Evaluate underdiagnosis and overdiagnosis bias of deep learning model on primary open-angle glaucoma diagnosis in under-served populations.

In the United States, primary open-angle glaucoma (POAG) is the leading cause of blindness, especially among African American and Hispanic individuals. Deep learning has been widely used to detect POAG using fundus images as its performance is comparable to or even surpasses diagnosis by clinicians. However, human bias in clinical diagnosis may be reflected and amplified in the widely-used deep learning models, thus impacting their performance. Biases may cause (1) underdiagnosis, increasing the risks of delayed or inadequate treatment, and (2) overdiagnosis, which may increase individuals' stress, fear, well-being, and unnecessary/costly treatment. In this study, we examined the underdiagnosis and overdiagnosis when applying deep learning in POAG detection based on the Ocular Hypertension Treatment Study (OHTS) from 22 centers across 16 states in the United States. Our results show that the widely-used deep learning model can underdiagnose or overdiagnose under-served populations. The most underdiagnosed group is female younger (< 60 yrs) group, and the most overdiagnosed group is Black older (≥ 60 yrs) group. Biased diagnosis through traditional deep learning methods may delay disease detection, treatment and create burdens among under-served populations, thereby, raising ethical concerns about using deep learning models in ophthalmology clinics.

Predicting Age-related Macular Degeneration Progression with Longitudinal Fundus Images Using Deep Learning.

Accurately predicting a patient's risk of progressing to late age-related macular degeneration (AMD) is difficult but crucial for personalized medicine. While existing risk prediction models for progression to late AMD are useful for triaging patients, none utilizes longitudinal color fundus photographs (CFPs) in a patient's history to estimate the risk of late AMD in a given subsequent time interval. In this work, we seek to evaluate how deep neural networks capture the sequential information in longitudinal CFPs and improve the prediction of 2-year and 5-year risk of progression to late AMD. Specifically, we proposed two deep learning models, CNN-LSTM and CNN-Transformer, which use a Long-Short Term Memory (LSTM) and a Transformer, respectively with convolutional neural networks (CNN), to capture the sequential information in longitudinal CFPs. We evaluated our models in comparison to baselines on the Age-Related Eye Disease Study, one of the largest longitudinal AMD cohorts with CFPs. The proposed models outperformed the baseline models that utilized only single-visit CFPs to predict the risk of late AMD (0.879 vs 0.868 in AUC for 2-year prediction, and 0.879 vs 0.862 for 5-year prediction). Further experiments showed that utilizing longitudinal CFPs over a longer time period was helpful for deep learning models to predict the risk of late AMD. We made the source code available at https://github.com/bionlplab/AMD_prognosis_mlmi2022 to catalyze future works that seek to develop deep learning models for late AMD prediction.

Enhancing convolutional neural network predictions of electrocardiograms with left ventricular dysfunction using a novel sub-waveform representation.

Background: Electrocardiogram (ECG) deep learning (DL) has promise to improve the outcomes of patients with cardiovascular abnormalities. In ECG DL, researchers often use convolutional neural networks (CNNs) and traditionally use the full duration of raw ECG waveforms that create redundancies in feature learning and result in inaccurate predictions with large uncertainties. Objective: For enhancing these predictions, we introduced a sub-waveform representation that leverages the rhythmic pattern of ECG waveforms (data-centric approach) rather than changing the CNN architecture (model-centric approach). Results: We applied the proposed representation to a population with 92,446 patients to identify left ventricular dysfunction. We found that the sub-waveform representation increases the performance metrics compared to the full-waveform representation. We observed a 2% increase for area under the receiver operating characteristic curve and 10% increase for area under the precision-recall curve. We also carefully examined three reliability components of explainability, interpretability, and fairness. We provided an explanation for enhancements obtained by heartbeat alignment mechanism. By developing a new scoring system, we interpreted the clinical relevance of ECG features and showed that sub-waveform representation further pushes the scores towards clinical predictions. Finally, we showed that the new representation significantly reduces prediction uncertainties within subgroups that contributes to individual fairness. Conclusion: We expect that this added control over the granularity of ECG data will improve the DL modeling for new artificial intelligence technologies in the cardiovascular space.

Supervised Pretraining through Contrastive Categorical Positive Samplings to Improve COVID-19 Mortality Prediction.

Clinical EHR data is naturally heterogeneous, where it contains abundant sub-phenotype. Such diversity creates challenges for outcome prediction using a machine learning model since it leads to high intra-class variance. To address this issue, we propose a supervised pre-training model with a unique embedded k-nearest-neighbor positive sampling strategy. We demonstrate the enhanced performance value of this framework theoretically and show that it yields highly competitive experimental results in predicting patient mortality in real-world COVID-19 EHR data with a total of over 7,000 patients admitted to a large, urban health system. Our method achieves a better AUROC prediction score of 0.872, which outperforms the alternative pre-training models and traditional machine learning methods. Additionally, our method performs much better when the training data size is small (345 training instances).

Relational Learning Improves Prediction of Mortality in COVID-19 in the Intensive Care Unit.

Traditional Machine Learning (ML) models have had limited success in predicting Coronoavirus-19 (COVID-19) outcomes using Electronic Health Record (EHR) data partially due to not effectively capturing the inter-connectivity patterns between various data modalities. In this work, we propose a novel framework that utilizes relational learning based on a heterogeneous graph model (HGM) for predicting mortality at different time windows in COVID-19 patients within the intensive care unit (ICU). We utilize the EHRs of one of the largest and most diverse patient populations across five hospitals in major health system in New York City. In our model, we use an LSTM for processing time varying patient data and apply our proposed relational learning strategy in the final output layer along with other static features. Here, we replace the traditional softmax layer with a Skip-Gram relational learning strategy to compare the similarity between a patient and outcome embedding representation. We demonstrate that the construction of a HGM can robustly learn the patterns classifying patient representations of outcomes through leveraging patterns within the embeddings of similar patients. Our experimental results show that our relational learning-based HGM model achieves higher area under the receiver operating characteristic curve (auROC) than both comparator models in all prediction time windows, with dramatic improvements to recall.

Contrastive learning improves critical event prediction in COVID-19 patients.

Deep learning (DL) models typically require large-scale, balanced training data to be robust, generalizable, and effective in the context of healthcare. This has been a major issue for developing DL models for the coronavirus disease 2019 (COVID-19) pandemic, where data are highly class imbalanced. Conventional approaches in DL use cross-entropy loss (CEL), which often suffers from poor margin classification. We show that contrastive loss (CL) improves the performance of CEL, especially in imbalanced electronic health records (EHR) data for COVID-19 analyses. We use a diverse EHR dataset to predict three outcomes: mortality, intubation, and intensive care unit (ICU) transfer in hospitalized COVID-19 patients over multiple time windows. To compare the performance of CEL and CL, models are tested on the full dataset and a restricted dataset. CL models consistently outperform CEL models, with differences ranging from 0.04 to 0.15 for area under the precision and recall curve (AUPRC) and 0.05 to 0.1 for area under the receiver-operating characteristic curve (AUROC).

Contrastive Learning Improves Critical Event Prediction in COVID-19 Patients.

Machine Learning (ML) models typically require large-scale, balanced training data to be robust, generalizable, and effective in the context of healthcare. This has been a major issue for developing ML models for the coronavirus-disease 2019 (COVID-19) pandemic where data is highly imbalanced, particularly within electronic health records (EHR) research. Conventional approaches in ML use cross-entropy loss (CEL) that often suffers from poor margin classification. For the first time, we show that contrastive loss (CL) improves the performance of CEL especially for imbalanced EHR data and the related COVID-19 analyses. This study has been approved by the Institutional Review Board at the Icahn School of Medicine at Mount Sinai. We use EHR data from five hospitals within the Mount Sinai Health System (MSHS) to predict mortality, intubation, and intensive care unit (ICU) transfer in hospitalized COVID-19 patients over 24 and 48 hour time windows. We train two sequential architectures (RNN and RETAIN) using two loss functions (CEL and CL). Models are tested on full sample data set which contain all available data and restricted data set to emulate higher class imbalance.CL models consistently outperform CEL models with the restricted data set on these tasks with differences ranging from 0.04 to 0.15 for AUPRC and 0.05 to 0.1 for AUROC. For the restricted sample, only the CL model maintains proper clustering and is able to identify important features, such as pulse oximetry. CL outperforms CEL in instances of severe class imbalance, on three EHR outcomes with respect to three performance metrics: predictive power, clustering, and feature importance. We believe that the developed CL framework can be expanded and used for EHR ML work in general.

Federated Learning of Electronic Health Records to Improve Mortality Prediction in Hospitalized Patients With COVID-19: Machine Learning Approach.

BACKGROUND: Machine learning models require large datasets that may be siloed across different health care institutions. Machine learning studies that focus on COVID-19 have been limited to single-hospital data, which limits model generalizability. OBJECTIVE: We aimed to use federated learning, a machine learning technique that avoids locally aggregating raw clinical data across multiple institutions, to predict mortality in hospitalized patients with COVID-19 within 7 days. METHODS: Patient data were collected from the electronic health records of 5 hospitals within the Mount Sinai Health System. Logistic regression with L1 regularization/least absolute shrinkage and selection operator (LASSO) and multilayer perceptron (MLP) models were trained by using local data at each site. We developed a pooled model with combined data from all 5 sites, and a federated model that only shared parameters with a central aggregator. RESULTS: The LASSOfederated model outperformed the LASSOlocal model at 3 hospitals, and the MLPfederated model performed better than the MLPlocal model at all 5 hospitals, as determined by the area under the receiver operating characteristic curve. The LASSOpooled model outperformed the LASSOfederated model at all hospitals, and the MLPfederated model outperformed the MLPpooled model at 2 hospitals. CONCLUSIONS: The federated learning of COVID-19 electronic health record data shows promise in developing robust predictive models without compromising patient privacy.

Federated Learning of Electronic Health Records Improves Mortality Prediction in Patients Hospitalized with COVID-19.

Machine learning (ML) models require large datasets which may be siloed across different healthcare institutions. Using federated learning, a ML technique that avoids locally aggregating raw clinical data across multiple institutions, we predict mortality within seven days in hospitalized COVID-19 patients. Patient data was collected from Electronic Health Records (EHRs) from five hospitals within the Mount Sinai Health System (MSHS). Logistic Regression with L1 regularization (LASSO) and Multilayer Perceptron (MLP) models were trained using local data at each site, a pooled model with combined data from all five sites, and a federated model that only shared parameters with a central aggregator. Both the federated LASSO and federated MLP models performed better than their local model counterparts at four hospitals. The federated MLP model also outperformed the federated LASSO model at all hospitals. Federated learning shows promise in COVID-19 EHR data to develop robust predictive models without compromising patient privacy.