TE-ESN: Time Encoding Echo State Network for Prediction Based on Irregularly Sampled Time Series Data

Sun, C. P.; Hong, Shenda; Song, Moxian; Chou, Yen-Hsiu; Sun, Yongyue; Cai, Derun; Li, Hongyan

doi:10.24963/ijcai.2021/414

Cited by 17 publications

(12 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, critically ill patients are often hospitalized for several months, and records often have hundreds of observations. Due to the change in patient's health status, the relevant measurement requirements are also changing, which may be several hours or days apart [25]. Thus, to model long-term dependency and eliminate the impact of uneven time intervals, we implemented Time-aware Long Short-Term Memory (T-LSTM) [26].…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Continuous Diagnosis and Prognosis by Controlling the Update Process of Deep Neural Networks

Sun¹,

Li²,

Song³

et al. 2022

Preprint

View full text Add to dashboard Cite

Continuous diagnosis and prognosis are essential for intensive care patients. It can provide more opportunities for timely treatment and rational resource allocation, especially for sepsis, a main cause of death in ICU, and COVID-19, a new worldwide epidemic. Although deep learning methods have shown their great superiority in many medical tasks, they tend to catastrophically forget, over fit, and get results too late when performing diagnosis and prognosis in the continuous mode. In this work, we summarized the three requirements of this task, proposed a new concept, continuous classification of time series (CCTS), and designed a novel model training method, restricted update strategy of neural networks (RU). In the context of continuous prognosis, our method outperformed all baselines and achieved the

show abstract

Section: Resultsmentioning

confidence: 99%

“…For example, early diagnosis helps for sepsis outcomes [57]. Nowadays, Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) have shown good performances for CTS and ECTS by modeling longterm dependencies [58], addressing data irregularities [25], learning frequency features [59], etc.…”

Section: A1 Single-shot Classificationmentioning

confidence: 99%

Continuous Diagnosis and Prognosis by Controlling the Update Process of Deep Neural Networks

Sun¹,

Li²,

Song³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…One limitation of modeling patient data in discretized time intervals using attention is that it may not capture the full complexity of the data. In many cases, medical data is recorded at irregular intervals, which can provide deeper insight into the progression of a disease (41). Additionally, discretizing the time intervals may make it more difficult to model the autoregressive nature of the data, where future time points are predicted based on previous ones.…”

Section: Discussionmentioning

confidence: 99%

hART: Deep Learning-Informed Lifespan Heart Failure Risk Trajectories

Moroz,

Li,

Marelli

2023

Preprint

View full text Add to dashboard Cite

Heart failure (HF) results in persistent risk and long-term comorbidities. This is particularly true for patients with lifelong HF sequelae of cardiovascular disease such as patients with congenital heart disease (CHD). We developedhART(heart failure Attentive Risk Trajectory), a deep-learning model to predict HF trajectories in CHD patients.hARTis designed to capture the contextual relationships between medical events within a patient’s history. It is trained to predict future HF risk by using the masked self-attention mechanism that forces it to focus only on the most relevant segments of the past medical events. To demonstrate the utility ofhART, we used a large cohort containing healthcare administrative data from the Quebec CHD database (137,493 patients, 35-year follow-up).hARTachieves an area under the precision-recall of 28% for HF risk prediction, which is 33% improvement over existing methods. Patients with severe CHD lesion showed a consistently elevated predicted HF risks throughout their lifespan, and patients with genetic syndromes exhibited elevated HF risks until the age of 50. The impact of the birth condition decreases on long-term HF risk. The timing of interventions such as arrhythmia surgery had varying impacts on the lifespan HF risk among the individuals. Arrhythmic surgery performed at a younger age had minimal long-term effects on HF risk, while surgeries during adulthood had a significant lasting impact. Together, we show thathARTcan detect meaningful lifelong HF risk in CHD patients by capturing both long and short-range dependencies in their past medical events. Our code is available athttps://github.com/li-lab-mcgill/hART-heart-failure-Attentive-Risk-Trajectory.

show abstract

“…With the advancement of medical technology, patients in the Intensive Care Unit (ICU) are monitored by different instruments at the bedside that measure different vital signals Sun et al (2021) about the patient’s health. Such as heart rate, systolic blood pressure, temperature, etc.…”

Section: Introductionmentioning

confidence: 99%

“…For example, Darabi et al (2020) used transformer networks and the recently proposed BERT language model to embed these clinical record data streams into a unified vector representation for downstream mortality prediction tasks. However, in a real-world clinical setting, time series data can be missing due to different collection frequencies Sun et al (2020) , and thus time series data may be incomplete. On the other hand, clinical records may suffer from spelling errors, non-standard abbreviations, and different writing styles Qiao et al (2019) , making it difficult to use clinical annotations for prediction.…”

Section: Introductionmentioning

confidence: 99%

Deep multi-modal intermediate fusion of clinical record and time series data in mortality prediction

Niu

Zhang²,

Peng³

et al. 2023

Front. Mol. Biosci.

View full text Add to dashboard Cite

In intensive care units (ICUs), mortality prediction is performed by combining information from these two sources of ICU patients by monitoring patient health. Respectively, time series data generated from each patient admission to the ICU and clinical records consisting of physician diagnostic summaries. However, existing mortality prediction studies mainly cascade the multimodal features of time series data and clinical records for prediction, ignoring thecross-modal correlation between the underlying features in different modal data. To address theseissues, we propose a multimodal fusion model for mortality prediction that jointly models patients’ time-series data as well as clinical records. We apply a fine-tuned Bert model (Bio-Bert) to the patient’s clinical record to generate a holistic embedding of the text part, which is then combined with the output of an LSTM model encoding the patient’s time-series data to extract valid features. The global contextual information of each modal data is extracted using the improved fusion module to capture the correlation between different modal data. Furthermore, the improved fusion module can be easily added to the fusion features of any unimodal network and utilize existing pre-trained unimodal model weights. We use a real dataset containing 18904 ICU patients to train and evaluate our model, and the research results show that the representations obtained by themodel can achieve better prediction accuracy compared to the baseline.

show abstract

TE-ESN: Time Encoding Echo State Network for Prediction Based on Irregularly Sampled Time Series Data

Cited by 17 publications

References 1 publication

Continuous Diagnosis and Prognosis by Controlling the Update Process of Deep Neural Networks

Continuous Diagnosis and Prognosis by Controlling the Update Process of Deep Neural Networks

hART: Deep Learning-Informed Lifespan Heart Failure Risk Trajectories

Deep multi-modal intermediate fusion of clinical record and time series data in mortality prediction

Contact Info

Product

Resources

About