-induced acute respiratory distress syndrome (ARDS) remains elusive. The aim of this study was to evaluate the predictive validity of NRBCs in these patients. METHODS: Daily NRBC values of adult patients with SARS-CoV-2-induced ARDS were assessed and their predictive validity for mortality was statistically evaluated. A cut-off level based on the patient's maximum NRBC value during ICU stay was calculated and further specified according to Youden's method. Based on this cut-off value, further analyses such as logistic regression models and survival were performed. RESULTS: 413 critically ill patients with SARS-CoV-2-induced ARDS were analyzed. Patients who did not survive had significantly higher NRBC values during their ICU stay compared to patients who survived (1090/µl [310; 3883] vs. 140/µl [20; 500]; p < 0.0001). Patients with severe ARDS (n = 374) had significantly higher NRBC values during ICU stay compared to patients with moderate ARDS (n = 38) (490/µl [120; 1890] vs. 30/µl [10; 476]; p < 0.0001). A cut-off level of NRBC ≥ 500/µl was found to best stratify risk and was associated with a longer duration of ICU stay (12 [8; 18] vs. 18 [13; 27] days; p < 0.0001) and longer duration of mechanical ventilation (10 [6; 16] vs. 17 [12; 26] days; p < 0.0001). Logistic regression analysis with multivariate adjustment showed NRBCs ≥ 500/µl to be an independent risk factor of mortality (odds ratio (OR) 4.72; 95% confidence interval (CI) 2.95-7.62, p < 0.0001). Patients with NRBC values below the threshold of 500/µl had a significant survival advantage over those above the threshold (median survival 32 [95% CI 8.7-43.3] vs. 21 days [95% CI 18.2-23.8], log-rank test, p < 0.05). Patients who once reached the NRBC threshold of ≥ 500/µl during their ICU stay had a significantly increased long-term mortality (median survival 489 days, log-rank test, p = 0.0029, hazard ratio (HR) 3.2, 95% CI 1.2-8.5). CONCLUSIONS: NRBCs predict mortality in critically ill patients with SARS-CoV-2-induced ARDS with high prognostic power. Further studies are required to confirm the clinical impact of NRBCs to eventually enhance decision making.

Supporting regional pandemic management by enabling self-service reporting-A case report.

BACKGROUND: The COVID-19 pandemic revealed a need for better collaboration among research, care, and management in Germany as well as globally. Initially, there was a high demand for broad data collection across Germany, but as the pandemic evolved, localized data became increasingly necessary. Customized dashboards and tools were rapidly developed to provide timely and accurate information. In Saxony, the DISPENSE project was created to predict short-term hospital bed capacity demands, and while it was successful, continuous adjustments and the initial monolithic system architecture of the application made it difficult to customize and scale. METHODS: To analyze the current state of the DISPENSE tool, we conducted an in-depth analysis of the data processing steps and identified data flows underlying users' metrics and dashboards. We also conducted a workshop to understand the different views and constraints of specific user groups, and brought together and clustered the information according to content-related service areas to determine functionality-related service groups. Based on this analysis, we developed a concept for the system architecture, modularized the main services by assigning specialized applications and integrated them into the existing system, allowing for self-service reporting and evaluation of the expert groups' needs. RESULTS: We analyzed the applications' dataflow and identified specific user groups. The functionalities of the monolithic application were divided into specific service groups for data processing, data storage, predictions, content visualization, and user management. After composition and implementation, we evaluated the new system architecture against the initial requirements by enabling self-service reporting to the users. DISCUSSION: By modularizing the monolithic application and creating a more flexible system, the challenges of rapidly changing requirements, growing need for information, and high administrative efforts were addressed. CONCLUSION: We demonstrated an improved adaptation towards the needs of various user groups, increased efficiency, and reduced burden on administrators, while also enabling self-service functionalities and specialization of single applications on individual service groups.

Hyperspectral Imaging during Normothermic Machine Perfusion-A Functional Classification of Ex Vivo Kidneys Based on Convolutional Neural Networks.

Facing an ongoing organ shortage in transplant medicine, strategies to increase the use of organs from marginal donors by objective organ assessment are being fostered. In this context, normothermic machine perfusion provides a platform for ex vivo organ evaluation during preservation. Consequently, analytical tools are emerging to determine organ quality. In this study, hyperspectral imaging (HSI) in the wavelength range of 550-995 nm was applied. Classification of 26 kidneys based on HSI was established using KidneyResNet, a convolutional neural network (CNN) based on the ResNet-18 architecture, to predict inulin clearance behavior. HSI preprocessing steps were implemented, including automated region of interest (ROI) selection, before executing the KidneyResNet algorithm. Training parameters and augmentation methods were investigated concerning their influence on the prediction. When classifying individual ROIs, the optimized KidneyResNet model achieved 84% and 62% accuracy in the validation and test set, respectively. With a majority decision on all ROIs of a kidney, the accuracy increased to 96% (validation set) and 100% (test set). These results demonstrate the feasibility of HSI in combination with KidneyResNet for non-invasive prediction of ex vivo kidney function. This knowledge of preoperative renal quality may support the organ acceptance decision.