首页|期刊导航|数字中医药(英文)|面向轻量化血压预测的任务导向重采样脉搏波压缩

面向轻量化血压预测的任务导向重采样脉搏波压缩OA

Task-oriented compression of resampled pulse waves for lightweight blood pressure prediction

中文摘要英文摘要

目的 提出一种任务导向的无创血压预测信号压缩范式,通过重采样消除人工特征工程,实现高性能且适用于嵌入式系统的轻量化模型.方法 采用配备压电传感器的 PDA-1脉诊仪于 2021 年 11 月 20 日至2022 年 11月 19 日在上海中医药大学附属曙光医院采集健康志愿者的桡动脉脉搏波.从每段 30 秒记录中提取最稳定的 8秒片段(包含多个完整脉搏周期).将原始信号(200 Hz)通过傅里叶变换、小波变换(db9)和三次样条插值重采样至 40 Hz.构建 3 组特征集:基线特征集(重采样数据和人口统计特征)、融合特征集(基线特征集加上手工提取的时域特征)、手工特征集(手工提取的特征和人口统计特征).采用贝叶斯优化进行超参数调优,建立 XGBoost和随机森林模型,并用十折交叉验证进行评估.依据英国高血压学会(BHS)和美国医疗器械促进协会(AAMI)标准进行预测准确性评估,重建质量以百分比均方根差(PRD)量化.采用相应的统计方法进行模型评估及不同配置间的多重比较.通过夏普利加性解释(SHAP)方法、频谱能量分布分析及小波多分辨率分解进行可视化与可解释性分析.结果 共纳入 348名健康志愿者.基于重采样数据构建的 XGBoost和随机森林模型均达到具有临床竞争力的性能水平,其中 XG-Boost 在准确性与计算效率方面表现更优.采用傅里叶变换重采样与基线特征集的 XGBoost模型在测试集上收缩压平均绝对误差(MAE)为(6.01±0.94)mmHg,舒张压为(4.60±0.34)mmHg,收缩压达到 BHS B级,舒张压达到 BHS A级,并符合 AAMI标准.三种重采样方法的预测准确性无统计学显著差异(P>0.05);然而其重建误差差异显著:小波变换误差最高(PRD=35.52%),其次为傅里叶变换(PRD=8.87%),插值最低(PRD=1.33%).与基线特征集相比,融合特征集仅将收缩压 MAE降低 0.09 mmHg(校正 P>0.05),同时 XGBoost训练时间增加 92.5%至 247.5%(P<0.001).手工特征集与基线集预测性能相当,表明在采用重采样信号时,人工特征工程是冗余的.结论 本研究验证了压缩信号用于临床级预测的充分性以及手工特征工程的冗余性.将面向任务的压缩与机器学习相结合,有望支持轻量型可穿戴连续血压监测设备的研发.

Objective This study proposes a task-oriented signal compression paradigm for non-inva-sive blood pressure(BP)prediction,eliminating manual feature engineering through resam-pling to achieve a high-performance lightweight model suitable for embedded systems. Methods The PDA-1 pulse diagnostic device equipped with a piezoelectric sensor was used to collect radial artery pulse waves from healthy volunteers at Shuguang Hospital Affiliated to Shanghai University of Traditional Chinese Medicine between November 20,2021 and November 19,2022.The most stable 8 s segment containing multiple complete pulse cycles was extracted from each 30 s recording.Original signals(200 Hz)were resampled to 40 Hz us-ing Fourier transform,wavelet transform(db9),and cubic spline interpolation.Three feature sets were constructed:baseline(resampled data and demographics),fusion(baseline plus handcrafted time-domain features),and handcrafted(handcrafted features and demograph-ics)feature sets.XGBoost and random forest(RF)models were developed with Bayesian opti-mization for hyperparameter tuning,and evaluated using 10-fold cross-validation.Prediction accuracy was assessed against British Hypertension Society(BHS)and Association for the Ad-vancement of Medical Instrumentation(AAMI)standards,while reconstruction quality was quantified by percentage root-mean-square difference(PRD).Model evaluation and multiple comparisons among different configurations were conducted using the corresponding statis-tical methods.Visual and interpretability analyses were performed using SHapley Additive ex-Planations(SHAP),frequency-energy distribution analysis,and wavelet multi-resolution de-composition. Results A total of 348 healthy volunteers were enrolled.XGBoost and RF models built on re-sampled data achieved clinically competitive performance,with XGBoost demonstrating su-perior accuracy and computational efficiency.The XGBoost model employing Fourier trans-form resampling and the baseline feature set achieved a mean absolute error(MAE)of(6.01±0.94)mmHg for systolic blood pressure(SBP)and(4.60±0.34)mmHg for diastolic blood pressure(DBP)on the test set,meeting BHS Grade B for SBP and BHS Grade A for DBP,as well as AAMI standards.No statistically significant differences in predictive accuracy were found among the three resampling methods(P>0.05);however,their reconstruction errors differed markedly:wavelet transform exhibited the highest error(PRD=35.52%),followed by Fourier transform(PRD=8.87%),and interpolation the lowest(PRD=1.33%).Compared with the baseline feature set,the fused feature set reduced SBP MAE by only 0.09 mmHg(correct-ed P>0.05)while increasing XGBoost training time by 92.5%-247.5%(P<0.001).The handcrafted feature set achieved comparable predictive performance to the baseline,indicat-ing that manual feature engineering is redundant when resampled signals are employed. Conclusion This study validated the sufficiency of compressed signals for clinical-grade pre-diction and the redundancy of handcrafted feature engineering.Integrating task-oriented compression with machine learning may support the development of lightweight wearable continuous BP monitoring devices.

万小龙;李克松;孙晓军;崔骥;许家佗;胡晓娟

上海中医药大学中医药人工智能学院,上海 201203,中国上海中医药大学中医药人工智能学院,上海 201203,中国上海中医药大学中医学院,上海 201203,中国上海中医药大学中医学院,上海 201203,中国上海中医药大学中医学院,上海 201203,中国上海中医药大学中医药人工智能学院,上海 201203,中国

脉搏波任务导向重采样方法特征工程血压预测轻量化

Pulse wavesTask-orientedResampling methodFeature engineeringBlood pressure predictionLightweight

《数字中医药(英文)》 2026 (3)

344-361,18

National Science and Technology Major Project of China(2025ZD0545000),High-level Traditional Chinese Medi-cine Key Discipline Construction Project of the National Administration of Traditional Chinese Medicine:Tradi-tional Chinese Medicine Diagnostics(ZYYZDXK2023069),and Shanghai Key Discipline Construction Project of Tra-ditional Chinese Medicine(Clinical)(shzyyzdxk-2024207).

10.1016/j.dcmed.2026.08.002

评论