Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Heart Failure with Attention Learning Techniques Utilizing Cardiovascular Data

Ershadul Haque, Manoranjan Paul, Faranak Tohidi

TL;DR

This work tackles heart failure prediction from cardiovascular EHR data by introducing an attention-based predictor built on a Transformer-like encoder–decoder. By focusing on key features such as serum creatinine and ejection fraction, and by systematically tuning multiple optimizers and learning rates, the study demonstrates competitive performance relative to LSTM approaches and highlights the importance of learning-rate selection for different predictive targets. The proposed attention architecture leverages multi-head attention and residual normalization to mitigate overfitting and improve feature integration from longitudinal cardiovascular data. The findings underscore the potential of attention-based models for HF risk stratification in clinical settings, while the conclusion section unexpectedly references a quantum image representation approach, indicating a possible content mix-up in the document.

Abstract

Cardiovascular diseases (CVDs) encompass a group of disorders affecting the heart and blood vessels, including conditions such as coronary artery disease, heart failure, stroke, and hypertension. In cardiovascular diseases, heart failure is one of the main causes of death and also long-term suffering in patients worldwide. Prediction is one of the risk factors that is highly valuable for treatment and intervention to minimize heart failure. In this work, an attention learning-based heart failure prediction approach is proposed on EHR(electronic health record) cardiovascular data such as ejection fraction and serum creatinine. Moreover, different optimizers with various learning rate approaches are applied to fine-tune the proposed approach. Serum creatinine and ejection fraction are the two most important features to predict the patient's heart failure. The computational result shows that the RMSProp optimizer with 0.001 learning rate has a better prediction based on serum creatinine. On the other hand, the combination of SGD optimizer with 0.01 learning rate exhibits optimum performance based on ejection fraction features. Overall, the proposed attention learning-based approach performs very efficiently in predicting heart failure compared to the existing state-of-the-art such as LSTM approach.

Predicting Heart Failure with Attention Learning Techniques Utilizing Cardiovascular Data

TL;DR

This work tackles heart failure prediction from cardiovascular EHR data by introducing an attention-based predictor built on a Transformer-like encoder–decoder. By focusing on key features such as serum creatinine and ejection fraction, and by systematically tuning multiple optimizers and learning rates, the study demonstrates competitive performance relative to LSTM approaches and highlights the importance of learning-rate selection for different predictive targets. The proposed attention architecture leverages multi-head attention and residual normalization to mitigate overfitting and improve feature integration from longitudinal cardiovascular data. The findings underscore the potential of attention-based models for HF risk stratification in clinical settings, while the conclusion section unexpectedly references a quantum image representation approach, indicating a possible content mix-up in the document.

Abstract

Cardiovascular diseases (CVDs) encompass a group of disorders affecting the heart and blood vessels, including conditions such as coronary artery disease, heart failure, stroke, and hypertension. In cardiovascular diseases, heart failure is one of the main causes of death and also long-term suffering in patients worldwide. Prediction is one of the risk factors that is highly valuable for treatment and intervention to minimize heart failure. In this work, an attention learning-based heart failure prediction approach is proposed on EHR(electronic health record) cardiovascular data such as ejection fraction and serum creatinine. Moreover, different optimizers with various learning rate approaches are applied to fine-tune the proposed approach. Serum creatinine and ejection fraction are the two most important features to predict the patient's heart failure. The computational result shows that the RMSProp optimizer with 0.001 learning rate has a better prediction based on serum creatinine. On the other hand, the combination of SGD optimizer with 0.01 learning rate exhibits optimum performance based on ejection fraction features. Overall, the proposed attention learning-based approach performs very efficiently in predicting heart failure compared to the existing state-of-the-art such as LSTM approach.
Paper Structure (11 sections, 5 figures)

This paper contains 11 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Related Works
  3. Proposed methodology
  4. Model architecture
  5. Encoder and decoder block
  6. Attention function
  7. Result and discussion
  8. Serum creatinine and ejection fraction-based prediction using RMSProp, SGD, and Adadelta optimizer with 0.01,0.001, and 0.0001 learning rate
  9. Ejection fraction-based prediction using RMSprop, SGD, Adam optimizer with 0.01, 0.001 and 0.001 learning rates
  10. Serum creatinine-based prediction using Adadelta and Adam's optimizer based on 0.01, 0.001, and 0.0001 learning rates
  11. conclusion

Figures (5)

  • Figure 1: Statistic on estimated death in 2021 from circulatory and heart diseases.heart2012
  • Figure 2: Proposed approach architecture
  • Figure 3: Prediction result analysis of the proposed approach using RMSProp, SGD, and Adadelta optimizer with 0.01, 0.001, and 0.0001 learning rates based on serum creatinine and ejection fraction.
  • Figure 4: Prediction result analysis of the proposed approach using RMSProp, SGD, and Adam optimizer based on ejection fraction using 0.1, 0.001, and 0.0001 learning rates.
  • Figure 5: Prediction result analysis of the proposed approach using Adadelta, Adam optimizer with different learning rates based on serum creatinine and age.