2 March 2026 in IEEE Journal of Biomedical and Health Informatics
Electrocardiogram (ECG) analysis represents a promising field for deep learning applications in clinical diagnostics. However, practical use of current methods is still constrained by their heavy reliance on large amounts of labeled data, as well as limitations in processing efficiency and signal quality. To address these challenges, we present BioFPT (Biosignal Feature Pyramid Transformer), a novel self-supervised learning framework designed for ECG signal. The proposed framework incorporates a Split Mask-Join (SMJ) transformation for a Pre-training strategy, complemented by an overlapping embedding mechanism that eliminates positional encoding requirements. The efficiency of the architectural design is enhanced through a Spatial Reduction Attention (SRA) transformer, which achieves a reduction in computational complexity without performance degradation. Comprehensive evaluation of seven public ECG datasets comprising over 94,000 sub jects demonstrates BioFPT's effectiveness, an accuracy improvement of 4.2% and a parameter reduction of 14.8% compared to state-of-the-art models. Furthermore, it maintains robust performance across diverse pathological conditions and signal qualities. The proposed architecture rep resents a significant advancement in self-supervised ECG analysis, particularly suitable for scenarios with limited labeled data availability. Moreover, its versatile architecture shows promise for broader applications across various biosignals.
10 July 2024 in IEEE Sensors Journal
Pulmonary hypertension (PH) is a heterogeneous and highly morbid disease, diagnosed with mean pulmonary artery pressure (PAP) measured by invasive right heart catheterization (RHC). However, RHC risks complications and cannot provide continuous PAP monitoring over long durations. Non-invasive assessment of PAP has gained interest for its low risk and convenience compared to invasive measurement. Recently, seismocardiogram (SCG), which unobtrusively records chest vibrations from cardiovascular motions, has demonstrated feasibility for PAP estimation. In this paper, we propose an attentive conditional adversarial generative network (CGAN) framework for the purpose of continuous PAP assessment using SCG. This is the first time that the PAP waveforms morphology has been accurately reconstructed from SCG. The proposed framework involves a U-net generator with a soft attention module to focus on salient SCG features, along with a PatchGAN discriminator for localized assessment. It uses Wasserstein distance in the loss function to improve training stability, which is compared with existing generative models using the same SCG-RHC dataset. Results show that our model yields superior PAP waveform reconstruction performance with Pearson correlation coefficient (PCC) of 0.982. The reconstructed waveforms exhibit high correlation with invasive RHC measurement, with quantitative statistics and physiological indices validating strong agreement. This work represents a novel approach for continuous, non-invasive PAP estimation, offering valuable improvements in the management of PH.