What is the clinical evidence from this study?

Study design: Other. Population: Cardiovascular and neurological abnormalities (Arrhythmia, Myocardial Infarction, EEG abnormalities, ICU phenotyping). Intervention: DDxNet (Deep Learning Model) vs. Problem-specific state-of-the-art machine learning models. Primary outcome: Diagnostic performance (Accuracy, Recall, Precision, F1-score, AUROC).

What does this research mean for the field?

DDxNet, a generic deep learning architecture, achieves state-of-the-art performance across multiple clinical diagnostic tasks involving time-varying data (EHR, ECG, EEG), including 99.7% accuracy for myocardial infarction detection using single-lead ECG. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

October 2, 2020Open Access

DDxNet: a deep learning model for automatic interpretation of electronic health records, electrocardiograms and electroencephalograms

Key Result

DDxNet, a generic deep learning architecture, achieved state-of-the-art performance across multiple clinical diagnostic tasks, including 99.7% accuracy for myocardial infarction detection using single-lead ECG.

Structured PICO

Does the DDxNet deep learning architecture improve diagnostic accuracy across different clinical modalities (ECG, EEG, EHR) compared to problem-specific baseline models?

Population

Multiple benchmark datasets including TUH abnormal EEG corpus (n=2993 patients), MIT-BIH arrhythmia database (n=47 subjects), Physionet PTB MI database (n=200 subjects: 148 MI, 52 normal), and MIMIC-III ICU EHR database (n=33,798 unique patients).

Intervention

DDxNet, a deep learning architecture with densely connected causal convolutions and adaptive dilation for time-varying clinical data

Comparator

Off-the-shelf generic architectures (e.g., ResNet, Temporal Convolutional Networks) and problem-specific state-of-the-art baseline models

Outcome

Diagnostic performance (accuracy, recall, precision, F1-score) on abnormality detection, arrhythmia classification, myocardial infarction detection, and acute care phenotyping

DDxNet is a generic deep learning architecture that achieves state-of-the-art diagnostic performance across diverse clinical time-series data (ECG, EEG, EHR) without requiring problem-specific architectural tuning.

Limitations

Retrospective classification without exact timestamps for billing codes in EHR data
Lack of uncertainty quantification
Need to better understand model behavior under domain shifts and ensure fairness

Abstract

Effective patient care mandates rapid, yet accurate, diagnosis. With the abundance of non-invasive diagnostic measurements and electronic health records (EHR), manual interpretation for differential diagnosis has become time-consuming and challenging. This has led to wide-spread adoption of AI-powered tools, in pursuit of improving accuracy and efficiency of this process. While the unique challenges presented by each modality and clinical task demand customized tools, the cumbersome process of making problem-specific choices has triggered the critical need for a generic solution to enable rapid development of models in practice. In this spirit, we develop DDxNet, a deep architecture for time-varying clinical data, which we demonstrate to be well-suited for diagnostic tasks involving different modalities (ECG/EEG/EHR), required level of characterization (abnormality detection/phenotyping) and data fidelity (single-lead ECG/22-channel EEG). Using multiple benchmark problems, we show that DDxNet produces high-fidelity predictive models, and sometimes even provides significant performance gains over problem-specific solutions.

AI에게 질문

Bookmark

View Full Paper

AI에게 질문

Bookmark

View Full Paper

Cite This Study

Thiagarajan et al. (Fri,) conducted a other in Cardiovascular and neurological abnormalities (Arrhythmia, Myocardial Infarction, EEG abnormalities, ICU phenotyping). DDxNet (Deep Learning Model) vs. Problem-specific state-of-the-art machine learning models was evaluated on Diagnostic performance (Accuracy, Recall, Precision, F1-score, AUROC). DDxNet, a generic deep learning architecture, achieved state-of-the-art performance across multiple clinical diagnostic tasks, including 99.7% accuracy for myocardial infarction detection using single-lead ECG.

synapsesocial.com/papers/6a154d6ba2f71238514e48c6 https://doi.org/https://doi.org/10.1038/s41598-020-73126-9

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: