An official website of the United States government
Abstract Sudden cardiac death from arrhythmia is a major cause of mortality worldwide. In this study, we developed a novel deep learning (DL) approach that blends neural networks and survival analysis to predict patient-specific survival curves from contrast-enhanced cardiac magnetic resonance images and clinical covariates for patients with ischemic heart disease. The DL-predicted survival curves offer accurate predictions at times up to 10 years and allow for estimation of uncertainty in predictions. The performance of this learning architecture was evaluated on multi-center internal validation data and tested on an independent test set, achieving concordance indexes of 0.83 and 0.74 and 10-year integrated Brier scores of 0.12 and 0.14. We demonstrate that our DL approach, with only raw cardiac images as input, outperforms standard survival models constructed using clinical covariates. This technology has the potential to transform clinical decision-making by offering accurate and generalizable predictions of patient-specific survival probabilities of arrhythmic death over time.
more »
« less
Neurological Prognostication of Post-Cardiac-Arrest Coma Patients Using EEG Data: A Dynamic Survival Analysis Framework with Competing Risks
Patients resuscitated from cardiac arrest who enter a coma are at high risk of death. Forecasting neurological outcomes of these patients (i.e., the task of neurological prognostication) could help with treatment decisions: which patients are likely to awaken from their coma and should be kept on life-sustaining therapies, and which are so ill that they would unlikely benefit from treatment? In this paper, we propose, to the best of our knowledge, the first dynamic framework for neurological prognostication of post-cardiac-arrest comatose patients using EEG data: our framework makes predictions for a patient over time as more EEG data become available, and different training patients’ available EEG time series could vary in length. Predictions themselves are phrased in terms of either time-to-event outcomes (time-to-awakening or time-to-death) or as the patient’s probability of awakening or of dying across multiple time horizons (e.g., within the next 24, 48, or 72 hours). Our framework is based on using any dynamic survival analysis model that supports competing risks in the form of estimating patient-level cumulative incidence functions. We consider three competing risks as to what happens first to a patient: awakening, being withdrawn from life-sustaining therapies (and thus deterministically dying), or dying (by other causes). For some patients, we do not know which of these happened first since they were still in a coma when data collection stopped (i.e., their outcome is censored). Competing risks models readily accommodate such patients. We demonstrate our framework by benchmarking three existing dynamic survival analysis models that support competing risks on a real dataset of 922 post-cardiac-arrest coma patients. Our main experimental findings are that: (1) the classical Fine and Gray model which only uses a patient’s static features and summary statistics from the patient’s latest hour’s worth of EEG data is highly competitive, achieving accuracy scores as high as the recently developed Dynamic-DeepHit model that uses substantially more of the patient’s EEG data; and (2) in an ablation study, we show that our choice of modeling three competing risks results in a model that is at least as accurate while learning more information than simpler models (using two competing risks or a standard survival analysis setup with no competing risks).
more »
« less
- Award ID(s):
- 2047981
- PAR ID:
- 10514114
- Publisher / Repository:
- Proceedings of Machine Learning Research
- Date Published:
- Journal Name:
- Proceedings of the 8th Machine Learning for Healthcare Conference
- ISSN:
- 2640-3498
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
ABSTRACT Breast cancer patients may experience relapse or death after surgery during the follow‐up period, leading to dependent censoring of relapse. This phenomenon, known as semi‐competing risk, imposes challenges in analyzing treatment effects on breast cancer and necessitates advanced statistical tools for unbiased analysis. Despite progress in estimation and inference within semi‐competing risks regression, its application to causal inference is still in its early stages. This article aims to propose a frequentist and semi‐parametric framework based on copula models that can facilitate valid causal inference, net quantity estimation and interpretation, and sensitivity analysis for unmeasured factors under right‐censored semi‐competing risks data. We also propose novel procedures to enhance parameter estimation and its applicability in practice. After that, we apply the proposed framework to a breast cancer study and detect the time‐varying causal effects of hormone‐ and radio‐treatments on patients' relapse and overall survival. Moreover, extensive numerical evaluations demonstrate the method's feasibility, highlighting minimal estimation bias and reliable statistical inference.more » « less
-
Abstract Objective.We aim to assess the severity of spatial neglect (SN) through detailing patients’ field of view (FOV) using EEG. Spatial neglect, a prevalent neurological syndrome in stroke patients, typically results from unilateral brain injuries, leading to inattention to the contralesional space. Commonly used Neglect detection methods like the Behavioral Inattention Test—conventional lack the capability to assess the full extent and severity of neglect. Although the Catherine Bergego Scale provides valuable clinical information, it does not detail the specific FOV affected in neglect patients.Approach.Building on our previously developed EEG-based brain–computer interface system, AR-guided EEG-based neglect detection, assessment, and rehabilitation system (AREEN), we aim to map neglect severity across a patient’s FOV. We have demonstrated that AREEN can assess neglect severity in a patient-agnostic manner. However, its effectiveness in patient-specific scenarios, which is crucial for creating a generalizable plug-and-play system, remains unexplored. This paper introduces a novel EEG-based combined spatio-temporal network (ESTNet) that processes both time and frequency domain data to capture essential frequency band information associated with SN. We also propose a FOV correction system using Bayesian fusion, leveraging AREEN’s recorded response times for enhanced accuracy by addressing noisy labels within the dataset.Main results.Extensive testing of ESTNet on our proprietary dataset has demonstrated its superiority over benchmark methods, achieving 79.62% accuracy, 76.71% sensitivity, and 86.36% specificity. Additionally, we provide saliency maps to enhance model explainability and establish clinical correlations.Significance.These findings underscore ESTNet’s potential combined with Bayesian fusion-based FOV correction as an effective tool for generalized neglect assessment in clinical settings.more » « less
-
An immunotherapy trial often uses the phase I/II design to identify the optimal biological dose, which monitors the efficacy and toxicity outcomes simultaneously in a single trial. The progression-free survival rate is often used as the efficacy outcome in phase I/II immunotherapy trials. As a result, patients developing disease progression in phase I/II immunotherapy trials are generally seriously ill and are often treated off the trial for ethical consideration. Consequently, the happening of disease progression will terminate the toxicity event but not vice versa, so the issue of the semi-competing risks arises. Moreover, this issue can become more intractable with the late-onset outcomes, which happens when a relatively long follow-up time is required to ascertain progression-free survival. This paper proposes a novel Bayesian adaptive phase I/II design accounting for semi-competing risks outcomes for immunotherapy trials, referred to as the dose-finding design accounting for semi-competing risks outcomes for immunotherapy trials (SCI) design. To tackle the issue of the semi-competing risks in the presence of late-onset outcomes, we re-construct the likelihood function based on each patient's actual follow-up time and develop a data augmentation method to efficiently draw posterior samples from a series of Beta-binomial distributions. We propose a concise curve-free dose-finding algorithm to adaptively identify the optimal biological dose using accumulated data without making any parametric dose–response assumptions. Numerical studies show that the proposed SCI design yields good operating characteristics in dose selection, patient allocation, and trial duration.more » « less
-
Abstract Burst suppression is a phenomenon in which the electroencephalogram (EEG) of a pharmacologically sedated patient alternates between higher frequency and amplitude bursting to lower frequency and amplitude suppression. The level of sedation can be quantified by the burst suppression ratio (BSR), which is defined as the amount of time that an EEG is suppressed over the amount of time measured. Maintaining a stable BSR in patients is an important clinical problem, which has led to an interest in the development of BSR-based closed-loop pharmacological control systems. Methods to address the problem typically involve pharmacokinetic (PK) modeling that describes the dynamics of drug infusion in the body as well as signal processing methods for estimating burst suppression from EEG measurements. In this regard, simulations, physiological modeling, and control design can play a key role in producing a solution. This paper seeks to add to prior work by incorporating a Schnider PK model with the Wilson–Cowan neural mass model to use as a basis for robust control design of biophysical burst suppression dynamics. We add to this framework actuator modeling, real-time burst suppression estimation, and uncertainty modeling in both the patient's physical characteristics and neurological phenomena to form a closed-loop system. Using the Ziegler–Nichols tuning method for proportional-integral-derivative (PID) control, we were able to control this system at multiple BSR set points over a simulation time of 18 h in both nominal, patient varied with noise added and with reduced performance due to BSR bounding when including patient variation and noise as well as neurological uncertainty.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
