Objective: In the complex landscape of ICU operations, accurate discharge decisions are crucial yet challenging, as premature discharge risks readmission and mortality while prolonged stays consume resources and heighten infection risk. The objective of this work is to develop a deep learning-based Discharge Readiness Score (DRS) model using minimal clinical features to predict ICU discharge readiness, and to highlight its application in estimating excess ICU stays for resource optimization. Methods and procedures: We utilized nearly 1.8 million ICU patient-stays from 2007-2023 across 300 US hospitals in the Philips eICU database. Six readily available features (age, mean arterial pressure, systolic pressure, heart rate, respiratory rate, and Glasgow Coma Scale) were used as inputs. A 5-layer neural network predicted patient mortality within 48 hours post-ICU discharge as a proxy for discharge readiness. The model was trained on 80% of data, validated on 10%, and tested on 10% (approximately 180,000 patients). We applied the model hourly to estimate excess ICU stays, defining excess stay as the time patients remained at low risk but continued in ICU. Results: The model achieved an AUC of 0.93 on the test set. Performance was consistent across years, ethnicities, ICU types, and admission groups. Using the model, we found that about 22% of patients had excess ICU time, with a median of 16 hours. The analysis highlighted trends over time and across ICU types, providing insights into resource utilization. Conclusion: The DRS model effectively predicts ICU discharge readiness using minimal features and can estimate excess ICU stays, aiding resource optimization. Clinical Impact- The model offers a practical tool for ICU discharge planning and resource utilization analysis, potentially improving patient outcomes and ICU operations

OBJECTIVES:To determine the impact of critical care telemedicine (tele-critical care [TCC]) implementation duration on clinical outcomes: ICU mortality, ICU length of stay (LOS), and mechanical ventilation utilization. DESIGN:Retrospective cohort study. SETTING:Thirty-five U.S. Department of Veterans Affairs (VA) hospitals (444 ICU beds) that used TCC between 2012 and 2020. PATIENTS:One hundred ninety-three thousand three hundred sixty-seven patient stays meeting specific inclusion criteria from 2012 to 2020 were included in the study. INTERVENTIONS:Critical care telemedicine (TCC) implementation. MEASUREMENTS AND MAIN RESULTS:The standardized ICU mortality rate was calculated by comparing patient outcomes to expected outcomes, utilizing critical care prediction models. ICU LOS was standardized for illness severity and case mix. The rate of invasive mechanical ventilation was analyzed, comparing ventilator days against predicted values. Longer TCC utilization was linked with a trend toward lower standardized ICU mortality rates, with statistically significant reductions after a 5-year period. ICU LOS also showed a significant decrease with prolonged TCC deployment. While the rate of invasive mechanical ventilation declined over time, it was not significantly related to the TCC deployment duration. CONCLUSIONS:Extended TCC implementation improves ICU mortality rates and reduces ICU LOS. Longer TCC deployment has clear benefits on patient outcomes in the VA healthcare system. Further research should explore long-term effects and factors influencing TCC adoption.

Understanding cellular disease processes like cancer is key for improving diagnosis and treatment. Single-cell RNA sequencing (scRNA-seq) enables modeling transitions between normal and diseased cellular states in complex tissues. However, interpolating between healthy and diseased states in high-dimensional scRNA-seq data poses computational challenges. We use the Class-Informed Variational Autoencoder (CI-VAE), a generative AI model, to learn low-dimensional cell-type-specific representations from scRNA-seq data. During inference, CI-VAE interpolates between normal and diseased cells, robustly predicting cell-type-specific gene expression trajectories from healthy to disease states. Applied to colon cancer data, CI-VAE closely predicted observed transitions by generating synthetic gene expression changes associated with cancer progression for each cell type, potentially offering insights into underlying molecular mechanisms for disease understanding, biomarker discovery, and targeted therapy design.

We propose a new variant of Variational Autoencoder (VAE), Class-Informed VAE (CI-VAE), that enables interpolation between arbitrary pairs of observations of the same class. CI-VAE combines the general VAE architecture with a linear discriminator layer on the latent space to enforce the construction of a latent space where observations from different classes are linearly separable. This allows for robust latent-space linear traversal and data generation between two arbitrary observations of the same class, which has potential applications in science and engineering. One specific application is to enhance understanding of the biological processes involving the development of diseases or cancer from healthy cells. We demonstrate the effectiveness of CI-VAE on the MNIST dataset of handwritten digits, showing that it significantly improves class-specific linear traversal and data augmentation compared to VAE while maintaining comparable reconstruction error. We also apply CI-VAE to a study of colon cancer single-cell genomics data, showing that interpolation between normal cells and tumor cells using CI-VAE may enhance our understanding of the mechanism of cancer development.

The goal of this work is to develop a Machine Learning model to predict the need for both invasive and non-invasive mechanical ventilation in intensive care unit (ICU) patients. Using the Philips eICU Research Institute (ERI) database, 2.6 million ICU patient data from 2010 to 2019 were analyzed. This data was randomly split into training (63%), validation (27%), and test (10%) sets. Additionally, an external test set from a single hospital from the ERI database was employed to assess the model's generalizability. Model performance was determined by comparing the model probability predictions with the actual incidence of ventilation use, either invasive or non-invasive. The model demonstrated a prediction performance with an AUC of 0.921 for overall ventilation, 0.937 for invasive, and 0.827 for non-invasive. Factors such as high Glasgow Coma Scores, younger age, lower BMI, and lower PaCO2 were highlighted as indicators of a lower likelihood for the need for ventilation. The model can serve as a retrospective benchmarking tool for hospitals to assess ICU performance concerning mechanical ventilation necessity. It also enables analysis of ventilation strategy trends and risk-adjusted comparisons, with potential for future testing as a clinical decision tool for optimizing ICU ventilation management.

OBJECTIVES:Enhancing critical care efficacy involves evaluating and improving system functioning. Benchmarking, a retrospective comparison of results against standards, aids risk-adjusted assessment and helps healthcare providers identify areas for improvement based on observed and predicted outcomes. The last two decades have seen the development of several models using machine learning (ML) for clinical outcome prediction. ML is a field of artificial intelligence focused on creating algorithms that enable computers to learn from and make predictions or decisions based on data. This narrative review centers on key discoveries and outcomes to aid clinicians and researchers in selecting the optimal methodology for critical care benchmarking using ML.METHODS:We used PubMed to search the literature from 2003 to 2023 regarding predictive models utilizing ML for mortality (592 articles), length of stay (143 articles), or mechanical ventilation (195 articles). We supplemented the PubMed search with Google Scholar, making sure relevant articles were included. Given the narrative style, papers in the cohort were manually curated for a comprehensive reader perspective.RESULTS:Our report presents comparative results for benchmarked outcomes and emphasizes advancements in feature types, preprocessing, model selection, and validation. It showcases instances where ML effectively tackled critical care outcome-prediction challenges, including nonlinear relationships, class imbalances, missing data, and documentation variability, leading to enhanced results.CONCLUSIONS:Although ML has provided novel tools to improve the benchmarking of critical care outcomes, areas that require further research include class imbalance, fairness, improved calibration, generalizability, and long-term validation of published models.

We proposed Class-Informed Variational Autoencoder (CI-VAE) to enable interpolation between arbitrary pairs of observations of the same class. CI-VAE combines the general VAE architecture with a linear discriminator layer on the latent space to enforce the construction of a latent space such that observations from different classes are linearly separable. In conventional VAEs, class overlapping on the latent space usually occurs. However, in CI-VAE, the enforced linear separability of classes on the latent space allows for robust latent-space linear traversal and data generation between two arbitrary observations of the same class. Class-specific data interpolation has extensive potential applications in science, particularly in biology, such as uncovering the biological trajectory of diseases or cancer. We used the MNIST dataset of handwritten digits as a case study to compare the performance of CI-VAE and VAE in class-specific data augmentation. We showed that CI-VAE significantly improved class-specific linear traversal and data augmentation compared with VAE while maintaining comparable reconstruction error. In a study of Colon cancer genomics data, we showed that the interpolation between normal cells and tumor cells using CI-VAE may enhance our understanding of cancer development.

Predicting the duration of ventilation in the ICU helps in assessing the risk of ventilator-induced lung injury, ensuring sufficient oxygenation, and optimizing resource allocation. Prior models provided a prediction of total duration without distinguishing between invasive and non-invasive ventilation. This work proposes two independent gradient boosting regression models for predicting the duration of invasive and non-invasive ventilation based on commonly available ICU features. These models are trained on 2.6 million patient stays across 350 US hospitals between 2010 to 2019. The mean absolute error (MAE) for the prediction of duration was 2.08 days for invasive ventilation and 0.36 days for non-invasive ventilation. The total ventilation duration predicted by our model had MAE of 2.38 days, which outperformed the gold standard (APACHE) with MAE of 3.02 days. The feature importance analysis of the trained models showed that, for invasive ventilation, high average heart rate, diagnosis of respiratory infection and admissions from locations other than the operating room were associated with longer ventilation durations. For non-invasive ventilation, higher respiratory rates and having any GCS measurement were associated with longer durations.

PURPOSE The availability of increasing volumes of multiomics, imaging, and clinical data in complex diseases such as cancer opens opportunities for the formulation and development of computational imaging genomics methods that can link multiomics, imaging, and clinical data. METHODS Here, we present the Imaging-AMARETTO algorithms and software tools to systematically interrogate regulatory networks derived from multiomics data within and across related patient studies for their relevance to radiography and histopathology imaging features predicting clinical outcomes. RESULTS To demonstrate its utility, we applied Imaging-AMARETTO to integrate three patient studies of brain tumors, specifically, multiomics with radiography imaging data from The Cancer Genome Atlas (TCGA) glioblastoma multiforme (GBM) and low-grade glioma (LGG) cohorts and transcriptomics with histopathology imaging data from the Ivy Glioblastoma Atlas Project (IvyGAP) GBM cohort. Our results show that Imaging-AMARETTO recapitulates known key drivers of tumor-associated microglia and macrophage mechanisms, mediated by STAT3, AHR, and CCR2, and neurodevelopmental and stemness mechanisms, mediated by OLIG2. Imaging-AMARETTO provides interpretation of their underlying molecular mechanisms in light of imaging biomarkers of clinical outcomes and uncovers novel master drivers, THBS1 and MAP2, that establish relationships across these distinct mechanisms. CONCLUSION Our network-based imaging genomics tools serve as hypothesis generators that facilitate the interrogation of known and uncovering of novel hypotheses for follow-up with experimental validation studies. We anticipate that our Imaging-AMARETTO imaging genomics tools will be useful to the community of biomedical researchers for applications to similar studies of cancer and other complex diseases with available multiomics, imaging, and clinical data.

Previous studies suggest that functional ankle instability (FAI) may be associated with deficits in the ability to sense muscle forces. We tested individuals with FAI to determine if they have reduced ability to control ankle muscle forces, which is a function of force sense. Our test was performed isometrically to minimize the involvement of joint position sense and kinesthesia. A FAI group and a control group were recruited to perform an ankle force control task using a platform-based ankle robot. They were asked to move a cursor to hit 24 targets as accurately and as fast as possible in a virtual maze. The cursor movement was based on the direction and magnitude of the forces applied to the robot. Participants underwent three conditions: pre-test (baseline), practice (skill acquisition), and post-test (post skill acquisition). The force control ability was quantified based on the accuracy performance during the task. The accuracy performance was negatively associated with the collision count of the cursor with the maze wall. The FAI group showed reduced ability to control ankle muscle forces compared to the control group in the pre-test condition, but the difference became non-significant in the post-test condition after practice. The change in performance before and after practice may be due to different degrees of reliance on force sense.

Ankle joint plays a critical role in daily activities involving interactions with environment using force and position control. Neuromechanical dysfunctions (e.g. due to stroke or brain injury), therefore, have a major impact on individuals’ quality of life. The effective design of neurorehabilitation protocols for robotic rehabilitation platforms, relies on understanding the control characteristics of the ankle joint in interaction with external environment using force and position. This is particularly of interest since the findings in upper-limb may not be generalizable to the lower-limb. This study aimed to characterize the skilled performance of ankle joint in visuomotor position and force control. A 2-degree of freedom (DOF) robotic footplate was used to measure individuals’ force and position. Healthy individuals (n = 27) used ankle force or position for point-to-point and tracking control tasks in 1-DOF and 2-DOF virtual game environments. Subjects’ performance was quantified as a function of accuracy and completion time. While the performance measures in 1-DOF control tasks were comparable, the subjects’ performance in 2-DOF tasks was significantly better with position control. Subjective questionnaires on the perceived difficulty matched the objective experimental results; suggesting that the poor performance in force control was not due to experimental setup or fatigue but can be attributed to the different levels of challenge needed in neural control. It is inferred that in visuomotor coordination, the neuromuscular specialization of ankle provides better control over position rather than force. These findings can inform the design of neuro-rehabilitation platforms, selection of effective tasks, and therapeutic protocols.

BACKGROUND & AIMS:The mechanisms of hepatitis C virus (HCV) infection, liver disease progression, and hepatocarcinogenesis are only partially understood. We performed genomic, proteomic, and metabolomic analyses of HCV-infected cells and chimeric mice to learn more about these processes. METHODS:Huh7.5.1dif (hepatocyte-like cells) were infected with culture-derived HCV and used in RNA sequencing, proteomic, metabolomic, and integrative genomic analyses. uPA/SCID (urokinase-type plasminogen activator/severe combined immunodeficiency) mice were injected with serum from HCV-infected patients; 8 weeks later, liver tissues were collected and analyzed by RNA sequencing and proteomics. Using differential expression, gene set enrichment analyses, and protein interaction mapping, we identified pathways that changed in response to HCV infection. We validated our findings in studies of liver tissues from 216 patients with HCV infection and early-stage cirrhosis and paired biopsy specimens from 99 patients with hepatocellular carcinoma, including 17 patients with histologic features of steatohepatitis. Cirrhotic liver tissues from patients with HCV infection were classified into 2 groups based on relative peroxisome function; outcomes assessed included Child-Pugh class, development of hepatocellular carcinoma, survival, and steatohepatitis. Hepatocellular carcinomas were classified according to steatohepatitis; the outcome was relative peroxisomal function. RESULTS:We quantified 21,950 messenger RNAs (mRNAs) and 8297 proteins in HCV-infected cells. Upon HCV infection of hepatocyte-like cells and chimeric mice, we observed significant changes in levels of mRNAs and proteins involved in metabolism and hepatocarcinogenesis. HCV infection of hepatocyte-like cells significantly increased levels of the mRNAs, but not proteins, that regulate the innate immune response; we believe this was due to the inhibition of translation in these cells. HCV infection of hepatocyte-like cells increased glucose consumption and metabolism and the STAT3 signaling pathway and reduced peroxisome function. Peroxisomes mediate β-oxidation of very long-chain fatty acids; we found intracellular accumulation of very long-chain fatty acids in HCV-infected cells, which is also observed in patients with fatty liver disease. Cells in livers from HCV-infected mice had significant reductions in levels of the mRNAs and proteins associated with peroxisome function, indicating perturbation of peroxisomes. We found that defects in peroxisome function were associated with outcomes and features of HCV-associated cirrhosis, fatty liver disease, and hepatocellular carcinoma in patients. CONCLUSIONS:We performed combined transcriptome, proteome, and metabolome analyses of liver tissues from HCV-infected hepatocyte-like cells and HCV-infected mice. We found that HCV infection increases glucose metabolism and the STAT3 signaling pathway and thereby reduces peroxisome function; alterations in the expression levels of peroxisome genes were associated with outcomes of patients with liver diseases. These findings provide insights into liver disease pathogenesis and might be used to identify new therapeutic targets.

Challenge and threat are biopsychological responses following an individual's evaluation of task demands relative to his or her available resources to cope with these demands. In this study, we aimed to investigate individual and group variation in physiological responding across a series of motivated performance tasks of varying difficulty. We specifically tested three hypotheses: (H1) individuals will express different sets of physiological patterns (features) across tasks of varying difficulty; (H2) there willbe groups of individuals who share common salient physiological features that dominate within-individual differentiation in physiological responding across tasks of varying difficulty; and (H3) the accuracy of predicting self-reported judgments of challenge and threat across individuals will be higher within each group with shared salient physiological features than across all groups or the entire sample. To test these hypotheses, we developed an integrated analytic framework for multimodal physiological data analysis. We employed data from an existing experiment in which participants completed three mental arithmetic tasks of increasing difficulty during which we collected different modalities of physiological data. Analyses revealed three groups of participants who shared common features that best differentiated their within-individual physiological response patterns across tasks. Support vector machine (SVM) classifiers were then trained using both shared features within each group and all computed features to predict challenge vs. threat states. Our results showed that, within-group classification model using person-specific features achieved higher self-report prediction accuracy comparing to the alternative model trained on data from all participants without feature selection.

Electrocardiogram, electrodermal activity, electromyogram, continuous blood pressure, and impedance cardiography are among the most commonly used peripheral physiological signals (biosignals) in psychological studies and healthcare applications, including health tracking, sleep quality assessment, disease early-detection/diagnosis, and understanding human emotional and affective phenomena. This paper presents the development of a biosignal-specific processing toolbox (Bio-SP tool) for preprocessing and feature extraction of these physiological signals according to the state-of-the-art studies reported in the scientific literature and feedback received from the field experts. Our open-source Bio-SP tool is intended to assist researchers in affective computing, digital and mobile health, and telemedicine to extract relevant physiological patterns (i.e., features) from these biosignals semi-automatically and reliably. In this paper, we describe the successful algorithms used for signal-specific quality checking, artifact/noise filtering, and segmentation along with introducing features shown to be highly relevant to category discrimination in several healthcare applications (e.g., discriminating patterns associated with disease versus non-disease). Further, the Bio-SP tool is a publicly-available software written in MATLAB with a user-friendly graphical user interface (GUI), enabling future crowd-sourced modification to these tools. The GUI is compatible with MathWorks Classification Learner app for inference model development, such as model training, cross-validation scheme farming, and classification result computation.

In this paper, we present a multimodal approach to simultaneously analyze facial movements and several peripheral physiological signals to decode individualized affective experiences under positive and negative emotional contexts, while considering their personalized resting dynamics. We propose a person-specific recurrence network to quantify the dynamics present in the person's facial movements and physiological data. Facial movement is represented using a robust head vs. 3D face landmark localization and tracking approach, and physiological data are processed by extracting known attributes related to the underlying affective experience. The dynamical coupling between different input modalities is then assessed through the extraction of several complex recurrent network metrics. Inference models are then trained using these metrics as features to predict individual's affective experience in a given context, after their resting dynamics are excluded from their response. We validated our approach using a multimodal dataset consists of (i) facial videos and (ii) several peripheral physiological signals, synchronously recorded from 12 participants while watching 4 emotion-eliciting video-based stimuli. The affective experience prediction results signified that our multimodal fusion method improves the prediction accuracy up to 19% when compared to the prediction using only one or a subset of the input modalities. Furthermore, we gained prediction improvement for affective experience by considering the effect of individualized resting dynamics.

Quasi-static, pulmonary pressure-volume (P-V) curves are combined with a respiratory system model to analyze characteristics of patients with acute respiratory distress syndrome (ARDS). It is shown that there exist distinct differences between healthy- and injured-respiratory system in the order of magnitudes of parameters of their P-V model equation. Four stages of ARDS (Early ARDS, Deep knee, Advanced ARDS and Baby lung) are defined quantitatively in terms of these parameters.

Quasi-static, pulmonary pressure-volume (P-V) curves were combined with a respiratory system model to analyze tidal pressure cycles, simulating mechanical ventilation of patients with acute respiratory distress syndrome (ARDS). Two important quantities including 1) tidal recruited volume and 2) tidal hyperinflated volume were analytically computed by integrating the distribution of alveolar elements over the affected pop-open pressure range. We analytically predicted the tidal recruited volume of four canine subjects and compared our results with similar experimental measurements on canine models for the validation. We then applied our mathematical model to the P-V data of ARDS populations in four stages of Early ARDS, Deep Knee, Advanced ARDS and Baby Lung to quantify the tidal recruited volume and tidal hyperinflated volume as an indicator of ventilator-induced lung injury (VILI). These quantitative predictions based on patient-specific P-V data suggest that the optimum parameters of mechanical ventilation including PEEP and Tidal Pressure (Volume) are largely varying among ARDS population and are primarily influenced by the degree in the severity of ARDS.