An Improved UNet++ Model for Congestive Heart Failure Diagnosis Using Short-Term RR Intervals.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Congestive heart failure (CHF) is a complex syndrome often diagnosed late due to ventricular dysfunction.
• Heart rate variability (HRV) shows promise as a prognostic indicator for CHF.
• Early detection of CHF remains a significant clinical challenge.

Purpose of the Study:

• To develop and evaluate an end-to-end deep learning model for the early detection of CHF using HRV signals.
• To enhance the UNet++ architecture with Squeeze-and-Excitation (SE) residual blocks for improved feature extraction.
• To assess the model's performance across different lengths of RR interval data.

Main Methods:

• An end-to-end encoder-decoder deep learning model, inspired by 2-D UNet++, was developed.
• The model incorporates Squeeze-and-Excitation (SE) residual blocks for hierarchical deep feature extraction.
• Performance was evaluated on two open-source databases using 500, 1000, and 2000 RR intervals.

Main Results:

• The proposed deep learning model achieved high accuracy in detecting CHF.
• Accuracy rates of 85.64%, 86.65%, and 88.79% were recorded with 500, 1000, and 2000 RR intervals, respectively.
• The model effectively distinguished between CHF patients and normal subjects.

Conclusions:

• Deep learning-based HRV analysis is a valuable tool for the early detection of CHF.
• The developed model demonstrates potential to assist clinicians in timely and accurate CHF diagnosis.
• Further research into deep learning applications for cardiovascular disease detection is warranted.