Timely diagnosis and treatment of heart blocks are critical for preventing fatal outcomes in patients with cardiac conduction disorders. Expert analysis of clinical 12-lead electro-cardiograms (ECG) remains the standard diagnosis apparatus. In this study, we propose a self-supervised deep learning model that can detect evidence of heart blocks in ECG signals. We build a multichannel residual neural network (KRes) and train this model in a self-supervised contrastive fashion leveraging 3.6 million ECG from patients at the Massachusetts General Hospital (MGH) to learn robust ECG representations. We evaluate the utility of such representations on a large public dataset, PTB-XL, through fine-tuning and linear probing for identifying heart blocks. To compare, we also build a baseline supervised model by adapting KRes. The training data contains a subset of 10.6 thou-sand PTB-XL ECG from patients with and without heart block patterns, and a holdout set of 1319 ECG is used for evaluation. We compare the performances of the finetuning and the linear probing of the self-supervised representations with that of the supervised model using the area under the receiver-operating curve (AUC), sensitivity, specificity, and predictive values, and found those performing equally well. Moreover, we repeat the training cycles of the three pipelines while reducing the number of training samples and demonstrate that the performance of the self-supervised representations remain steady even when labeled data is limited. This research underscores the potential of self-supervised learning in cardiac diagnostics, emphasizing generalizability and performance across diverse datasets and clinical settings, especially in data-scarce paradigms.

Self-supervised learning (SSL) for clinical time series data has received significant attention in recent literature, since these data are highly rich and provide important information about a patient's physiological state. However, most existing SSL methods for clinical time series are limited in that they are designed for unimodal time series, such as a sequence of structured features (e.g., lab values and vitals signs) or an individual high-dimensional physiological signal (e.g., an electrocardiogram). These existing methods cannot be readily extended to model time series that exhibit multimodality, with structured features and high-dimensional data being recorded at each timestep in the sequence. In this work, we address this gap and propose a new SSL method --- Sequential Multi-Dimensional SSL --- where a SSL loss is applied both at the level of the entire sequence and at the level of the individual high-dimensional data points in the sequence in order to better capture information at both scales. Our strategy is agnostic to the specific form of loss function used at each level -- it can be contrastive, as in SimCLR, or non-contrastive, as in VICReg. We evaluate our method on two real-world clinical datasets, where the time series contains sequences of (1) high-frequency electrocardiograms and (2) structured data from lab values and vitals signs. Our experimental results indicate that pre-training with our method and then fine-tuning on downstream tasks improves performance over baselines on both datasets, and in several settings, can lead to improvements across different self-supervised loss functions.

Supervised machine learning applications in health care are often limited due to a scarcity of labeled training data. To mitigate the effect of small sample size, we introduce a pre-training approach, Patient Contrastive Learning of Representations (PCLR), which creates latent representations of electrocardiograms (ECGs) from a large number of unlabeled examples using contrastive learning. The resulting representations are expressive, performant, and practical across a wide spectrum of clinical tasks. We develop PCLR using a large health care system with over 3.2 million 12-lead ECGs and demonstrate that training linear models on PCLR representations achieves a 51% performance increase, on average, over six training set sizes and four tasks (sex classification, age regression, and the detection of left ventricular hypertrophy and atrial fibrillation), relative to training neural network models from scratch. We also compared PCLR to three other ECG pre-training approaches (supervised pre-training, unsupervised pre-training with an autoencoder, and pre-training using a contrastive multi ECG-segment approach), and show significant performance benefits in three out of four tasks. We found an average performance benefit of 47% over the other models and an average of a 9% performance benefit compared to best model for each task. We release PCLR to enable others to extract ECG representations at https://github.com/broadinstitute/ml4h/tree/master/model_zoo/PCLR.

Although oversampling methods are widely used to deal with class imbalance problems, most only utilize observed samples in the minority class and ignore the rich information available in the majority class. In this work, we use an oversampling method that leverages information in both the majority and minority classes to mitigate the class imbalance problem. Experimental results on two clinical datasets with highly imbalanced outcomes demonstrate that prediction models can be significantly improved using data obtained from this oversampling method when the number of minority class samples is very small.

UNLABELLED: Intrinsically disordered proteins (IDPs) play central roles in many biological processes. Consequently, an accurate description of the disordered state is an important step towards a comprehensive understanding of a number of important biological functions. In this work we describe a new web server, Mollack, for the automated construction of unfolded ensembles that uses both experimental and molecular simulation data to construct models for the unfolded state. An important aspect of the method is that it calculates a quantitative estimate of the uncertainty in the constructed ensemble, thereby providing an objective measure of the quality of the final model. Overall, Mollack facilitates structure-function studies of disordered proteins. AVAILABILITY AND IMPLEMENTATION: http://cmstultz-mollack.mit.edu CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

In this article, we propose a methodology for identifying predictive physiological patterns in the absence of prior knowledge. We use the principle of conservation to identify activity that consistently precedes an outcome in patients, and describe a two-stage process that allows us to efficiently search for such patterns in large datasets. This involves first transforming continuous physiological signals from patients into symbolic sequences, and then searching for patterns in these reduced representations that are strongly associated with an outcome.Our strategy of identifying conserved activity that is unlikely to have occurred purely by chance in symbolic data is analogous to the discovery of regulatory motifs in genomic datasets. We build upon existing work in this area, generalizing the notion of a regulatory motif and enhancing current techniques to operate robustly on non-genomic data. We also address two significant considerations associated with motif discovery in general: computational efficiency and robustness in the presence of degeneracy and noise. To deal with these issues, we introduce the concept of active regions and new subset-based techniques such as a two-layer Gibbs sampling algorithm. These extensions allow for a framework for information inference, where precursors are identified as approximately conserved activity of arbitrary complexity preceding multiple occurrences of an event.We evaluated our solution on a population of patients who experienced sudden cardiac death and attempted to discover electrocardiographic activity that may be associated with the endpoint of death. To assess the predictive patterns discovered, we compared likelihood scores for motifs in the sudden death population against control populations of normal individuals and those with non-fatal supraventricular arrhythmias. Our results suggest that predictive motif discovery may be able to identify clinically relevant information even in the absence of significant prior knowledge.

Tau is a natively unfolded protein that forms intracellular aggregates in the brains of patients with Alzheimer's disease. To decipher the mechanism underlying the formation of tau aggregates, we developed a novel approach for constructing models of natively unfolded proteins. The method, energy-minima mapping and weighting (EMW), samples local energy minima of subsequences within a natively unfolded protein and then constructs ensembles from these energetically favorable conformations that are consistent with a given set of experimental data. A unique feature of the method is that it does not strive to generate a single ensemble that represents the unfolded state. Instead we construct a number of candidate ensembles, each of which agrees with a given set of experimental constraints, and focus our analysis on local structural features that are present in all of the independently generated ensembles. Using EMW we generated ensembles that are consistent with chemical shift measurements obtained on tau constructs. Thirty models were constructed for the second microtubule binding repeat (MTBR2) in wild-type (WT) tau and a DeltaK280 mutant, which is found in some forms of frontotemporal dementia. By focusing on structural features that are preserved across all ensembles, we find that the aggregation-initiating sequence, PHF6*, prefers an extended conformation in both the WT and DeltaK280 sequences. In addition, we find that residue K280 can adopt a loop/turn conformation in WT MTBR2 and that deletion of this residue, which can adopt nonextended states, leads to an increase in locally extended conformations near the C-terminus of PHF6*. As an increased preference for extended states near the C-terminus of PHF6* may facilitate the propagation of beta-structure downstream from PHF6*, these results explain how a deletion at position 280 can promote the formation of tau aggregates.