Xingguang Geng

This study integrated ancient Traditional Chinese Medicine (TCM) pulse diagnosis techniques with modern machine learning to advance contemporary medical diagnostics. A portable intelligent TCM pulse diagnostic device was developed using MEMS and CMOS technologies to collect detailed pulse waveform data from three groups: healthy females, females with coronary heart disease (CHD), and pregnant females in the first trimester. Three feature selection methods—Random Forest, Recursive Feature Elimination, and correlation analysis were adopted to identify key pulse features. Thirteen classification algorithms using various combinations of feature selection and machine learning approaches were evaluated. This study achieved a maximum classification accuracy of 80% using solely pulse data. Incorporating clinical parameters such as temperature and age significantly increased classification accuracy to 91%. The Gradient Boosting Decision Tree model with correlation analysis-based feature selection and Random Forest-based dimensionality reduction demonstrated the highest accuracy. Key pulse features identified included those related to the Liver Meridian, Pericardium Meridian, and Heart Meridian, which are crucial for differentiating the three groups. This study validated the clinical utility of pulse diagnostic devices when combined with machine learning algorithms. This study demonstrates concrete clinical utility for the pulse diagnostic device, while empirically validating foundational TCM concepts. It establishes a rigorous framework to advance evidence-driven modernization of pulse diagnosis protocols.

Objective: This study presents, for the first time, an ultrasound method combining focused and single plane wave transmissions to achieve single calibration, non-invasive, cuffless blood pressure monitoring. Methods: High-quality echo signals obtained from focused transmissions were first used to identify the optimal segment of the radial artery wall. Then, single plane wave signals were utilized to track this segment frame-by-frame and calculate local pulse wave velocity (PWV). Finally, the M-K equation was applied to estimate systolic and diastolic blood pressures. Results: To validate the effectiveness of the proposed algorithm, 25 volunteers participated in the study. Data were collected using both an oscillometric cuff-based blood pressure monitor and our ultrasound system in four states: morning, evening, pre-exercise, and post-exercise. A total of 300 sets of comparative data were obtained. The overall mean deviation for systolic blood pressure was 2.2 ± 2.1 mmHg, and for diastolic blood pressure, it was 2.1 ± 2.2 mmHg. Conclusion: To our knowledge, this is the first study to achieve real-time blood pressure calculation within seconds using a single-array ultrasound transducer. Compared to existing non-invasive continuous blood pressure monitoring methods, this approach is faster and provides more accurate vascular wall information. Significance: Combined with affordable hardware, this method holds significant potential for the development of wearable devices for real-time cardiovascular health monitoring.

The single-cycle pulse waves’ morphological characteristics can reflect rich physiological and pathological information about the human cardiovascular system, playing an important role in continuous blood pressure monitoring and cardiovascular disease risk warning. Pulse signals belong to weak physiological signals, and the acquisition process is easily affected by factors such as adjacent channel interference and motion artifacts, leading to abnormal cycles and affecting subsequent feature extraction. Therefore, based on support vector machines (SVM), a single-cycle pulse waveform recognition algorithm is proposed. Using a multi-channel pulse wave acquisition device to collect radial artery pulse signals from 150 subjects. Preprocess the signal to obtain a single-cycle pulse waveform. A total of 54 frequency domain and time-frequency domain feature parameters were extracted by using Fourier transform and wavelet transform to extract the frequency domain and time-frequency domain features. Use the principal components analysis (PCA) method to reduce parameter dimensions. Using the SVM for model training to achieve the classification of single-cycle effective pulse signals, pseudo noise and interference signals. The results show that the algorithm’s classification accuracy is 97.42%, indicating that the algorithm can effectively identify pulse signals, pseudo noise and interference signals, laying the foundation for modeling and suppressing interference signals in the next step.

Pulse signals carry comprehensive information regarding human cardiovascular physiology and pathology, providing a noninvasive and continuous method to assess cardiovascular health status in blood pressure monitoring. The blood pressure measurement method based on the pulse signal needs to extract the features of the single-cycle pulse signal, while the pulse signal pertains to the weak physiological signal of body surface. The acquisition process is susceptible to various factors leading to abnormal cycles, especially adjacent channel interference, affecting the subsequent feature extraction. To address this problem, this paper conducts an analysis of the formation mechanism of adjacent channel interference and proposes a single-cycle pulse signal recognition algorithm based on a one-dimensional deep convolutional neural network (1D-CNN) model. Radial pulse signals were collected from 150 subjects by pulse bracelet, and a dataset comprising 3446 single-cycle signals was extracted in total after denoising, single-cycle segmentation, and standardized preprocessing. The 1D-CNN model is trained to classify input signals into three categories: effective pulse signals, distortion, and interference signals. This classification is achieved by evaluating the waveform morphology of the signals within a single cycle. The results show that the overall classification accuracy of the algorithm on the test set is 98.26%, in which the classification accuracy of pulse waves is 99.8%, indicating that it can effectively recognize single-cycle pulse waves, which lays the foundation for subsequent continuous blood pressure measurement.

Delay multiply and sum beamforming (DMAS) is a non-linear method used in ultrasound imaging which offers superior performance to conventional delay and sum beamforming (DAS). While the combination of DMAS and coherence factor (CF) can further improve single plane-wave imaging lateral resolution, by using CF to weight the DMAS output, the main lobe width and aberration effects can be suppressed, which will improve the disadvantage of low lateral resolution when imaging with a single plane-wave. However, in low signal-to-noise ratio (SNR) environments, the speckle variance of the image increases, and there are black area artifacts around high echo objects. To improve the quality of the scatter without significantly reducing the lateral resolution of the DMAS-CF, this paper proposes an adaptive spatio-temporally smoothed coherence factor (GSTS-CF) combined with delay multiply and sum beamformer (DMAS + GSTS-CF), which uses the generalized coherence factor (GCF) as a local coherence detection tool to adaptively determine the subarray length to obtain an improved adaptive spatio-temporally smoothed factor, and uses this factor to weight the output of DMAS. The simulation and experimental data show that the proposed method improves lateral resolution (20 mm depth) by 86.87% compared to DAS, 52.13% compared to DMAS, 15.84% compared to DMAS + STS-CF, and has a full width at half maxima (FWHM) similar to DMAS-CF. The proposed method improves the speckle signal-to-noise ratio (sSNR) by 87.85% (simulation) and 77.84% (in carotid) compared to DMAS-CF, 20.37% (simulation) and 40.74% (in carotid) compared to DMAS, 15.03% (simulation) and 13.46% (in carotid) compared to DMAS + STS-CF, and has sSNR and scatter variance similar to DAS. This indicates that the method improves scatter quality (lower scatter variance and higher sSNR) without significantly reducing lateral resolution.

The radial artery reflects the largest amount of physiological and pathological information about the human body. However, ultrasound signal processing involves a large number of complex functions, and traditional digital signal processing can hardly meet the requirements of real-time processing of ultrasound data. The research aims to improve computational accuracy and reduce the hardware complexity of ultrasound signal processing systems. Firstly, this paper proposes to apply the coordinate rotation digital computer (CORDIC) algorithm to the whole radial artery ultrasound signal processing, combines the signal processing characteristics of each sub-module, and designs the dynamic filtering module based on the radix-4 CORDIC algorithm, the quadrature demodulation module based on the partitioned-hybrid CORDIC algorithm, and the dynamic range transformation module based on the improved scale-free CORDIC algorithm. A digital radial artery ultrasound imaging system was then built to verify the accuracy of the three sub-modules. The simulation results show that the use of the high-performance CORDIC algorithm can improve the accuracy of data processing. This provides a new idea for the real-time processing of ultrasound signals. Finally, radial artery ultrasound data were collected from 20 volunteers using different probe scanning modes at three reference positions. The vessel diameter measurements were averaged to verify the reliability of the CORDIC algorithm for radial artery ultrasound imaging, which has practical application value for computer-aided clinical diagnosis.

The accurate localization of S1 and S2 is essential for heart sound segmentation and classification. However, current direct heart sound segmentation algorithms have poor noise immunity and low accuracy. Therefore, this paper proposes a new optimal heart sound segmentation algorithm based on K-means clustering and Haar wavelet transform. The algorithm includes three parts. Firstly, this method uses the Viola integral method and Shannon’s energy-based algorithm to extract the function of the envelope of the heart sound energy. Secondly, the time–frequency domain features of the acquired envelope are extracted from different dimensions and the optimal peak is searched adaptively based on a dynamic segmentation threshold. Finally, K-means clustering and Haar wavelet transform are implemented to localize S1 and S2 of heart sounds in the time domain. After validation, the recognition rate of S1 reached 98.02% and that of S2 reached 96.76%. The model outperforms other effective methods that have been implemented. The algorithm has high robustness and noise immunity. Therefore, it can provide a new method for feature extraction and analysis of heart sound signals collected in clinical settings.

In pulse wave analysis, the changing curve of pulse wave strength with continuous increasing pressure, that is, the P-S (pressure-strength) curve, contains abundant human physiological information, but there is no accurate model to describe the formation mechanism of the curve. Therefore, this paper proposes a modeling method of the radial artery P-S curve based on the radial vibration of the vascular wall. The modeling method includes three parts. Firstly, based on hemodynamics, we proposed the blood motion equation in the pulsation process of healthy people. Secondly, the motion equation of the vascular wall based on the fluid–structure interaction between blood motion and vascular wall was established. Finally, according to the elastic theory of the vascular wall, the relationship between pulse strength and extravascular pressure of blood vessels was found. To verify the accuracy and applicability of the model, this paper simulated the changes in the vascular wall stress and the intravascular pressure with the extravascular pressure during the process of vascular deformation. In addition, 69 healthy volunteers were selected to participate in this study. Based on the gradient compression, the pulse strength envelope under the continuous pressure sequence of the radial artery, namely the pulse P-S curve, was extracted. We also analyzed the relationship between the individual P-S curve difference and BMI. The results show that the actual human body data collection and analysis results are consistent with the theoretical model established in this paper, which indicates that the model can provide a novel idea for the evaluation of the state of the human body.

Continuous monitoring of pulse waves plays a significant role in reflecting physical conditions and disease diagnosis. However, the current collection equipment cannot simultaneously achieve wearable and continuous monitoring under varying pressure and provide personalized pulse wave monitoring targeted different human bodies. To solve the above problems, this paper proposed a novel wearable and real-time pulse wave monitoring system based on a novel flexible compound sensor. Firstly, a custom-packaged pressure sensor, a signal stabilization structure, and a micro pressurization system make up the flexible compound sensor to complete the stable acquisition of pulse wave signals under continuously varying pressure. Secondly, a real-time algorithm completes the analysis of the trend of the pulse wave peak, which can quickly and accurately locate the best pulse wave for different individuals. Finally, the experimental results show that the wearable system can both realize continuous monitoring and reflecting trend differences and quickly locate the best pulse wave for different individuals with the 95% accuracy. The weight of the whole system is only 52.775 g, the working current is 46 mA, and the power consumption is 160 mW. Its small size and low power consumption meet wearable and portable scenarios, which has significant research value and commercialization prospects.