Reconstruction of central arterial pressure waveform based on CBL-iTransformer model from radial arterial pressure waveform

The central arterial pressure (CAP) serves as a crucial parameter for assessing cardiovascular health and evaluating the risk of related diseases. Its non-invasive, continuous, and accurate reconstruction is crucial for the evaluation and prevention of cardiovascular diseases. However, traditional approaches often exhibit limited accuracy, while certain deep learning models face challenges in feature extraction, which limits their widespread use and clinical adoption. This study introduces a novel waveform reconstruction model (CBL-iTransformer), which is built upon an enhanced iTransformer architecture. It integrates bidirectional long short-term memory networks (BiLSTM) and convolutional neural networks (CNN) to augment the feature extraction efficiency and precision during the waveform reconstruction process. The performance of the CBL-iTransformer model in reconstructing the CAP waveform is discussed and validated using radial arterial pressure (RAP) waveform and CAP waveform data obtained from 62 patients who underwent invasive measurements before and after medication. The evaluation metrics, including the mean absolute error (MAE) and root mean square error (RMSE), are compared against tranditional method as well as a range of deep learning models. The research findings demonstrate that the model achieves a robust reconstruction of the CAP waveform ( : 0.93 ± 0.90 mmHg , : 1.29 ± 0.90 mmHg), while also yielding reliable reconstruction results for the central aortic systolic pressure (CASP) and central aortic diastolic pressure (CADP) waveforms ( : 1.44 ± 0.84 mmHg , : 1.30 ± 0.78 mmHg). The CBL-iTransformer model demonstrates excellent CAP reconstruction performance and is expected to be applied to clinical practice in the future.