Search

Within the intensive care unit (ICU), a wealth of patient data, including clinical measurements and clinical notes, is readily available. This data is a valuable resource for comprehending patient health and informing medical decisions, but it also contains many challenges in analysis. Deep learning models show promise in extracting meaningful patterns, but they require extensive labeled data, a challenge in critical care. To address this, we propose a novel approach employing self-supervised pretraining, focusing on the alignment of clinical measurements and notes. Our approach combines contrastive and masked token prediction tasks during pretraining. Semi-supervised experiments on the MIMIC-III dataset demonstrate the effectiveness of our self-supervised pretraining. In downstream tasks, including in-hospital mortality prediction and phenotyping, our pretrained model outperforms baselines in settings where only a fraction of the data is labeled, emphasizing its ability to enhance ICU data analysis. Notably, our method excels in situations where very few labels are available, as evidenced by an increase in the AUC-ROC for in-hospital mortality by 0.17 and in AUC-PR for phenotyping by 0.1 when only 1% of labels are accessible. This work advances self-supervised learning in the healthcare domain, optimizing clinical insights from abundant yet challenging ICU data.

In wearable sensing applications, data is inevitable to be irregularly sampled or partially missing, which pose challenges for any downstream application. An unique aspect of wearable data is that it is time-series data and each channel can be correlated to another one, such as x, y, z axis of accelerometer. We argue that traditional methods have rarely made use of both times-series dynamics of the data as well as the relatedness of the features from different sensors. We propose a model, termed as DynImp, to handle different time point's missingness with nearest neighbors along feature axis and then feeding the data into a LSTM-based denoising autoencoder which can reconstruct missingness along the time axis. We experiment the model on the extreme missingness scenario ($>50\%$ missing rate) which has not been widely tested in wearable data. Our experiments on activity recognition show that the method can exploit the multi-modality features from related sensors and also learn from history time-series dynamics to reconstruct the data under extreme missingness., Comment: 5 pages, 2 figures, accepted in ICASSP'2022

Advances in self-supervised learning (SSL) have shown that self-supervised pretraining on medical imaging data can provide a strong initialization for downstream supervised classification and segmentation. Given the difficulty of obtaining expert labels for medical image recognition tasks, such an "in-domain" SSL initialization is often desirable due to its improved label efficiency over standard transfer learning. However, most efforts toward SSL of medical imaging data are not adapted to video-based medical imaging modalities. With this progress in mind, we developed a self-supervised contrastive learning approach, EchoCLR, catered to echocardiogram videos with the goal of learning strong representations for efficient fine-tuning on downstream cardiac disease diagnosis. EchoCLR leverages (i) distinct videos of the same patient as positive pairs for contrastive learning and (ii) a frame re-ordering pretext task to enforce temporal coherence. When fine-tuned on small portions of labeled data (as few as 51 exams), EchoCLR pretraining significantly improved classification performance for left ventricular hypertrophy (LVH) and aortic stenosis (AS) over other transfer learning and SSL approaches across internal and external test sets. For example, when fine-tuning on 10% of available training data (519 studies), an EchoCLR-pretrained model achieved 0.72 AUROC (95% CI: [0.69, 0.75]) on LVH classification, compared to 0.61 AUROC (95% CI: [0.57, 0.64]) with a standard transfer learning approach. Similarly, using 1% of available training data (53 studies), EchoCLR pretraining achieved 0.82 AUROC (95% CI: [0.79, 0.84]) on severe AS classification, compared to 0.61 AUROC (95% CI: [0.58, 0.65]) with transfer learning. EchoCLR is unique in its ability to learn representations of medical videos and demonstrates that SSL can enable label-efficient disease classification from small, labeled datasets.

Medical events of interest, such as mortality, often happen at a low rate in electronic medical records, as most admitted patients survive. Training models with this imbalance rate (class density discrepancy) may lead to suboptimal prediction. Traditionally this problem is addressed through ad-hoc methods such as resampling or reweighting but performance in many cases is still limited. We propose a framework for training models for this imbalance issue: 1) we first decouple the feature extraction and classification process, adjusting training batches separately for each component to mitigate bias caused by class density discrepancy; 2) we train the network with both a density-aware loss and a learnable cost matrix for misclassifications. We demonstrate our model's improved performance in real-world medical datasets (TOPCAT and MIMIC-III) to show improved AUC-ROC, AUC-PRC, Brier Skill Score compared with the baselines in the domain.

Sustained high levels of blood glucose in type 2 diabetes (T2DM) can have disastrous long-term health consequences. An essential component of clinical interventions for T2DM is monitoring dietary intake to keep plasma glucose levels within an acceptable range. Yet, current techniques to monitor food intake are time intensive and error prone. To address this issue, we are developing techniques to automatically monitor food intake and the composition of those foods using continuous glucose monitors (CGMs). This article presents the results of a clinical study in which participants consumed nine standardized meals with known macronutrients amounts (carbohydrate, protein, and fat) while wearing a CGM. We built a multitask neural network to estimate the macronutrient composition from the CGM signal, and compared it against a baseline linear regression. The best prediction result comes from our proposed neural network, trained with subject-dependent data, as measured by root mean squared relative error and correlation coefficient. These findings suggest that it is possible to estimate macronutrient composition from CGM signals, opening the possibility to develop automatic techniques to track food intake.

In many machine learning tasks, input features with varying degrees of predictive capability are acquired at varying costs. In order to optimize the performance-cost trade-off, one would select features to observe a priori. However, given the changing context with previous observations, the subset of predictive features to select may change dynamically. Therefore, we face the challenging new problem of foresight dynamic selection (FDS): finding a dynamic and light-weight policy to decide which features to observe next, before actually observing them, for overall performance-cost trade-offs. To tackle FDS, this paper proposes a Bayesian learning framework of Variational Foresight Dynamic Selection (VFDS). VFDS learns a policy that selects the next feature subset to observe, by optimizing a variational Bayesian objective that characterizes the trade-off between model performance and feature cost. At its core is an implicit variational distribution on binary gates that are dependent on previous observations, which will select the next subset of features to observe. We apply VFDS on the Human Activity Recognition (HAR) task where the performance-cost trade-off is critical in its practice. Extensive results demonstrate that VFDS selects different features under changing contexts, notably saving sensory costs while maintaining or improving the HAR accuracy. Moreover, the features that VFDS dynamically select are shown to be interpretable and associated with the different activity types. We will release the code.

Goal: To achieve-high quality comprehensive feature extraction from physiological signals that enables precise physiological parameter estimation despite evolving waveform morphologies. Methods: We propose Boosted-SpringDTW, a probabilistic framework that leverages dynamic time warping (DTW) and minimal domain-specific heuristics to simultaneously segment physiological signals and identify fiducial points that represent cardiac events. An automated dynamic template adapts to evolving waveform morphologies. We validate Boosted-SpringDTW performance with a benchmark PPG dataset whose morphologies include subject- and respiratory-induced variation. Results: Boosted-SpringDTW achieves precision, recall, and F1-scores over 0.96 for identifying fiducial points and mean absolute error values less than 11.41 milliseconds when estimating IBI. Conclusion: Boosted-SpringDTW improves F1-Scores compared to two baseline feature extraction algorithms by 35 percent on average for fiducial point identification and mean percent difference by 16 percent on average for IBI estimation. Significance: Precise hemodynamic parameter estimation with wearable devices enables continuous health monitoring throughout a patients' daily life.

Electronic Health Record (EHR) systems provide critical, rich and valuable information at high frequency. One of the most exciting applications of EHR data is in developing a real-time mortality warning system with tools from survival analysis. However, most of the survival analysis methods used recently are based on (semi)parametric models using static covariates. These models do not take advantage of the information conveyed by the time-varying EHR data. In this work, we present an application of a highly scalable survival analysis method, BoXHED 2.0 to develop a real-time in-ICU mortality warning indicator based on the MIMIC IV data set. Importantly, BoXHED can incorporate time-dependent covariates in a fully nonparametric manner and is backed by theory. Our in-ICU mortality model achieves an AUC-PRC of 0.41 and AUC-ROC of 0.83 out of sample, demonstrating the benefit of real-time monitoring.

Machine learning continues to grow in popularity due to its ability to learn increasingly complex tasks. However, for many supervised models, the shift in a data distribution or the appearance of a new event can result in a severe decrease in model performance. Retraining a model from scratch with updated data can be resource intensive or impossible depending on the constraints placed on an organization or system. Continual learning methods attempt to adapt models to new classes instead of retraining. However, many of these methods do not have a detection method for new classes or make assumptions about the distribution of classes. In this paper, we develop an attention based Gaussian Mixture, called GMAT, that learns interpretable representations of data with or without labels. We incorporate this method with existing Neural Architecture Search techniques to develop an algorithm for detection new events for an optimal number of representations through an iterative process of training a growing. We show that our method is capable learning new representations of data without labels or assumptions about the distributions of labels. We additionally develop a method that allows our model to utilize labels to more accurately develop representations. Lastly, we show that our method can avoid catastrophic forgetting by replaying samples from learned representations., Comment: 8 pages, 5 figures

The recent breakthrough achieved by contrastive learning accelerates the pace for deploying unsupervised training on real-world data applications. However, unlabeled data in reality is commonly imbalanced and shows a long-tail distribution, and it is unclear how robustly the latest contrastive learning methods could perform in the practical scenario. This paper proposes to explicitly tackle this challenge, via a principled framework called Self-Damaging Contrastive Learning (SDCLR), to automatically balance the representation learning without knowing the classes. Our main inspiration is drawn from the recent finding that deep models have difficult-to-memorize samples, and those may be exposed through network pruning. It is further natural to hypothesize that long-tail samples are also tougher for the model to learn well due to insufficient examples. Hence, the key innovation in SDCLR is to create a dynamic self-competitor model to contrast with the target model, which is a pruned version of the latter. During training, contrasting the two models will lead to adaptive online mining of the most easily forgotten samples for the current target model, and implicitly emphasize them more in the contrastive loss. Extensive experiments across multiple datasets and imbalance settings show that SDCLR significantly improves not only overall accuracies but also balancedness, in terms of linear evaluation on the full-shot and few-shot settings. Our code is available at: https://github.com/VITA-Group/SDCLR., Comment: ICML 2021

Modern applications of survival analysis increasingly involve time-dependent covariates. The Python package BoXHED2.0 is a tree-boosted hazard estimator that is fully nonparametric, and is applicable to survival settings far more general than right-censoring, including recurring events and competing risks. BoXHED2.0 is also scalable to the point of being on the same order of speed as parametric boosted survival models, in part because its core is written in C++ and it also supports the use of GPUs and multicore CPUs. BoXHED2.0 is available from PyPI and also from www.github.com/BoXHED., Comment: 27 pages

Problem lists are intended to provide clinicians with a relevant summary of patient medical issues and are embedded in many electronic health record systems. Despite their importance, problem lists are often cluttered with resolved or currently irrelevant conditions. In this work, we develop a novel end-to-end framework that first extracts diagnosis and procedure information from clinical notes and subsequently uses the extracted medical problems to predict patient outcomes. This framework is both more performant and more interpretable than existing models used within the domain, achieving an AU-ROC of 0.710 for bounceback readmission and 0.869 for in-hospital mortality occurring after ICU discharge. We identify risk factors for both readmission and mortality outcomes and demonstrate that our framework can be used to develop dynamic problem lists that present clinical problems along with their quantitative importance. We conduct a qualitative user study with medical experts and demonstrate that they view the lists produced by our framework favorably and find them to be a more effective clinical decision support tool than a strong baseline., Comment: To appear in the proceedings of the Machine Learning for Healthcare Conference (MLHC) 2020. Accepted papers can be viewed at https://www.mlforhc.org/accepted-papers

Blood pressure monitoring is an essential component of hypertension management and in the prediction of associated comorbidities. Blood pressure is a dynamic vital sign with frequent changes throughout a given day. Capturing blood pressure remotely and frequently (also known as ambulatory blood pressure monitoring) has traditionally been achieved by measuring blood pressure at discrete intervals using an inflatable cuff. However, there is growing interest in developing a cuffless ambulatory blood pressure monitoring system to measure blood pressure continuously. One such approach is by utilizing bioimpedance sensors to build regression models. A practical problem with this approach is that the amount of data required to confidently train such a regression model can be prohibitive. In this paper, we propose the application of the domain-adversarial training neural network (DANN) method on our multitask learning (MTL) blood pressure estimation model, allowing for knowledge transfer between subjects. Our proposed model obtains average root mean square error (RMSE) of $4.80 \pm 0.74$ mmHg for diastolic blood pressure and $7.34 \pm 1.88$ mmHg for systolic blood pressure when using three minutes of training data, $4.64 \pm 0.60$ mmHg and $7.10 \pm 1.79$ respectively when using four minutes of training data, and $4.48 \pm 0.57$ mmHg and $6.79 \pm 1.70$ respectively when using five minutes of training data. DANN improves training with minimal data in comparison to both directly training and to training with a pretrained model from another subject, decreasing RMSE by $0.19$ to $0.26$ mmHg (diastolic) and by $0.46$ to $0.67$ mmHg (systolic) in comparison to the best baseline models. We observe that four minutes of training data is the minimum requirement for our framework to exceed ISO standards within this cohort of patients., Comment: Published at MLHC 2020

The proliferation of medical monitoring devices makes it possible to track health vitals at high frequency, enabling the development of dynamic health risk scores that change with the underlying readings. Survival analysis, in particular hazard estimation, is well-suited to analyzing this stream of data to predict disease onset as a function of the time-varying vitals. This paper introduces the software package BoXHED (pronounced 'box-head') for nonparametrically estimating hazard functions via gradient boosting. BoXHED 1.0 is a novel tree-based implementation of the generic estimator proposed in Lee, Chen, Ishwaran (2017), which was designed for handling time-dependent covariates in a fully nonparametric manner. BoXHED is also the first publicly available software implementation for Lee, Chen, Ishwaran (2017). Applying BoXHED to cardiovascular disease onset data from the Framingham Heart Study reveals novel interaction effects among known risk factors, potentially resolving an open question in clinical literature., Comment: 10 pages, 3 figures, 5 tables

Activity recognition in wearable computing faces two key challenges: i) activity characteristics may be context-dependent and change under different contexts or situations; ii) unknown contexts and activities may occur from time to time, requiring flexibility and adaptability of the algorithm. We develop a context-aware mixture of deep models termed the {\alpha}-\b{eta} network coupled with uncertainty quantification (UQ) based upon maximum entropy to enhance human activity recognition performance. We improve accuracy and F score by 10% by identifying high-level contexts in a data-driven way to guide model development. In order to ensure training stability, we have used a clustering-based pre-training in both public and in-house datasets, demonstrating improved accuracy through unknown context discovery., Comment: 10 pages, 5 figures, accepted by AISTATS 2020

Clinical notes contain a large amount of clinically valuable information that is ignored in many clinical decision support systems due to the difficulty that comes with mining that information. Recent work has found success leveraging deep learning models for the prediction of clinical outcomes using clinical notes. However, these models fail to provide clinically relevant and interpretable information that clinicians can utilize for informed clinical care. In this work, we augment a popular convolutional model with an attention mechanism and apply it to unstructured clinical notes for the prediction of ICU readmission and mortality. We find that the addition of the attention mechanism leads to competitive performance while allowing for the straightforward interpretation of predictions. We develop clear visualizations to present important spans of text for both individual predictions and high-risk cohorts. We then conduct a qualitative analysis and demonstrate that our model is consistently attending to clinically meaningful portions of the narrative for all of the outcomes that we explore., Comment: Machine Learning for Health (ML4H) at NeurIPS 2019 - Extended Abstract

Cardiovascular disorders account for nearly 1 in 3 deaths in the United States. Care for these disorders are often determined during visits to acute care facilities, such as hospitals. While the length of stay in these settings represents just a small proportion of patients' lives, they account for a disproportionately large amount of decision making. To overcome this bias towards data from acute care settings, there is a need for longitudinal monitoring in patients with cardiovascular disorders. Longitudinal monitoring can provide a more comprehensive picture of patient health, allowing for more informed decision making. This work surveys the current field of sensing technologies and machine learning analytics that exist in the field of remote monitoring for cardiovascular disorders. We highlight three primary needs in the design of new smart health technologies: 1) the need for sensing technology that can track longitudinal trends in signs and symptoms of the cardiovascular disorder despite potentially infrequent, noisy, or missing data measurements; 2) the need for new analytic techniques that model data captured in a longitudinal, continual fashion to aid in the development of new risk prediction techniques and in tracking disease progression; and 3) the need for machine learning techniques that are personalized and interpretable, allowing for advancements in shared clinical decision making. We highlight these needs based upon the current state-of-the-art in smart health technologies and analytics and discuss the ample opportunities that exist in addressing all three needs in the development of smart health technologies and analytics applied to the field of cardiovascular disorders and care., Comment: 32 pages, 3 figures, to be submitted to ACM Transactions on Computing for Healthcare (HEALTH), Special Issue on Wearable Technologies for Smart Health 2019

Visual summarization of clinical data collected on patients contained within the electronic health record (EHR) may enable precise and rapid triage at the time of patient presentation to an emergency department (ED). The triage process is critical in the appropriate allocation of resources and in anticipating eventual patient disposition, typically admission to the hospital or discharge home. EHR data are high-dimensional and complex, but offer the opportunity to discover and characterize underlying data-driven patient phenotypes. These phenotypes will enable improved, personalized therapeutic decision making and prognostication. In this work, we focus on the challenge of two-dimensional patient projections. A low dimensional embedding offers visual interpretability lost in higher dimensions. While linear dimensionality reduction techniques such as principal component analysis are often used towards this aim, they are insufficient to describe the variance of patient data. In this work, we employ the newly-described non-linear embedding technique called uniform manifold approximation and projection (UMAP). UMAP seeks to capture both local and global structures in high-dimensional data. We then use Gaussian mixture models to identify clusters in the embedded data and use the adjusted Rand index (ARI) to establish stability in the discovery of these clusters. This technique is applied to five common clinical chief complaints from a real-world ED EHR dataset, describing the emergent properties of discovered clusters. We observe clinically-relevant cluster attributes, suggesting that visual embeddings of EHR data using non-linear dimensionality reduction is a promising approach to reveal data-driven patient phenotypes. In the five chief complaints, we find between 2 and 6 clusters, with the peak mean pairwise ARI between subsequent training iterations to range from 0.35 to 0.74.

Emerging wearable sensors have enabled the unprecedented ability to continuously monitor human activities for healthcare purposes. However, with so many ambient sensors collecting different measurements, it becomes important not only to maintain good monitoring accuracy, but also low power consumption to ensure sustainable monitoring. This power-efficient sensing scheme can be achieved by deciding which group of sensors to use at a given time, requiring an accurate characterization of the trade-off between sensor energy usage and the uncertainty in ignoring certain sensor signals while monitoring. To address this challenge in the context of activity monitoring, we have designed an adaptive activity monitoring framework. We first propose a switching Gaussian process to model the observed sensor signals emitting from the underlying activity states. To efficiently compute the Gaussian process model likelihood and quantify the context prediction uncertainty, we propose a block circulant embedding technique and use Fast Fourier Transforms (FFT) for inference. By computing the Bayesian loss function tailored to switching Gaussian processes, an adaptive monitoring procedure is developed to select features from available sensors that optimize the trade-off between sensor power consumption and the prediction performance quantified by state prediction entropy. We demonstrate the effectiveness of our framework on the popular benchmark of UCI Human Activity Recognition using Smartphones., Comment: to appear in AISTATS 2019