Self-Attention MHDNet: A Novel Deep Learning Model for the Detection of R-Peaks in the Electrocardiogram Signals...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Signal Processing

Background:

• Magnetic Resonance Imaging (MRI) is crucial for medical diagnosis and image-guided surgery.
• Electrocardiogram (ECG) data is vital for patient monitoring and gating during MRI scans.
• The Magnetohydrodynamic (MHD) effect in MRI scanners significantly distorts ECG signals, causing irregular heartbeats and hindering accurate QRS complex detection.

Purpose of the Study:

• To develop a reliable method for detecting R-peaks in ECG waveforms acquired within 3 Tesla (T) and 7T magnetic fields.
• To overcome the challenges posed by MHD-induced distortions in ECG signals during MRI procedures.
• To enhance the accuracy of cardiovascular functional MRI by improving the gating trigger pulse detection.

Main Methods:

• A novel deep learning model, Self-Attention MHDNet, was developed for R-peak detection.
• The model utilizes 1D-segmentation to process ECG signals corrupted by the MHD effect.
• The model was evaluated on ECG data acquired in both 3T and 7T MRI environments.

Main Results:

• Self-Attention MHDNet achieved a recall of 99.83% and precision of 99.68% in 3T magnetic fields.
• In 7T magnetic fields, the model demonstrated superior performance with a recall of 99.87% and precision of 99.78%.
• The proposed model effectively detects R-peaks despite significant MHD-induced ECG distortions.

Conclusions:

• The Self-Attention MHDNet model provides a robust solution for accurate R-peak detection in MHD-corrupted ECG signals.
• This advancement enables precise gating for cardiovascular functional MRI, even in high magnetic field environments.
• The model's high performance in 3T and 7T settings signifies its potential for clinical application in MRI-guided procedures.