Wireless, non-invasive, high-resolution thrill sensor for continuous vascular access monitoring of hemodialysis patients

Abstract

Hemodialysis is the primary life-sustaining therapy for patients with end-stage renal disease (ESRD). Diagnosing and monitoring the health condition of vascular access (VA) for hemodialysis involves the use of ultrasonic imaging and angiography. These infrequent tests cannot, however, identify signs of mild stenosis and sudden onset to facilitate early intervention. Hence, approximately 70% of hemodialysis patients experience VA failure each year, rendering dialysis difficult and posing life-threatening risks. Here, we report wireless, non-invasive thrill sensors incorporating high-resolution accelerometers designed to be conformally placed on VA skin for the continuous monitoring of VA thrill signals, which we tested in multi-platform validation including an in vitro benchtop system, canine models and patients (n = 9). Local measurements of VA thrill intensity via continuous wear demonstrated that a decrease in thrill intensity along the entire VA region is indicative of diminished VA function, while an abnormally heightened thrill intensity at a certain location along the VA indicates localized structural lesions. These thrill sensors may aid the early treatment of VA stenosis upon the transition from subjective thrill palpation to quantitative hemodynamic phenotyping.

Introduction

End-stage renal disease (ESRD), the final stage of chronic kidney disease, has an extremely high mortality rate^1,2,3,4, and is expected to become the fifth leading cause of shortened life expectancy as the world ages^5,6. Renal replacement therapies, including hemodialysis, are imperative for sustaining life in patients with ESRD^7,8,9,10. Current trends indicate that the number of patients undergoing hemodialysis worldwide could reach 5.4 million by 2030¹¹, with a projected annual growth rate of 4%¹². Hemodialysis requires the special blood circulation for inflow and outflow, achieved through vascular access (VA) — a surgically established direct connection (circulatory short-circuit) between a selected vein and artery in the patient’s forearm or upper arm^13,14,15,16 (see Fig. 1a). However, maintaining long-term patency and optimal VA function is challenging¹⁷, as repeated punctures during hemodialysis and potential vascular abnormalities from specialized blood flow¹⁸ may lead to intimal hyperplasia¹⁹. Approximately 70% of patients experience VA failure annually²⁰, primarily due to untreated stenosis. This complication can leave patients unable to undergo hemodialysis, posing a life-threatening risk^21,22. Thus, timely VA monitoring is critical²³. Presently, direct assessment of VA blood flow rate using a Doppler ultrasound machine is the most useful indicator of VA stenosis²⁴. Nevertheless, it remains of a clinical nature and possesses limited potential for daily, at-home, rapid and continuous monitoring of VA health conditions²⁵, representing an unmet healthcare need.

Fig. 1: The thrill sensor for continuous monitoring of vascular access (VA) health conditions.

a Schematic illustration of a patient with vascular access (VA) in the forearm, equipped with a wireless thrill sensor. b Schematic of a well-functioning VA and its corresponding high-frequency, broadband thrill acceleration signal. c Schematic of early-stage stenosis in the VA, resulting in reduced-amplitude, more irregular thrill signal. d Schematic of advanced-stage stenosis in the VA, resulting in a markedly diminished thrill signal. e Simultaneous recordings from a canine model with well-functioning VA, VA thrill acceleration (purple) by the thrill sensor, VA thrill sound (blue) by a commercial digital stethoscope, and VA pleth (black) by an electrocardiographic (ECG) monitor. f Continuous monitoring of VA thrill changes during dynamic movement: a patient gradually raises his arm from 0° to 90°. g Comparison of thrill signals before and after percutaneous transluminal angioplasty (PTA) surgery for venous anastomosis stenosis: pre-PTA, only a pulse-like vibration; post-PTA, thrill signal recovery. h Distinct thrill signatures in different stenosis locations: venous anastomosis stenosis, prominent pulse signal; arterial puncture-site stenosis: weakened thrill without pulse.

In long-term medical practice, clinicians have collated a compendium of empirical methods for evaluating VA health conditions. Due to the distinctiveness of the blood flow inside the VA²⁶, high-frequency vibrations are generated on the skin surface in its vicinity, which are distinctly discernible from the pulsation of arterial blood vessels. This phenomenon is referred to as VA thrill²⁷ and exhibits significant variability among different VA health conditions²⁸ (Fig. 1b–d). Palpation and auscultation of the thrill²⁹ are the most common methods of rapid assessment of VA health conditions and are the only methods that patients can self-assess at home, as outlined in the recommendations for routine maintenance of VA³⁰. Nevertheless, this method is highly subjective³¹, and the accuracy of the assessment is contingent on the operator’s experience. The complex characteristics of VA thrill signals that vary with the degree of stenosis pose significant challenges to its quantitative measurement and corresponding diagnostic strategy. Researchers have sought to develop quantitative methods to monitor VA thrill, including the use of photoplethysmography (PPG) sensors and digital stethoscopes. Specifically, following continuous development^32,33,34,35, the PPG sensor has been shown to indirectly assess VA thrill by monitoring changes in arterial pulse along the VA. However, it cannot directly and precisely measure VA thrill and is vulnerable to serious motion artifacts and ambient light interference, resulting in inaccurate assessments. Recent works^36,37,38,39 have employed a commercial digital stethoscope for signal acquisition and investigated the screening method of VA stenosis using deep learning signal analysis methods. Nonetheless, the probe of the commercial equipment is large and difficult to wear, making it unsuitable for continuous monitoring of VA health conditions. There is an urgent need for the solution that can expeditiously, objectively and continuously monitor VA health. While arteriovenous fistulas (AVFs) represent the majority of VA cases⁴⁰, this study focuses on arteriovenous grafts (AVGs) because (1) AVGs exhibit more complex stenosis characteristics than AVFs, occurring at arterial/venous anastomosis, mid-graft, or puncture sites⁴¹; (2) AVGs’ variable depth makes thrill assessment via palpation particularly challenging⁴², creating an unmet need for objective monitoring tools that the KDOQI guidelines (Section 13.2, 13.5, 13.7)³⁰ highlight, which recommends regular physical examination while discouraging routine AVG surveillance by imaging; (3) AVGs become essential and life-sustaining when AVFs fail in long-term hemodialysis patients⁴³; and (4) AVGs’ standardized PTFE material⁴⁴ facilitates sensor development.

Therefore, to bridge this critical clinical gap, we developed a wireless, non-invasive thrill sensor integrating dual MEMS accelerometers and motion-artifact cancellation methods, which is capable of high-resolution spatial-temporal mapping of VA thrill. Through multi-platform validation, encompassing in vitro benchtop systems, canine models, and human studies, we systematically decode thrill signatures as quantitative biomarkers for VA health conditions and establish two diagnostic signatures: (1) thrill intensity exhibits a persistent positive correlation with blood flow rate in well-functioning VAs, while dysfunctional VAs show an over 20% reduction; (2) pathognomonic focal singularities of heightened intensity precisely localize stenosis sites. These enable a dual-parameter diagnostic strategy: global thrill intensity quantifies VA functional decline, and thrill singularity localizes VA structural lesions. The thrill sensor transforms subjective palpation into objective hemodynamic phenotyping, offering the paradigm for continuous, patient-administered VA surveillance.

Results

Thrill sensor for continuous monitoring vascular access health condition

The proposed thrill sensor can be noninvasively attached to the skin surface above VA, facilitating the wireless, rapid, quantitative monitoring of high-frequency and weak thrill signals. This approach enables the real-time assessment of VA health conditions (Supplementary Fig. 1). In comparison to Doppler ultrasound probes or stethoscopes, our sensor is more compact (30 × 15 × 5 mm³, representing less than 5% of the ultrasound probe’s volume) and lighter (~20 g, approximately 4% of the ultrasound probe’s weight), with minimal mass and mechanical loading (Supplementary Figs. 2, 3, and 4). The thrill sensor is able to accurately monitor the VA thrill signal, as illustrated in Fig. 1e (from a canine’s well-functioning VA). This signal is shown to be consistent with the recording and the cardiac cycle synchronously acquired by the electrocardiographic (ECG) monitor and the commercial digital stethoscope (3 M™ Littmann® CORE Stethoscope, Eko). Furthermore, the thrill sensor exhibits rapid and sensitive responses to continuous variations in thrill at the patient’s VA. As the patient continuously raised the VA arm, the sensor was able to detect an increase in the VA thrill signal (Fig. 1f), thereby demonstrating its excellent dynamic performance. Additionally, comparison between two groups of distinct thrill signals revealed (i) the recovery of the thrill signal after percutaneous transluminal angioplasty (PTA) surgery for venous anastomosis stenosis (Fig. 1g and Supplementary Fig. 5, Supplementary Movies 1 and 2); (ii) the emergence of the pulse signal at venous anastomosis stenosis, whereas this phenomenon did not exist in arterial puncture-site stenosis (Fig. 1h and Supplementary Fig. 6); (iii) the change in the thrill signal in a normal VA patient pre/post-hemodialysis (Supplementary Fig. 7); and (iv) the comparison of the thrill measurement results between patient self-attachment and doctor attachment of the thrill sensor (Supplementary Fig. 8 and Supplementary Movie 3). The combination of size, mass and sensitivity achieved with this thrill sensor is unattainable with alternative technologies (Supplementary Fig. 9 and Supplementary Table 1).

The core module of the thrill sensor (see Fig. 2a) consists of a pair of MEMS accelerometers (ADXL355, Analog Devices and LIS2DW12, STMicroelectronics). The former (main, noise density ~25 μg/√Hz) is located in close proximity to the VA surface, facilitating high-precision monitoring of the thrill; the latter (auxiliary, noise density ~85 μg/√Hz) is positioned adjacent to the VA, enabling measurement of limb motion. In addition, powered by a Li-ion battery (UFX351525, 90 mAh, 4.2 V), the thrill sensor’s low power consumption (1600 Hz sampling mode and Bluetooth low energy mode) enables 50 min of continuous monitoring and wireless data transmission (Supplementary Fig. 10, and see “Methods” for detailed information on system design and preparation).

Fig. 2: Details of the thrill sensor design and the fundamental monitoring results.

a Top view of the unencapsulated thrill sensor holds between the thumb and index finger, with a schematic of its electronic system structure. b Carotid pulse signals measured by the thrill sensor in a healthy male subject under different physiological conditions: sitting, speaking, and walking. c Simultaneous recording of vascular access (VA) thrill and motion artifacts in a male patient with a well-functioning VA. The raw signal (pink) contains both the VA thrill and limb motion (blue), and the real thrill signal (purple) is isolated after applying the motion-artifact cancellation method. d Schematic decomposition of the real thrill signal from a well-functioning VA. The signal is composed of two components: the arterial pulse, which contributes to the overall amplitude, and the high-frequency, weak vibration which is responsible for the rapid fluctuations.

Laboratory tests revealed that the thrill sensor is capable of measuring carotid pulse signals in various conditions (sitting, walking), thereby verifying its accuracy in monitoring typical biological vibrations (see Fig. 2b). Further environmental robustness testing under skin temperature variations (36–46 °C) and sweating conditions confirmed stable pulse detection (see Supplementary Fig. 11). VA thrill is considerably more complex than the arterial pulse generated by the heartbeat. This difference arises because blood flow in arteries is typically laminar⁴⁵, whereas blood flow within the VA (formed by the direct connection of a vein and an artery) is inherently turbulent⁴⁶, even when the VA functions well. Figure 2c illustrates the thrill signal of a patient (male, 53 years old) with a well-functioning VA. The thrill sensor recorded both the VA thrill with motion (pink lines, SNR = −26.95 dB) and arm movement (blue lines) simultaneously and then subtracted these to eliminate motion artifacts, obtaining the real VA thrill signal (purple line, SNR = 9.65 dB). According to Supplementary Note 1, a significant 36.6 dB SNR improvement is demonstrated through our dual-accelerometer motion-artifact cancellation method, which establishes a robust foundation for more accurate VA thrill measurement. By extracting a single cardiac cycle of the real VA thrill signal for signal decomposition, it becomes evident that the thrill of a well-functioning VA is composed of two parts: the arterial pulse and the VA high-frequency weak vibration. The arterial pulse contributes to the amplitude of the VA thrill signal, while the VA high-frequency weak vibration, with an amplitude approximately one-third of that of the arterial pulse, contributes to the high-frequency fluctuation (see Fig. 2d). When VA dysfunction occurs, the VA thrill signal becomes more complex and variable because the blood flow becomes more irregular due to abnormal changes in vascular morphology (such as stenosis)⁴⁷. At present, thrill-based diagnosis relies heavily on the medical experience of clinicians. To this end, we employed VA thrill sensors to systematically and quantitatively characterize VA thrill under varying degrees of stenosis using multi-platform experimental approaches comprising in vitro benchtop systems, animal models, and human studies.

In vitro experiments

A cardiovascular benchtop system containing mimic VA was established for in vitro experiments on the thrill sensor (refer to Fig. 3a and Supplementary Fig. 12). A pulsatile water pump was utilized to emulate cardiac function, with a mimic heart rate ranging from 60 to 150 bpm and a flow rate from 0.1 to 1.9 L/min (the blood flow rate in healthy VAs varies from 0.6 to 1.5 L/min^30,48). The mimic tissue with VA inside was made from silicone (Dragon Skin 00-10), where VA (a direct connection between a selected vein and artery) was simulated by a Y-shaped tubing cavity located 3 mm beneath the surface (for more details, refer to “Methods”). VA scenarios revealed that the simulated VA thrill signals on the benchtop system, measured by the thrill sensor, exhibited a high degree of consistency with the real VA thrill signals from patients (Supplementary Fig. 13). This demonstrated that the benchtop system had the capacity to simulate VA thrill and can be utilized for subsequent in vitro experiments on the subject.

Fig. 3: Performance of the thrill sensor on the in vitro benchtop system.

a Schematic of the vascular access (VA) benchtop system, illustrating the principle and components. b Schematic of the measurement point (“*”) and stenosis positions (“=”) on the mimic VA. Stenosis is induced by inserting polytetrafluorethylene (PTFE) cylindrical tubes into the vein (blue) or artery (red) section of the mimic VA to simulate venous and arterial stenosis, respectively. c Time-domain signals of the VA thrill acceleration for a well-functioning VA (0% stenosis, purple) and a dysfunctional VA (60% stenosis, yellow). The amplitude of the thrill signal is significantly reduced in the stenotic condition. d Spectrograms of the VA thrill signal for a well-functioning VA and a dysfunctional VA. The characteristic center frequency shifts from 200 Hz to 40 Hz, indicating a reduction in high-frequency thrill. e Thrill intensity as a function of stenosis degree. The thrill intensity decreases progressively with increasing stenosis, showing a significant reduction at advanced stages (>50%). f Spectrogram and time-domain signal of the VA thrill in severe vein stenosis. The prominent 1–2 Hz spike corresponds to the emergence of an arterial pulse signal. g Spectrogram and time-domain signal of the VA thrill in severe artery stenosis, with no emergence of the pulse.

VA stenosis mainly includes stenosis at VA venous outflow (vein stenosis) and stenosis at VA arterial inflow (artery stenosis), and the majority of VA stenosis is vein stenosis⁴¹. As shown in Fig. 3b, polytetrafluorethylene (PTFE) cylindrical tubes with varying inner and outer diameter ratios were positioned into the mimic VA in the benchtop system to simulate different degrees of vein stenosis (tiny, blue tube in Fig. 3b) and artery stenosis (tiny, red tube in Fig. 3b), separately. With respect to VA scenarios of mimic vein stenosis, Fig. 3c shows the acceleration signals of the simulated VA thrill when the mimic VA was non-stenosis or in 60% stenosis, with a constant simulated heart rate of 60 bpm and a simulated blood flow rate of 600 mL/min. In 60% stenosis, the maximum amplitude of the thrill acceleration time-domain signal was reduced to 1/5 of that when the mimic VA was non-stenosis. The time-domain signal is Fourier transformed to obtain the VA thrill acceleration signal spectrogram (Fig. 3d). The characteristic center frequency of the spectrogram was reduced from approximately 200 Hz at full patency to approximately 40 Hz at 60% stenosis, reflecting the intrinsic frequency of the simulated tissues (Supplementary Fig. 14). This showed that in cases of VA stenosis, the low-frequency vibration of the biological tissue due to cardiac activity gradually replaces the high-frequency thrill and dominates. Figure 3e shows the thrill intensity calculated by the simulated VA thrill (Supplementary Fig. 15) when the mimic VA was non-stenosis or in 30%, 60%, and 80% stenosis. The thrill intensity gradually decreased with increasing stenosis level, and stenosis VA exhibited a significant reduction in thrill intensity compared to non-stenotic VA (n = 4, t = 3.598, df = 3.031, p = 0.018 < 0.05, Cohen’s d = 2.544, one-tailed Welch’s t-test, see “Methods”). Specifically, at early-stage stenosis (<50%), VA thrill intensity decreased insignificantly, making it difficult to diagnose VA stenosis by palpation or auscultation. However, VA thrill intensity declined rapidly at advanced-stage stenosis (>50%), certainly leading to a significant change in palpation, but it was too late to undergo minor medical intervention at that time.

Figure 3f shows the spectrogram and extracted time-domain signal of simulated VA thrill (Supplementary Fig. 16) in severe vein stenosis (>80%). A sharp spike in the spectrogram was observed at a frequency ranging from 1 to 2 Hz, which revealed the emergence of an arterial pulse signal. This rule was consistent with the empirical assessment of clinicians during palpation of the VA; that is, severe stenosis at VA venous outflow results in a palpable water-hammer pulse⁴⁹. In contrast, severe stenosis at VA arterial inflow results in diminished intensity both in VA thrill and pulse⁵⁰, as illustrated in Fig. 3g.

Animal experiments

The animal experiments were performed on 5 canine models to verify the adaptability of the thrill sensor in real scenarios (Figs. 4–6, Supplementary Figs. 17 and 18). Figure 4a illustrates an example of the thrill sensor employed for well-functioning VA (the arteriovenous graft, AVG) thrill monitoring in a labrador. Figure 4b shows the thrill signal on the VA in the left leg and the pulse signal on the superficial femoral artery in the right leg, measured simultaneously. Both the time-domain signal and the time-frequency map can reveal that the VA thrill is the high-frequency vibration of the blood vessel superimposed on the arterial pulse. This is consistent with the signal characteristics of the real patient (refer to Fig. 2c) and can verify the animal model. The spectrogram indicates that the characteristic frequency band of VA thrill is 100–200 Hz. Consequently, the spectral mean value in this band (denoting the energy of the vibration signal, see “Methods”) was selected to characterize the intensity of the thrill.

Fig. 4: Animal application of the thrill sensor and vascular access (VA) stenosis assessment.

a Schematic of the animal model and sensor placement. b Comparison of the vascular access (VA) thrill signal (purple) and the normal artery pulse signal (red) recorded simultaneously from the same canine, showing the VA thrill as a high-frequency vibration superimposed on the arterial pulse. The spectrogram reveals a characteristic frequency band of 100–200 Hz for VA thrill. c Results from Canine 1 with a well-functioning VA (arteriovenous graft, AVG). The thrill sensor (purple) and commercial digital stethoscope (blue) are used to monitor the VA thrill signal at four different stenosis levels (0%, 37%, 68%, 76% flow reduction), with Doppler ultrasound measurements (PS, ED, TAMEAN, VolFlow) for reference. d Results from Canine 2 with a repaired VA (AVG) after percutaneous transluminal angioplasty (PTA) surgery. The thrill sensor (purple) and commercial digital stethoscope (blue) are used to monitor the VA thrill signal at four different stenosis levels (0%, 48%, 66%, 79% flow reduction), with Doppler ultrasound measurements for reference.

Fig. 5: Animal application of the thrill sensor during rapid and continuous vascular access (VA) changes.

Schematic of vein-side stenosis (a) and artery-side stenosis (b) in Canine 1 induced by finger pressure. The thrill sensor (purple) and commercial digital stethoscope (blue) simultaneously monitor the vascular access (VA) thrill signal. The Doppler ultrasound image shows a 56%/57% flow reduction, and the thrill signal decreases continuously, demonstrating the sensor’s rapid response. c Time-domain thrill signals monitored by the thrill sensor (purple) and commercial digital stethoscope (blue), and Doppler ultrasound images during percutaneous transluminal angioplasty (PTA) surgery in Canine 2, including: occlusion (no blood flow detected), half revascularization (402.36 mL/min), and full revascularization (808.07 mL/min). d Histogram of thrill intensity variation calculated from the above cases of Canine 1 and Canine 2. The thrill intensity is significantly lower in the high-risk state than in the healthy state (critical flow rate = 500 mL/min, p = 0.0015, one-tailed Welch’s t-test). Data are presented as mean values ±  SD from n = 9 individual cardiac cycles in each flow rate state. e Time-domain thrill signals and the histogram of thrill intensity during the rapid process of severe stenosis (more than 80% stenosis within 8 s) in Canine 3. The thrill intensity decreases from 0.98 to 0.06 in vein-side stenosis (blue) and from 0.79 to 0.02 in artery-side stenosis (purple). f Comparison of thrill signals during severe stenosis in Canine 3. The pulse signal emerges at severe vein-side stenosis (blue), but not in severe artery-side stenosis (purple).

Fig. 6: Animal application of the thrill sensor for spatial mapping the thrill distribution along the vascular access (VA, arteriovenous graft, AVG).

a Photo and Doppler ultrasound images of Canine 1’s well-functioning vascular access (VA, arteriovenous graft, AVG). The monitoring points (P1–P10) are distributed along the graft from the arterial anastomosis (P1) to the venous anastomosis (P10), with P6 at the U-turn. b Time-domain thrill signals at each monitoring point (P1–P10) along the AVG of Canine 1. c Distribution of thrill intensity, depth, and blood flow rate along the AVG of Canine 1. The thrill intensity decreases from P1 to P6, reaching a minimum, and then increases towards P9. Data are presented as mean values ±  SD from n = 9 individual cardiac cycles in each monitoring point. d Photo and Doppler ultrasound images of Canine 2’s repaired VA (AVG) with residual stenosis at venous anastomosis (P10). e Time-domain thrill signals at each monitoring point (P1–P10) along the AVG of Canine 2. f Distribution of thrill intensity, depth, and blood flow rate along the AVG of Canine 2. The thrill intensity shows an overall reduction but a singularity increase at P10. Data are presented as mean values ±  SD from n = 9 individual cardiac cycles in each monitoring point. g Colormap of thrill intensity along the newly implanted VA (AVG, Canine 5) on Day 1 and Day 7. The thrill intensity increases significantly from Day 1 to Day 7. h Colormap of thrill intensity along the well-functioning and 6-month mature VA (AVG, Canine 1, 3, and 4). The thrill intensity is highest at the arterial anastomosis, decreases towards the U-turn, and then increases towards the venous anastomosis. i Colormap of thrill intensity along the abnormal VA (AVG, Canine 2) with residual stenosis. The thrill intensity is significantly reduced overall, but there is a notable increase at the venous anastomosis.

Two canines, one designated Canine 1, which had a well-functioning AVG surgically created at least 6 months previously; and another designated Canine 2, which had a repaired AVG after percutaneous transluminal angioplasty (PTA) surgery yesterday, were measured using both the thrill sensor and commercial digital stethoscope for comparison (more experimental details refer to Supplementary Fig. 19). A hemostatic ribbon was employed to induce stenosis in the femoral artery upstream of the AVG (Supplementary Fig. 20). The AVG blood flow rate, as measured by Doppler ultrasound, revealed four distinct stenosis levels (Fig. 4c for Canine 1 and Fig. 4d for Canine 2). Whether Canine 1 had a high initial blood flow rate (984.93 mL/min) or Canine 2 had a low initial blood flow rate (737.41 mL/min), according to the time-domain signals from the thrill sensor and the commercial digital stethoscope, as the flow drop was less than 40%, the measurement results were highly consistent. It also revealed that the thrill attenuation was not significant in mild stenosis, which is consistent with the result of the benchtop system (refer to Fig. 3e). As the flow dropped between 40% and 80%, there was a significant decline in the thrill signal, but the signal measured by the thrill sensor was much stronger than that of the commercial digital stethoscope. Particularly, at 76%, the thrill signal measured by the stethoscope became virtually indistinguishable. When the flow drop was more than 80%, the thrill signal was rendered extremely diminished, thereby limiting both devices to effectively measure such thrill signals. The primary benefit of the thrill sensor stems from the high-resolution of the embedded accelerometer, its compact size and small pick-up area, and its compatibility with interference from the surrounding environment.

By applying continuous finger pressure (Supplementary Fig. 21) to the venous side (Fig. 5a) and arterial side (Fig. 5b) of the graft (Canine 1’s VA), the result was an approximately 60% decrease in the blood flow rate within 30 s, as measured by Doppler ultrasound, reaching moderate stenosis. The change in the thrill signal measured by the thrill sensor and the commercial digital stethoscope exhibited a high degree of consistency, following successive stenosis and decreasing continuously. In particular, when the stenosis reached over 50%, the thrill signal measured by the thrill sensor was significantly stronger than that of the stethoscope. This verified the rapid response and continuous measurement capability of the thrill sensor. Furthermore, we performed thrill monitoring and analysis for the graft (Canine 2’s VA) before and after PTA surgery (Fig. 5c and Supplementary Fig. 22). The VA underwent three phases of occlusion (no blood flow detected), half revascularization (blood flow of 402.36 mL/min), and full revascularization (blood flow of 808.07 mL/min), as measured by Doppler ultrasound. The thrill sensor was able to accurately reflect the changing process of the thrill signal. The process began in the occlusion phase (similar to the arterial pulse signal), gradually superimposed the high-frequency vibration with the increase in blood flow rate during thrombolysis and ultimately returned to the normal level. In summary, as shown in Fig. 5d, the histogram of thrill intensity variation calculated from the above cases (stenosis or revascularization process of Canine 1 and Canine 2) also demonstrates that, for the same individual, the quantitative thrill intensity measured by the thrill sensor is correlated with the VA blood flow rate, which can reflect the current VA health condition to a certain extent. The abnormal VA exhibited a significant reduction in thrill intensity compared to well-functioning VA (n = 4, t = 7.873, df = 3.258, p = 0.0015 < 0.05, Cohen’s d = 5.567, one-tailed Welch’s t-test, see “Methods”).

Moreover, based on the thrill sensor, the change of VA thrill in the rapid process of severe stenosis (more than 80% stenosis within 8 s on Canine 3’s well-functioning VA) was analyzed. The thrill intensity decreased from 0.98 to 0.06 during vein-side stenosis and from 0.79 to 0.02 during artery-side stenosis (Fig. 5e). A comparison between the two thrill signals in 7 ~ 8 s revealed the emergence of pulse signals at severe vein-side stenosis (blue curve in Fig. 5f), whereas this phenomenon did not exist in severe artery-side stenosis (purple curve in Fig. 5f). These results were consistent with the findings obtained from in vitro experiments via the benchtop system.

Having established the efficacy of the thrill sensor for localized, single-point assessment of VA thrill characteristics under varying hemodynamic conditions, we extended our investigation to evaluate the sensor’s capability for spatial mapping, that is, the quantitative profiling of thrill distribution along the entire length of the VA arteriovenous graft (AVG). An e-PTFE artificial blood vessel (approximately 20-30 cm in length) is employed in the AVG; in patients undergoing maintenance hemodialysis, stenosis inside the graft is also a prevalent type⁴¹. Therefore, quantitative measurement and continuous monitoring of thrill distribution along the AVG are critically important for early stenosis detection. However, current clinical tools face inherent limitations when measuring thrill distribution: ultrasound probes, despite their high-resolution 3D imaging capability, have a large detection footprint (typically ~40 mm in length), which restricts portability and continuous monitoring, while commercial stethoscopes also have bulky auscultation heads (6–8 times the graft diameter, Supplementary Fig. 23). Furthermore, certain existing wearable solutions for VA monitoring may suffer from large-area and long-time attachment, potentially interfering with hemodialysis cannulation and leading to reduced acceptance by patients and nursing staff. In contrast, the design of the thrill sensor incorporates a compact detection area (10 × 10 mm²) which is capable of confining measurements to a small region. This miniaturized footprint enables flexible placement away from critical cannulation sites, minimizing interference and improving clinical utility and user acceptance with temporary monitoring sessions (e.g., pre-dialysis or interdialytic assessment). Although this resolution cannot pinpoint and image sub-5 mm lesions with ultrasound-level precision, it can detect localized abnormal vibration intensity averaged over the 10 × 10 mm² area. Such distributed monitoring offers potential for pinpointing the approximate location of stenosis, assessing overall VA health conditions and complementing existing modalities.

The time-domain signals of the thrill via the thrill sensor were obtained from a total of 10 monitoring points on Canine 1’s well-functioning AVG (Fig. 6a, b and Supplementary Fig. 24). Ten points were distributed uniformly along the VA, from the arterial anastomosis (P1) through the U-turn (P6) to the venous anastomosis (P10), as shown in Fig. 6a. Doppler ultrasound measurements (Supplementary Figs. 25 and 26) demonstrated that the AVG depth fluctuated within acceptable limits, except for the anomalous depth of monitoring point P10, and the flow rate was consistently within the normal range. Figure 6c shows the distribution of the thrill intensity, depth and blood flow rate along the AVG of Canine 1. The thrill intensity decreased from 7.87 at the arterial anastomosis (P1), reached a minimum value of 0.42 at the U-turn (P6), and then increased from point to point, reaching 1.13 near the venous anastomosis (P9). Supplementary Figs. 27 and 28 show the thrill distribution along the AVG in another two experimental dogs, Canine 3 and Canine 4. In summary, in cases of well-functioning AVGs, the thrill intensity exhibited continuous variation. It underwent a gradual weakening from the arterial anastomosis to the U-turn, followed by a subsequent gradual strengthening from the U-turn to the venous anastomosis, and the greatest intensity was observed at the arterial anastomosis.

In the case of abnormal AVG, as evidenced in Canine 2 (Fig. 6d and Supplementary Figs. 29 and 30), which exhibited 26% residual stenosis at the venous anastomosis (P10) after PTA surgery, the thrill intensity distribution along the AVG was markedly reduced overall. However, the thrill intensity at the venous anastomosis (P10) was abnormally enhanced, which can be considered a singularity on the intensity distribution, and was significantly stronger than the other monitoring points on the entire AVG (Fig. 6e, f). Supplementary Fig. 31 presents the results of thrill monitoring in a recently established AVG (Canine 5). As the incision healed, there was an increase in overall blood flow rate. On the 7^th day after implantation, the thrill intensity at the same monitoring point on the AVG was up to twice as high as that on the day after implantation. Figure 6g-i summarizes the AVG thrill intensity colormap of 5 experimental dogs. The dogs were divided into three distinct groups: the newly implanted AVG (Canine 5), well-functioning and 6-month mature AVGs (Canine 1, 3, and 4) and the abnormal AVG (Canine 2). The findings reveal that the well-functioning group exhibits a significantly stronger thrill intensity in comparison to the newly implanted and abnormal groups (Supplementary Fig. 32), while the abnormal group shows a unique increase in thrill intensity across the arterial to venous anastomosis.

Human studies

The human study explores the applicability of the thrill sensor in a clinical setting. A total of 9 hemodialysis patients were included in this study. The average age of the patients is 57.5 ±  16.5 years, and 4 (44.4%) are male. The basic patient information and clinical characteristics are summarized in Supplementary Table 2.

First, as shown in Fig. 7a, 3 patients with well-functioning VA (arteriovenous graft, AVG, e.g., Patient A: Supplementary Figs. 33, 34, 35) were assessed by the thrill sensor and digital subtraction angiography (DSA) (Patient A and Patient B, whose AVGs returned to normal after repair; Patient C, whose AVG remained healthy for over 6 months). Figure 7b shows the thrill intensity histogram of 10 uniformly distributed monitoring points and thrill signals at key points (P1 and P10: the anastomosis; P6: the U-turn) for each patient. It is evident that the thrill intensity at the arterial anastomosis (P1) is significantly greater than that at the other monitoring points. The thrill intensity exhibits continuous change along the AVG, transitioning from strong (the arterial anastomosis) to weak (the U-turn) and subsequently from weak (the U-turn) to strong (the venous anastomosis). This finding aligns with the typical characteristics of thrill in well-functioning AVGs, as previously found in animal experiments.

Fig. 7: Clinical validation of the thrill sensor in hemodialysis patients.

a Photo and digital subtraction angiography (DSA) or Doppler ultrasound images of well-functioning arteriovenous grafts (AVGs) in three patients (Patient A, B, and C). The monitoring points (P1–P10) are distributed along the graft from the arterial anastomosis (P1) to the venous anastomosis (P10), with P6 at the U-turn. b Thrill intensity histogram and time-domain thrill signals at key monitoring points (P1, P6, P10) for three patients of well-functioning AVGs. The thrill intensity is highest at P1 and decreases towards P6, reaching a minimum, and then increases towards P10. Data are presented as mean values ±  SD from n = 9 individual cardiac cycles in each monitoring point. c Photo and DSA images of abnormal AVGs with residual stenosis at venous anastomosis in three patients (Patient D, E, and F). d Thrill intensity histogram and time-domain thrill signals at key monitoring points for three patients of abnormal AVGs with residual stenosis. The thrill intensity is reduced overall, with a notable increase singularity at P10. Data are presented as mean values ±  SD from n = 9 individual cardiac cycles in each monitoring point. e Photo and DSA images of abnormal AVGs with irregular lesions in three patients (Patient G, H, and I). Patient G’s AVG exhibits a morphological 90° turn, Patient H’s AVG has graft ectasia, and Patient I’s AVG has a non-drug-eluting stent implanted near the venous anastomosis. f Thrill intensity histogram and time-domain thrill signals at key monitoring points for three patients of abnormal AVGs with irregular lesions. The thrill intensity is reduced overall, with a notable increase singularity at the lesion point. Data are presented as mean values ±  SD from n = 9 individual cardiac cycles in each monitoring point. g Correspondence between the thrill intensity and blood flow rate in the AVGs of 9 patients. The thrill intensity exhibits a positive correlation with the blood flow rate in well-functioning AVGs (Patient A–C) and a significant thrill intensity reduction in abnormal AVGs (Patient D–I). h Diagram of the thrill intensity distribution along the AVGs of 9 patients.

Furthermore, 6 patients with abnormal AVG were assessed. Figure 7c shows 3 patients (Patient D, Patient E and Patient F) whose AVG had residual stenosis at the venous anastomosis (P10) after repair (e.g., Patient D: Supplementary Figs. 36, 37, 38). According to Fig. 7d, their thill intensity exhibits a reduction of more than 20% in comparison with the well-functioning AVGs (n = 6, t = 2.758, df = 4.222, p = 0.024 < 0.05, Cohen’s d = 1.927, one-tailed Welch’s t-test, see “Methods”). However, the thrill is abnormally enhanced at the residual stenosis position (P10), which becomes a singularity in the thrill intensity distribution. Figure 7e shows 3 patients (Patient G: Supplementary Figs. 39, 40, 41, Patient H: Supplementary Figs. 42, 43, 44 and Patient I: Supplementary Figs. 45, 46, 47), whose AVG has an irregular lesion inside (Patient G’s AVG exhibits a morphological 90° turn, Patient H’s AVG has graft ectasia in the vicinity of the arterial puncture point, and Patient I’s AVG undergoes a non-drug-eluting stent implanting near the venous anastomosis). According to Fig. 7f, the presence of an irregular lesion in the AVG not only results in an overall weaker thrill intensity compared to the well-functioning AVG but also leads to a singularity of enhanced thrill at the lesion point. Specifically, in Patient G’s AVG, the inside morphological 90° turn is located between the monitoring points P1 and P2, which triggers the Bernoulli effect, resulting in the abnormally enhanced thrill at the downstream position (P2). Additionally, in Patient H’s AVG, the graft ectasia is located between monitoring points P5 and P7, which intensifies turbulent flow, resulting in the abnormally enhanced thrill at the center of the ectasia (P6). In Patient I’s AVG, the stent suppresses the majority of the total energy, but its thrill intensity still experiences tiny enhancement near the stent position (P10). Consequently, in instances where the AVG is abnormal, the predominant feature of thrill is characterized by a diminished overall intensity, with a distinct singularity of thrill intensity observed at the lesion point. This finding also aligns with the characteristics of thrill along the abnormal AVG in animal experiments.

According to Supplementary Fig. 48, the Bland‒Altman plots demonstrate the comparability between normalized thrill intensity and dimensionless blood flow rate. Based on clinical assessments using the thrill sensor, the human studies revealed two defining characteristics of the thrill related to vascular access health conditions: (1) thrill intensity exhibits a persistent positive correlation with blood flow rate in well-functioning AVGs (Patient A-C), and abnormal AVGs (Patient D-I) face over 20% reduction in global thrill intensity compared to well-functioning AVGs (see Fig. 7g); (2) focal singularities of heightened thrill intensity precisely localized at the lesion point in abnormal AVGs (see Fig. 7h).

Discussion

This paper introduces a wireless, non-invasive thrill sensor that enables high-resolution spatial-temporal mapping of vascular access (VA) thrill for continuous hemodialysis access monitoring. The device integrates dual MEMS accelerometers and motion-artifact cancellation, achieving the sensitivity (25 μg/√Hz noise floor) within a miniaturized form factor (30 × 15 × 5 mm³). Its capacity for real-time, quantitative thrill assessment coupled with Bluetooth low-energy transmission addresses the critical unmet need for objective, patient-administered VA surveillance beyond intermittent clinical ultrasound.

Our multi-platform validation reveals fundamental hemodynamic signatures of VA health conditions. In vitro experiments established that the thrill intensity decays nonlinearly with stenosis severity. Animal experiments and human studies further demonstrated that well-functioning VA exhibits a persistent positive correlation between thrill intensity and blood flow rate, and continuous spatial thrill variation with peaking at arterial anastomosis. Conversely, abnormal VA consistently presents over 20% global thrill intensity reduction accompanied by focal singularities of heightened intensity precisely localized with lesions. These singularities provide a spatial marker for pinpointing stenosis location or type, enabling a dual-parameter diagnostic strategy: global thrill intensity quantifies functional decline, while thrill singularity localizes structural lesions.

Clinically, this approach converts subjective palpation into objective hemodynamic phenotyping, enabling continuous surveillance of thrill-flow coupling and earlier detection of functional decline or focal stenosis. The technology aligns with the KDOQI guidelines (Section 13.2, 13.5, 13.7)³⁰ by bridging the gap between manual examination and instrumental surveillance—offering quantitative results while avoiding unnecessary routine imaging. Its integration into clinical workflows is streamlined through ECG-synchronized thrill signal analysis for temporal correlation, multi-point monitoring for stenosis localization, and dynamic trend tracking with VA health conditions. The sensor’s potential extends to (i) alerting patients and clinicians to impending dysfunction before critical flow loss, (ii) guiding targeted imaging or intervention to the lesion site, and (iii) reducing unnecessary whole graft revisions and associated morbidity. Further clinical validation and testing may lead to broad adoption in expanded patient cohorts with diverse VA types and situations.

Methods

Fabrication of the thrill sensor

A laminate composite film of copper/polyimide/copper (18 μm/12.5 μm/18 μm, Shenzhen JLC Technology Group Co., Ltd.) served as the substrate for the thrill sensor. The copper layer is processed through the wet etching process to pattern the traces and bond pads. The substrate is drilled by micro twist drill bits and electroplated by graphite conductive adhesive (Ausbond A528). Successive cleaning with stainless-steel liquid flux (Worthington Inc.) and deionized water prepared the resulting flexible printed circuit board (f-PCB) for electronic components by removing surface oxides. The surface mount components on the f-PCB are soldered with low-temperature lead-free solder paste (LF999, KELLYSHUN) by a reflow oven (T937, Tai’an Puhui Electric Technology Co., Ltd.) to establish electrical contact between the components and the copper pads.

Design of the wireless electronic system

A pair of MEMS accelerometers (ADXL355, Analog Devices and LIS2DW12, STMicroelectronics) collect the thrill and motion artifact signals, and transmit the data to the BLE/Wi-Fi dual-mode SoC (ESP32C3-FH4, Espressif Systems) through the SPI communication protocol (written by Arduino IDE 1.8.19, Arduino). The signal is transmitted to the user’s mobile device (PyCharm on the laptop computer and BLE Assist on the smartphone) or the MQTT server (Local server) through a 2.4 GHz ceramic antenna (2450AT18A100E, JOHANSON). The system is powered by a Li-ion battery (UFX351525, 90 mAh, 4.2 V, Guangdong Ufine New Energy Co., Ltd.), and a linear voltage regulator LP5912-3.3DRVR (Texas Instruments) steadily reduces the voltage to 3.3 V for use by the electronic system.

Data acquisition and instrumentation

Simultaneous data acquisition for several thrill sensors occurs through Bluetooth low energy (BLE) with a host laptop (Legion Y9000X, Lenovo). The data are recorded via a custom interface (PyCharm 2023.1.2, JetBrains) and processed with custom algorithms (MATLAB R2022b, MathWorks). Vibration samples are acquired at a sampling frequency of 1600 Hz.

Signal filtering and data analysis

A second order Butterworth bandpass filter with a lower cut-off frequency of 100 Hz and a higher cut-off frequency of 200 Hz was applied to the raw thrill signals to extract the characteristic signal of VA thrill, which was quite distinct from the normal physical vibration including arterial pulse and limb motion. The filtered signals were then processed to quantify the thrill intensity through spectral energy analysis. They were segmented into 2-second epochs (50% overlap), multiplied by a Hamming window, and transformed to the frequency domain via fast Fourier transform (FFT). Power spectral density (PSD) was computed, and the thrill intensity (I_thrill) was quantified as the mean PSD (the spectral mean value) over the 100–200 Hz band,

$${I}_{{thrill}}={\int _{{f}_{1}}^{{f}_{2}}}{PSD}\left(f\right){df}/\left({f}_{1}-{f}_{2}\right),$$

where f₁ = 200 Hz, f₂ = 100 Hz.

Benchtop vascular access model

The model is a simulated vascular access system with a modulus of elasticity close to human skin. First, a mold with a cuboid cavity is designed for a specific vascular access depth and diameter. Then, a 3D printed polyvinyl alcohol (PVA, water-soluble) tube, shaped like the vascular access, is placed into the mold. Next, a uniform mixture of the two components of Dragon Skin 00-10 (Smooth-On, Inc., mass of A: mass of B = 1:1) was poured, and the model was left at room temperature (25 °C) for 5 h. Finally, the cured cuboid model was removed from the mold and immersed in hot water (90 °C) for 24 h to dissolve the PVA tube, and the benchtop vascular access model was obtained.

The flow through vascular access model is controlled with a high-precision pulsatile pump (PLPTXG003, Shanghai Preclinic Medtech Co., Ltd.) at a representative blood flow rate (Q = 0.1 to 1.9 L/min) and simulated heart rate (60 to 150 bpm). A recirculating water bath (3 L) maintained at 36 °C simulated the artificial blood temperature. The thrill sensor is mounted on the benchtop vascular access model surface along the path of flow for vibration sensing.

Benchtop model stenosis

Different degrees of stenosis are simulated through PTFE cylindrical tubes (2 mm length) with different wall thicknesses (inner/outer diameter ratio = 3/16, 1/2, 13/16) placed into the benchtop vascular access model. The position of the PTFE cylindrical tubes was adjusted to create different types of stenosis.

Animal experiment protocol

Animal experiments were approved by the Laboratory Animal Welfare and Ethics Committee of Pudong Hospital, Fudan University (No. 20240329-002). Three 4-year-old male Labrador dogs (Canine 1:4 years/28 kg, Canine 3:4 years/29 kg, Canine 4:4 years/27 kg) from Sunnatech Shanghai Co., Ltd. were used for experiments of well-functioning VA. A 4-year-old male Labrador dog (Canine 2:4 years/27 kg) from Sunnatech Shanghai Co., Ltd. was used for experiments of abnormal VA (with PTA surgery). A 4-year-old male Labrador dog (Canine 5:4 years/26 kg) from Weifang Shengnuo Experimental Animal Breeding Co., Ltd. was used for newly implanted VA experiments. This breed of dog is usually used in preclinical large animal studies of vascular access. They were premedicated using intramuscular injection of 2 mg/kg tiletamine-zolazepam (Zoletil 50) followed by endotracheal intubation. Animals were placed in a supine position, and anesthesia was maintained using inhalation of 1–3% isoflurane (RWD R510) at a 1 L/min flow rate of 100% O₂. Throughout the procedure, electrocardiogram, transcutaneous oxyhemoglobin saturation (SpO₂), and inspired oxygen fraction were continuously monitored. After the experiment, inject butorphanol 0.05 mg/kg subcutaneously and use a thermal blanket to keep the dogs warm. Clean stainless-steel cages were used for breeding. After all appropriate measurements were obtained, surgically remove the implanted VA vessel, continue to raise the dog and seek adoption.

When measuring VA thrill, the thrill sensor was mounted on the surface of the canine’s body, directly above the vascular access, using ultrathin medical adhesive tape (3 M™ 1524). A commercial digital stethoscope (3 M™ Littmann® CORE Stethoscope, Eko) was placed near the thrill sensor to simultaneously measure VA thrill sounds. Standard Doppler ultrasound imaging of the VA was performed with a 4D cardiovascular ultrasound system (Vivid E95, GE HealthCare) by applying the probe directly onto the skin over the VA region to measure the blood flow rate and blood vessel depth. The standard device operation protocol consisted of the synchronized thrill sensing and stethoscope auscultating step (10 - 30 s), and ultrasound step (300 s). The measurement of a single monitoring point was repeated 3 times.

Thrill assessment in changing situations of canine’s VA

In the first situation, a hemostatic ribbon was employed to induce stenosis in the femoral artery upstream of the VA. The femoral artery and vein bundles were identified and traced on Doppler ultrasound. A 5 cm longitudinal superficial skin incision in the groin cranial to this point was made using electrocautery, and the subcutaneous tissues were dissected down using blunt-tip scissors. The femoral artery was exposed using blunt dissection, and a hemostatic ribbon was wrapped around the artery so the blood vessel was tightened to a certain ratio as needed to control the blood flow rate. The thrill sensor and the digital stethoscope were placed 1 cm proximal to the arterial anastomosis. The device operation protocol consisted of the synchronized thrill sensing and stethoscope auscultating step (10 s), and ultrasound step (300 s).

In the second situation, continuous stenosis was introduced by applying continuous finger pressure to the venous side and arterial side of the VA. The point of compression was 1 cm proximal to the venous or arterial anastomosis. The thrill sensor and the digital stethoscope were placed 3 cm proximal to the arterial anastomosis. The device operation protocol consisted of the synchronized thrill sensing and stethoscope auscultating step (30 s), and ultrasound step (at the beginning and end of the continuous compression process).

In the third situation, PTA surgery was performed to repair a blocked VA. Over a 0.035-inch guidewire, a 6-8 millimeter mustang balloon dilatation catheter (Boston Scientific) was coaxially delivered to the VA vessel under fluoroscopic guidance. The thrill sensor was placed 3 cm proximal to the arterial anastomosis. The device operation protocol consisted of the thrill sensing step (10 s) and ultrasound step (300 s) at occlusion, half revascularization and full revascularization.

Thrill distribution mapping of canine’s VA

Each experiment to map the thrill distribution of canine’s VA was carried out at 10 monitoring points. Ten points were distributed uniformly along the VA, from the arterial anastomosis to the venous anastomosis. Each monitoring point is repeated 3 times to form a single measurement. At each monitoring point, the device operation protocol consisted of the thrill sensing step (10 s) and ultrasound step (300 s).

Inclusion and exclusion criteria in the human studies

The human studies were approved by the Institutional Review Board of Huashan Hospital (HIRB), Fudan University (No. 2025-1099). We performed the human study at Huashan Hospital, Fudan University between December 2024 and June 2025. All patients undergoing hemodialysis with arteriovenous graft (AVG) were eligible for the study. The only exclusion criterion was lack of informed consent. The following information was collected from 9 enrolled participants: age, self-reported gender, chief complaint and past history. Written informed consent was obtained from all participants, which included authorization for the publication of identifiable personal information. All participants received extra physical examination for free.

Thrill distribution mapping of the patient’s AVG

The thrill sensor and the digital stethoscope were epidermally adhered at 10 selected monitoring points over the patient’s AVG, and 10 points were distributed uniformly along the looped graft, from the arterial anastomosis to the venous anastomosis. Each monitoring point is repeated 3 times to form a single measurement. At each monitoring point, the device operation protocol consisted of the synchronized thrill sensing and stethoscope auscultating step (10 s), and ultrasound step (300 s).

Statistics and reproducibility

No statistical method was used to predetermine the sample size. No data were excluded from the analysis. There was no blinding in this study. This study employed three calibrated devices used interchangeably across each experiment without bias. Animal studies involved five canines, and human studies included nine patients.

Statistical validation

In vitro experiments, according to Fig. 3e, our claim is that abnormal VAs exhibit a significant reduction in thrill intensity compared to well-functioning VAs. The Welch’s t-test comparing the two groups’ thrill intensity (well-functioning VA: n = 4, μ = 2.365, σ = 0.035; abnormal VA: n = 4, μ = 1.490, σ = 0.485) was performed by SPSS software, with the null hypothesis (H₀): the abnormal VA’s mean had no significant reduction of the well-functioning VA. The result yields a significant difference (p = 0.018 < 0.05), thus rejecting H₀ and supporting our claim. In animal experiments, according to Fig. 5d, our claim is that abnormal VA (Canine 2) exhibits a significant reduction in thrill intensity compared to well-functioning VA (Canine 1). The Welch’s t-test comparing the two groups’ thrill intensity (well-functioning VA: n = 4, μ = 4.5958, σ = 1.070; abnormal VA: n = 4, μ = 0.2946, σ = 0.222) was performed, with the null hypothesis (H₀): the abnormal VA’s mean had no significant reduction of the well-functioning VA. The result yields a significant difference (p = 0.0015 < 0.05), thus rejecting H₀ and supporting our claim. In human studies, according to Fig. 7d, our claim is that dysfunctional VAs exhibit a > 20% reduction in thrill intensity compared to well-functioning VAs. The Welch’s t-test comparing the two groups’ thrill intensity (well-functioning VA: n = 3, μ = 5.022, σ = 1.560; abnormal VA: n = 6, μ = 1.556, σ = 1.289) is performed, with the null hypothesis (H₀): the abnormal VA’s mean was ≥80% of the well-functioning group. The result yields a significant difference (p = 0.024 < 0.05), thus rejecting H₀ and supporting our claim.

Ethics

Every experiment involving animals, human participants, or clinical samples have been carried out following a protocol approved by an ethical commission. Each participant gave informed written consent. The first author, Y.W., consent to the publication of his photo (Fig. 1a) as an illustrative schematic diagram of the device placed on the arm.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.