Label noise is omnipresent in the annotations process and has an impact on supervised learning algorithms. This work focuses on the impact of label noise on the performance of learning models by examining the effect of random and class-dependent label noise on a binary classification task: quality assessment for photoplethysmography (PPG). PPG signal is used to detect physiological changes and its quality can have a significant impact on the subsequent tasks, which makes PPG quality assessment a particularly good target for examining the impact of label noise in the field of biomedicine. Random and class-dependent label noise was introduced separately into the training set to emulate the errors associated with fatigue and bias in labeling data samples. We also tested different representations of the PPG, including features defined by domain experts, 1D raw signal and 2D image. Three different classifiers are tested on the noisy training data, including support vector machine (SVM), XGBoost, 1D Resnet and 2D Resnet, which handle three representations, respectively. The results showed that the two deep learning models were more robust than the two traditional machine learning models for both the random and class-dependent label noise. From the representation perspective, the 2D image shows better robustness compared to the 1D raw signal. The logits from three classifiers are also analyzed, the predicted probabilities intend to be more dispersed when more label noise is introduced. From this work, we investigated various factors related to label noise, including representations, label noise type, and data imbalance, which can be a good guidebook for designing more robust methods for label noise in future work.

Population-based and case-control studies in different ethnicities have linked a polymorphism, C825T, in exon 10 of GNB3 gene to hypertension and several additional diseases. The 825T allele is associated with alternative splicing and results in a shortened Gβ3 protein, referred to as Gβ3s, which loses 41 amino acids encompassing one WD40 repeat domain. The mechanism of how Gβ3 C825T polymorphism is associated with hypertension has remained unclear, but an impairment of its canonical function in G-protein-coupled receptor signaling has been ruled out. Here, we report that Gβ3, like other Gβ proteins, binds to DDB1 and assembles a DDB1-CUL4A-ROC1 E3 ubiquitin ligase (CRL4A(Gβ3)) to target GRK2 ubiquitination. The loss of the 41 amino-acid residues disrupts the Gβ3-DDB1 binding and impairs the function of Gβ3s to ubiquitinate GRK2. GRK2 ubiquitination levels were decreased and protein levels were accumulated in the blood samples of Gβ3 825T allele carriers. Deletion of Cul4a in mice resulted in systolic pressure increased and weakened heart function in male mice that can be partially rescued by the deletion of one Grk2 allele. These results reveal a mechanism explaining the link between Gβ3 C825T polymorphism and hypertension.

Bedside patient monitors are ubiquitous tools in modern critical care units to provide timely patient status. However, current systems suffer from high volume of false alarms leading to alarm fatigue, one of top technical hazards in clinical settings. Many studies are racing to develop improved algorithms towards precision patient monitoring, while little has been done to investigate the aspect of algorithm generalizability across different health institutions. Our group has been developing an evolving framework termed SuperAlarm that extracts multivariate patterns in data streams (monitor alarms, electronic health records and physiologic waveforms) of modern health enterprise to predict patient deterioration and has demonstrated great potential in mitigating alarm fatigue. In this study, we further investigate the generalizability of SuperAlarm by designing a comprehensive approach to achieve performance comparison in predicting in-hospital code blue (CB) events across two health institutions. SuperAlarm model trained with alarm data in one institution is tested on both internal and external test sets. Results show comparable performance with sensitivity up to 80% within one-hour window of events and over 90% in reduction of false alarms in both institutions. Cross-institutional performance agreement can be further improved by predicting a more stringent CB subtype (cardiopulmonary arrest), with internal sensitivity lying within 95% confident interval of external one up to 8-hour before event onset. The cross-institutional performance comparison offers first-hand knowledge on both advantages and challenges in generalizing a prediction algorithm across different institutions, which hold key information to guide the design of model training and deployment strategy.

Background

Patients hospitalized for suspected acute coronary syndrome (ACS) are at risk for transient myocardial ischemia. During the "rule-out" phase, continuous ECG ST-segment monitoring can identify transient myocardial ischemia, even when asymptomatic. However, current ST-segment monitoring software is vastly underutilized due to false positive alarms, with resultant alarm fatigue. Current ST algorithms may contribute to alarm fatigue because; (1) they are not designed with a delay (minutes), rather alarm to brief spikes (i.e., turning, heart rate changes), and (2) alarm to changes in a single ECG lead, rather than contiguous leads.

Purpose

This study was designed to determine sensitivity, and specificity, of ST algorithms when accounting for; ST magnitude (100μV vs 200μV), duration, and changes in contiguous ECG leads (i.e., aVL, I, - aVR, II, aVF, III; V1, V2, V3, V4, V5, V6, V6, I).

Methods

This was a secondary analysis from the COMPARE Study, which assessed occurrence rates for transient myocardial ischemia in hospitalized patients with suspected ACS using 12-lead Holter. Transient myocardial ischemia was identified from Holter using >100μV ST-segment ↑ or ↓, in >1 ECG lead, >1min. Algorithms tested against Holter transient myocardial ischemia were done using the University of California San Francisco (UCSF) ECG algorithm and included: (1)100μV vs 200μV any lead during a 5-min ST average; (2)100μV vs 200μV any lead >5min, (3) 100μV vs 200μV any lead during a 5-min ST average in contiguous leads, and (4) 100μV vs 200μV>5min in contiguous leads (Table below).

Results

In 361 patients; mean age 63+12years, 63% male, 56% prior CAD, 43 (11%) had transient myocardial ischemia. Of the 43 patients with transient myocardial ischemia, 17 (40%) had ST-segment elevation events, and 26 (60%) ST-segment depression events. A higher proportion of patients with ST segment depression has missed ischemic events. Table shows sensitivity and specificity for the four algorithms tested.

Conclusions

Sensitivity was highly variable, due to the ST threshold selected, with the 100μV measurement point being superior to the 200μV amplitude threshold. Of all the algorithms tested, there was moderate sensitivity and specificity (70% and 68%) using the 100μV ST-segment threshold, integrated ST-segment changes in contiguous leads during a 5-min average.

Patients with suspected acute coronary syndrome (ACS) are at risk of transient myocardial ischemia (TMI), which could lead to serious morbidity or even mortality. Early detection of myocardial ischemia can reduce damage to heart tissues and improve patient condition. Significant ST change in the electrocardiogram (ECG) is an important marker for detecting myocardial ischemia during the rule-out phase of potential ACS. However, current ECG monitoring software is vastly underused due to excessive false alarms. The present study aims to tackle this problem by combining a novel image-based approach with deep learning techniques to improve the detection accuracy of significant ST depression change. The obtained convolutional neural network (CNN) model yields an average area under the curve (AUC) at 89.6% from an independent testing set. At selected optimal cutoff thresholds, the proposed model yields a mean sensitivity at 84.4% while maintaining specificity at 84.9%.

Photoplethysmography (PPG) is a ubiquitous physiological measurement that detects beat-to-beat pulsatile blood volume changes and hence has a potential for monitoring cardiovascular conditions, particularly in ambulatory settings. A PPG dataset that is created for a particular use case is often imbalanced, due to a low prevalence of the pathological condition it targets to predict and the paroxysmal nature of the condition as well. To tackle this problem, we propose log-spectral matching GAN (LSM-GAN), a generative model that can be used as a data augmentation technique to alleviate the class imbalance in a PPG dataset to train a classifier. LSM-GAN utilizes a novel generator that generates a synthetic signal without a up-sampling process of input white noises, as well as adds the mismatch between real and synthetic signals in frequency domain to the conventional adversarial loss. In this study, experiments are designed focusing on examining how the influence of LSM-GAN as a data augmentation technique on one specific classification task - atrial fibrillation (AF) detection using PPG. We show that by taking spectral information into consideration, LSM-GAN as a data augmentation solution can generate more realistic PPG signals.

Photoplethysmography (PPG) is a noninvasive way to monitor various aspects of the circulatory system, and is becoming more and more widespread in biomedical processing. Recently, deep learning methods for analyzing PPG have also become prevalent, achieving state of the art results on heart rate estimation, atrial fibrillation detection, and motion artifact identification. Consequently, a need for interpretable deep learning has arisen within the field of biomedical signal processing. In this paper, we pioneer novel explanatory metrics which leverage domain-expert knowledge to validate a deep learning model. We visualize model attention over a whole testset using saliency methods and compare it to human expert annotations. Congruence, our first metric, measures the proportion of model attention within expert-annotated regions. Our second metric, Annotation Classification, measures how much of the expert annotations our deep learning model pays attention to. Finally, we apply our metrics to compare between a signal based model and an image based model for PPG signal quality classification. Both models are deep convolutional networks based on the ResNet architectures. We show that our signal-based one dimensional model acts in a more explainable manner than our image based model; on average 50.78% of the one dimensional model's attention are within expert annotations, whereas 36.03% of the two dimensional model's attention are within expert annotations. Similarly, when thresholding the one dimensional model attention, one can more accurately predict if each pixel of the PPG is annotated as artifactual by an expert. Through this testcase, we demonstrate how our metrics can provide a quantitative and dataset-wide analysis of how explainable the model is.

Early detection of Atrial Fibrillation (AFib) is crucial to prevent stroke recurrence. New tools for monitoring cardiac rhythm are important for risk stratification and stroke prevention. As many of new approaches to long-term AFib detection are now based on photoplethysmogram (PPG) recordings from wearable devices, ensuring high PPG signal-to-noise ratios is a fundamental requirement for a robust detection of AFib episodes. Traditionally, signal quality assessment is often based on the evaluation of similarity between pulses to derive signal quality indices. There are limitations to using this approach for accurate assessment of PPG quality in the presence of arrhythmia, as in the case of AFib, mainly due to substantial changes in pulse morphology. In this paper, we first tested the performance of algorithms selected from a body of studies on PPG quality assessment using a dataset of PPG recordings from patients with AFib. We then propose machine learning approaches for PPG quality assessment in 30-s segments of PPG recording from 13 stroke patients admitted to the University of California San Francisco (UCSF) neuro intensive care unit and another dataset of 3764 patients from one of the five UCSF general intensive care units. We used data acquired from two systems, fingertip PPG (fPPG) from a bedside monitor system, and radial PPG (rPPG) measured using a wearable commercial wristband. We compared various supervised machine learning techniques including k-nearest neighbors, decisions trees, and a two-class support vector machine (SVM). SVM provided the best performance. fPPG signals were used to build the model and achieved 0.9477 accuracy when tested on the data from the fPPG exclusive to the test set, and 0.9589 accuracy when tested on the rPPG data.