Alok Choudhary

Purpose: Ventilator-associated pneumonia (VAP) remains one of the most serious hospital-acquired infections in the intensive care unit (ICU), with high morbidity and mortality. Early identification of patients at risk for developing VAP could enable timely diagnostics and intervention. However, current clinical tools are limited in their ability to detect early physiologic signals preceding VAP onset. We aimed to build supervised machine learning models to predict short term onset of VAP. Methods: We analyzed electronic health record data from a prospective observational cohort of ICU patients, where VAP was adjudicated using a standardized published protocol by a panel of critical care physicians. Clinical features (including vital signs, ventilator settings, laboratory values, and support devices) were extracted for each patient-ICU-day. We explored unsupervised clustering to characterize feature dynamics associated with VAP onset. We built multiple machine learning models across different prediction windows (3, 5, 7 days before VAP). We examined model performance in two external cohorts, MIMIC-IV and secondary analysis of the AMIKINHAL trial. Results were evaluated with discrimination metrics such as AUROC. Results: The internal cohort included 507 patients with BAL-confirmed diagnoses: 261 developed VAP and 246 did not have VAP. Visualization using clustering identified distinct physiologic states enriched for VAP-labeled days. The best-performing model achieved an AUROC of 0.866 in predicting VAP up to seven days before clinical diagnosis. Temporal model probability trajectories showed rising model confidence in the days leading up to VAP. On external validation in MIMIC-IV, the best model achieved an AUROC of 0.817 for forecasting VAP within five days. There was low feature overlap with the AMIKINHAL trial data, leading to poor model performance. Feature analysis revealed that platelet count, positive end-expiratory pressure (PEEP), ventilator duration, and inflammatory markers were key drivers of model predictions. Conclusions: Machine learning models trained on routinely collected ICU data with careful labeling can anticipate VAP onset up to a week in advance with strong predictive performance. Model performance generalized to data from an entirely different hospital system despite differences in practice and labeling patterns, but did not perform well when there was poor feature overlap. Future work should focus on real-time prospective evaluation.

Background: Interstitial lung disease (ILD) affects > 40% of patients with systemic sclerosis (SSc/scleroderma) and is the leading cause of disease-related mortality. Although therapies may slow progression, outcomes remain poor, partly because ILD is often detected after irreversible lung injury has occurred. Although chest computed tomography (CT) is a sensitive tool for ILD detection and is recommended at SSc diagnosis, it is oftentimes not performed and even less often performed serially. We sought to develop tools to predict ILD and mortality in patients with SSc using data routinely available in the electronic health record (EHR) to inform medical decision-making. Methods: We analyzed longitudinal EHR data from two SSc cohorts: Northwestern University (1,169 participants; derivation cohort) and Yale University (376 participants; validation cohort). We identified clinical features from existing cohort-linked EHR queries composing a convenience sample of data from participants spanning decades rather than employing a single unified data collection effort. Three ILD experts independently reviewed CT reports and classified each as having or lacking ILD. To explore derivation cohort data structure, patients with > =3 forced vital capacity (FVC) results available were identified and stratified according to prevalent or absent ILD. Using unsupervised trajectory-based clustering exploratory analyses, we determined standardized patterns across groups. ML models were then developed using clinical EHR data as predictor variables and prevalent ILD and all-cause mortality as outcome variables. Model performance was assessed using area under the receiver operating characteristic curve (AUC). Results: Seventy-four clinical features with low missingness, including demographic, vital sign, laboratory, and pulmonary function test data, were utilized for analyses. Four robust PFT trajectory clusters were identified that were associated with ILD prevalence and mortality in exploratory analyses. A ML model for ILD detection achieved an AUC of 0.832 and retained performance in the Yale cohort (AUC 0.754). In addition to established predictors such as autoantibodies and pulmonary function, the model identified routine laboratory measurements, including red cell distribution width (RDW), white blood cell count, and serum chloride, as important contributors. One-year mortality prediction achieved AUCs of 0.904 in the North-western cohort and 0.910 in the Yale cohort. Among patients with SSc-ILD, one-year mortality was predicted with AUCs of 0.744 and 0.902 in the Northwestern and Yale cohorts, respectively. Unexpectedly, we found that subtle laboratory abnormalities (such as change in RDW) contributed to predicting mortality. Conclusions: Our prediction models comprised of widely available EHR data are useful tools to identify SSc patients at high risk for prevalent ILD and all-cause mortality. Integration of these models into clinical practice could enable scalable risk stratification and inform individualized ILD screening and monitoring strategies for SSc patients.

Recent innovations have made it possible to produce millions of distinct nanoparticles on a chip. These vast volumes of data are impossible to analyze manually, necessitating the development of automated tools. In previous work, we created a binary classification machine learning model to select quality nanoparticle images for downstream analysis. In this work, we show that adding a custom image preprocessing step before model training can produce significantly higher-performing models in a fraction of the time and make the model more robust to different image noise levels and microscope acquisition settings. The proposed image processing pipeline effectively cleans raw nanoparticle images, enhances key features, and allows us to use much lower resolution images and simpler neural network model architectures, resulting in higher performance and significant cost savings. Experiments demonstrate superior performance relative to our baseline, including a 15% improvement in recall and more than a 10% increase in accuracy. Given the high cost of downstream analysis, it is critical to minimize false positives in our application, and our best-performing model obtains a precision of 97.3% and weighted F-score of 95.9% on an unseen test set. Additionally, model training time is reduced from 15.5 hours to 32 seconds. We expect that adopting this pipeline for AI-driven automated nanoparticle characterization will offer a considerable speedup in the laboratory, allowing researchers to rapidly and accurately analyze much greater volumes of data and accelerate materials discovery.

Background Criteria to identify patients who are ready to be liberated from mechanical ventilation are imprecise, often resulting in prolonged mechanical ventilation or reintubation, both of which are associated with adverse outcomes. Daily protocol-driven assessment of the need for mechanical ventilation leads to earlier extubation but requires dedicated personnel. We sought to determine whether machine learning applied to the electronic health record could predict successful extubation. Methods We examined 37 clinical features from patients from a single-center prospective cohort study of patients in our quaternary care medical ICU who required mechanical ventilation and underwent a bronchoalveolar lavage for known or suspected pneumonia. We also tested our models on an external test set from a community hospital ICU in our health care system. We curated electronic health record data aggregated from midnight to 8AM and labeled extubation status. We deployed three data encoding/imputation strategies and built XGBoost, LightGBM, logistic regression, LSTM, and RNN models to predict successful next-day extubation. We evaluated each model’s performance using Area Under the Receiver Operating Characteristic (AUROC), Area Under the Precision Recall Curve (AUPRC), Sensitivity (Recall), Specificity, PPV (Precision), Accuracy, and F1-Score. Results Our internal cohort included 696 patients and 9,828 ICU days, and our external cohort had 333 patients and 2,835 ICU days. The best model (LSTM) predicted successful extubation on a given ICU day with an AUROC 0.87 (95% CI 0.834-0.902) and the internal test set and 0.87 (95% CI 0.848-0.885) on the external test set. A Logistic Regression model performed similarly (AUROC 0.86 internal test, 0.83 external test). Across multiple model types, measures previously demonstrated to be important in determining readiness for extubation were found to be most informative, including plateau pressure and Richmond Agitation Sedation Scale (RASS) score. Our model often predicted patients to be stable for extubation in the days preceding their actual extubation, with 63.8% of predicted extubations occurring within three days of true extubation. We also tested the best model on cases of failed extubations (requiring reintubation within two days) not seen by the model during training. Our best model would have identified 35.4% (17/48) of these cases in the internal test set and 48.1% (13/27) cases in the external test set as unlikely to be successfully extubated. Conclusions Machine learning models can accurately predict the likelihood of extubation on a given ICU day from data available in the electronic health record. Predictions from these models are driven by clinical features that have been associated with successful extubation in clinical trials.

Large-scale parallel applications can face significant I/O performance bottlenecks, making efficient I/O crucial. This work presents a comparative study of several parallel I/O implementations in the Weather Research and Forecasting model, including PnetCDF blocking and non-blocking I/O options, netCDF4, HDF5 Log VOL, and ADIOS. For I/O methods creating files in a canonical data layout, PnetCDF's non-blocking option offers up to 2x improvement over its blocking option and up to 4.5x over HDF5 via netCDF4, demonstrating the effectiveness of the write request aggregation technique. The HDF5 Log VOL outperforms ADIOS with a 4x improvement in write performance when creating files in the log layout, although both require non-negligible time to convert the file back to canonical order for post-run analysis. From these results we extract some observations that can guide I/O strategies for modern parallel codes.

Prediction of the autogenous shrinkage referred to as the reduction of apparent volume of concrete under seal and isothermal conditions is of great significance in the service life analysis and design of durable concrete structures, especially with the increasing use of concrete with low water-to-cement ratios. However, due to the highly complex mechanism of autogenous shrinkage, it is hard to design accurate mechanistic models for it. Existing state-of-the-art models for autogenous shrinkage do not perform well for several reasons such as not being able to capture faster shrinkage change at early ages (swelling), coefficients used are derived using statistical optimization methods to fit certain databases only, and mechanism to identify the most influencing factors on autogenous shrinkage is not present. Moreover, it is also challenging to deploy a machine learning framework directly to perform predictive analysis due to the sparse and noisy nature of the available experimental dataset. In this paper, we study and propose a method to combine the physics-based knowledge and the predictive ability of deep regression neural networks to mitigate the shortcomings of the existing models. We introduce a novel data augmentation technique that utilizes physics based knowledge to improve the accuracy while maintaining the characteristics of autogenous shrinkage in its predictions simultaneously. Using state-of-the-art B4 model, a genetic algorithm, and a deep neural network trained using raw data for comparison, we show that the proposed methods help improve the accuracy of the model as compared to other methods. We also observe that the proposed method is able to successfully learn and predict the swelling component of the shrinkage strain curve as well, which cannot be predicted using the existing state-of-the-art models.