Cardiovascular disease is the world's leading cause of death¹. According to the World Health Organization, 17.9 million people die every year due to cardiovascular diseases such as coronary heart disease, cerebrovascular disease, peripheral arterial disease, thromboembolic disease, etc. For the prewarning and accurate treatment of cardiovascular diseases, it is important to monitor hemodynamic parameters continuously, including blood pressure (BP), heart rate (HR), peripheral resistance (PR), and vascular elasticity. Several available physiological signals, such as electrocardiogram (ECG) signals², phonocardiogram (PCG) signals³, and pulse waves^{4, 5} can be used to calculate specific hemodynamic parameters and diagnose diseases such as arrhythmia and myocardial infarction. Pulse transit time (PTT), the time it takes for an arterial pulse wave to reach the periphery, can continuously and noninvasively evaluate BP without a cuff^6-11. Soft wearable devices are well suited to monitoring physiological signals with the advantages of real-time operation capability, skin-like mechanical properties, and high-SNR sensing capability^12-22. However, the human cardiovascular system is complicated and distributed. Monolithic hemodynamic parameters achieved by current wearable devices cannot adequately and precisely reflect the health status of regional vasculature. A spatiotemporal hemodynamic monitoring technique is urgently needed to satisfy the ever-growing demand for clinical treatment and daily health management of the cardiovascular system.

The distributed optical fiber (DOF) sensing technique represented by the fiber Bragg grating (FBG) is ideally suited for spatiotemporal hemodynamic monitoring. Its spatially distributed multichannel sensing capability and lack of electromagnetic interference lay a foundation for multiple high-SNR physiological signal monitoring^23-25. The excellent time synchronization ensures PTT detection accuracy considering the fast pulse wave velocity (PWV). In addition, the flexible networking of FBGs can customize a sensing net on the body according to hemodynamic monitoring requirements. However, the traditional optical fiber has a large-distinct mechanical property with the skin considering its rigid and brittle silica material and thick diameter of 125 μm, making it difficult to be worn on the body stably and comfortably. Flexible packaging technology has been used to address the mechanical mismatch^26-28. Nevertheless, excessive thick encapsulation and the low sensitivity of commercial FBG devices pose obstacles in detecting subtle physiological signals, thereby limiting their potential applications in wearable devices. Optical microfibers have been proven to have excellent flexibility, configurability, and large evanescent fields for high sensitivity sensing. In recent years, flexible devices based on microfibers have realized physiological signal monitoring^{19, 29}, voice detection²¹, and human-computer interaction³⁰. However, these devices are difficult to achieve spatially distributed, time-synchronized, and multi-parameter sensing capability without wavelength encoding strategy.

Here, we report a configurable soft microfiber Bragg grating (μFBG) group to monitor spatiotemporal hemodynamics. The precise tapering technology, flexible packaging technology, and advanced femtosecond laser direct-writing technique were employed to prepare the skin-like μFBG patch. Due to the ultrafine diameter and ultrathin thickness, the patch has excellent flexibility and enhanced sensitivity by two orders of magnitude. A ballistocardiogram (BCG) signal is successfully detected by a μFBG patch attached to the chest skin at the tricuspid site. Pulse waves at different superficial artery sites are detected with a high signal-to-noise ratio. The soft μFBG group can be constructed by connecting μFBG patches in series. Benefiting from the flexible layout of the soft μFBG group on the body, systemic PTTs are calculated, revealing the health status of regional arteries. HR and BP, two important hemodynamic parameters, are monitored when the user exercises. Moreover, the variation in regional PR can be detected. Relying on the soft μFBG group, the spatiotemporal hemodynamic monitoring technique opens up new possibilities in daily health management, early screening of disease, and precise treatment in the clinic. Table S1 summarizes the differences of this work from previous studies based on the microfiber sensors.

The optical microfiber contains a femtosecond laser-inscribed FBG embedded in a skin-like thin polydimethylsiloxane (PDMS) patch and is connected end-to-end with two commercial single-mode optical fibers (SMFs) for light propagation, as shown in Fig. 1(a). The optical micrograph in Fig. 1(b) and scanning electron micrograph (SEM) in Fig. 1(c) show that the microfiber has a controllable and uniform diameter of 12 μm using the flame brushing method³¹ (for the detailed fabrication process, see Supplementary information Fig. S1). The smooth surface of the microfiber ensures the low loss of the μFBG patch. The optical microfiber has great flexibility and configurability due to its ultrafine diameter^{29, 32}. The ultrathin PDMS patch prepared by the scraper-coating flexible packaging technique provides a similar elastic modulus to human skin and a large lamination area, ensuring strong adhesion. Additionally, PDMS has a lower refractive index than the microfiber and serves as the cladding layer of the optical waveguide. This restricts the evanescent field of the microfiber and ensures low loss of the μFBG patch. As shown in Fig. 1(d), the skin-like μFBG patch is capable of attaching to a model hand conformally. Using the femtosecond laser direct-writing technique^{33, 34}, the FBG is inscribed at the axis of the microfiber point by point, avoiding the damage of the peripheral PDMS patch. The working wavelength of FBG can be controlled by the grating period when the microfiber’s diameter is fixed. As shown in Fig. 1(e), three soft μFBG patches with different working wavelengths are connected in series by a commercial single-mode fiber, making multichannel operation over the whole body possible. A red light propagates along the optical fiber and scatters out at the three FBG positions. The reflective optical spectra of the three soft μFBG patches can be acquired with a single measurement using the optical signal acquisition such as the optical spectrum analyzer (OSA) and FBG interrogator, ensuring excellent temporal synchronization. When bending one of the soft μFBG patches, its grating period changes, resulting in the wavelength shift of the corresponding reflective peak in the optical spectrum without interference with the other reflective peaks (Fig. 1(f)). The total transmission loss of the three μFBG patches is 1.95 dB (0.65 dB each on average), indicating the low loss of the skin-like μFBG patch. As shown in Fig. 1(g−i), regardless of bending, twisting, or waving the soft μFBG patch, it can withstand deformation and guide light, showing its high potential for skin integration applications.

The skin-like μFBG patch converts strain to a shift in the working wavelength, realizing force/vibration sensing. Its response to the stress and vibration was characterized by the stress test equipment (Supplementary information Fig. S2). The soft μFBG patch was attached to a thick Ecoflex (smooth-on 00-30) substrate, which had a low elastic modulus (69 kPa under 100% strain) and was used to simulate human skin. An Ecoflex contactor was employed to compress the soft μFBG patch, resulting in different degrees of deformation. Figure 2(a) shows the reflective optical spectrum under different stresses. The raw curves (light color) were fitted with the Lorentz function to extract the peak wavelengths. The reflective peak shifted right significantly when the stress increased. According to Fig. 2(b) and Fig. 2(c), the soft μFBG patch had a high and linear sensitivity of 5.26 nm/N under a small stress range (within 50 mN) and a high sensitivity of 2.86 nm/N under a large stress range (0.05–0.45 N). The sensitivity guaranteed the signal quality for human physiological signals such as a pulse wave (10–50 mN). The stress response curves when stress increased and decreased are identical in Fig. 2(d), indicating the low hysteresis property. Figure 2(e) demonstrates the dynamic response under multiple different stresses. The wavelength shift was identical to the stress calibrated by a commercial force meter and showed good repeatability and stability. Using the same characterization method, the commercial FBG with diameters of 125 μm and 80 μm had low sensitivities of 0.02 nm/N and 0.04 nm/N, respectively (Fig. 2(b), inset). The stress response of the soft μFBG patch was enhanced by 2 orders of magnitude compared with that of the commercial FBG sensors in Fig. 2(f) because the ultrafine diameter (12 μm) increased the strain under the common stress and the evanescent field^{35, 36} introduced the variation in the microfiber’s effective refractive index.

The soft μFBG patch also had a great response to low-frequency vibration, benefiting from the ultrathin thickness of the PDMS patch. As shown in Fig. 2(g−i), the waveform was sinusoidal, and the amplitude was identical when detecting vibrations with frequencies of 1 Hz, 2 Hz, and 5 Hz, ensuring waveform accuracy when detecting human physiological mechanical signals, which tend to be low-frequency vibration signals. Also, the μFBG patch can successfully measure vibrations up to 100 Hz (Supplementary information Fig. S3). By designing the structure of the microfiber and the demodulation equipment, higher frequency vibration signals can be acquired³⁷. A 10000-circle repetition test was performed with the same stress at a frequency of 1 Hz, as shown in Fig. 2(j), showing the great repeatability and robustness of the soft μFBG patch.

Owing to the skin-like mechanical compliance and excellent stress/vibration response, the skin-like μFBG patch can maintain stable contact with human skin in different body locations and detect multiple physiological signals. The subjects were instructed to relax their bodies and to remain calm during the measurement. Ballistocardiogram (BCG) signals, as the near-end signals of the cardiovascular system and pulse waves at multiple far-end body sites, were detected to analyze the propagation process of pulse waves and monitor hemodynamic parameters. An FBG interrogator was employed to record the wavelength shift of the soft μFBG patch caused by micro stress on the skin surface due to cardiovascular activity with a sampling frequency of 2500 Hz.

As shown in Fig. 3(a), the soft μFBG patch was attached to the chest skin at the tricuspid site to detect the BCG signal. A small amount of medical alcohol was applied to make the μFBG patch adhere conformally to the skin surface. After the alcohol evaporated, the patch maintained close contact with the skin. The raw BCG signal (Supplementary information Fig. S4(a)) was mixed with the respiration signal caused by the periodic expansion of the chest, which was represented as baseline shift and had a low frequency (< 1 Hz). By employing the short-time Fourier transform (STFT), the time-frequency diagram of the raw BCG signal was obtained (Supplementary information Fig. S4(b)). The frequency components related to heart rate were clearly identified in the time-frequency diagram. Therefore, a 3–15 Hz bandpass filter was employed to eliminate the low-frequency respiration signal and the high-frequency noise. After filtering, the BCG signal relative to periodic cardiovascular activity stood out both in the time domain and in the time-frequency domain (Supplementary information Fig. S4(c) and S4(d)). The heart rate period can be distinguished significantly (see the triangle marks in Fig. 3(b)). Figure 3(c) shows the details of BCG signal. Among the feature points, the I wave represents the beginning of cardiac ejection. The purple region identified by the I wave and K wave indicates the systole of the heart rhythm. The spectrum signature of the BCG signal is illustrated in Fig. 3(d).

When detecting the pulse wave, the soft μFBG patch was attached to the skin at the superficial artery sites, as shown in Fig. 3(e). The typical pulse wave at the radial artery is shown in Fig. 3(f). As shown in Supplementary Fig. S4(e) and S4(f), the raw pulse wave contained the baseline shift caused by the movement of the arm and obvious high-frequency noise. Therefore, a 0.5–10 Hz bandpass filter was employed to remove them (Supplementary information Fig. S4(g) and S4(h)). Additionally, the details of the pulse wave are shown in Fig. 3(g). The B wave tends to represent the beginning of cardiac ejection, and the purple region identified by the B wave and D wave indicates the systole of the heart rhythm. The spectrum signature of the pulse wave is illustrated in Fig. 3(h). The I wave of the BCG signal and the B wave of the pulse wave both represent the beginning of cardiac ejection. Therefore, we use the two feature points to calculate the PTT and to analyze the propagation process of the pulse wave.

The signal amplitude defined by the maximum wavelength shift is a key factor in determining the signal quality. Benefiting from the high sensitivity of the soft μFBG patch, the signal amplitudes of the BCG signal and radial pulse wave were up to 35 pm and 80 pm, respectively, guaranteeing the accuracy of hemodynamic monitoring. Using the same detection method, the commercial FBG had a signal amplitude of less than 2 pm, which was hard to detect by optical signal acquisitions, resulting in poor waveform accuracy (Supplementary information Fig. S5).

Owing to the multichannel and time-synchronized operation capability, the soft μFBG group was built by connecting two μFBG patches in series. The μFBG group was capable of detecting the BCG signal and pulse wave synchronously. The I wave of the BCG signal and the B wave of the pulse wave were used to calculate the PTT. By detecting the pulse wave at different superficial artery sites, systemic PTT was available to thoroughly evaluate the status of the cardiovascular system. As shown in Fig. 4(a), the PTTs of three different cases (heart to carotid artery (CA), heart to radial artery (RA), and heart to pedal artery (PA)) were detected and calculated on a healthy man with a body height of 172 cm. Figure 4(b−g) shows the waveforms and details of the synchronous BCG signal and pulse wave. The purple regions indicate each PTT.

The morphology of pulse waves at the three superficial artery sites differed from each other. The pulse wave at limbs tended to possess a steeper waveform compared with the one at neck. This phenomenon is caused by the backward propagation of pulse waves at arterioles, which is called the amplification effect³⁸. Compared to the central site, such as the neck, the reflected pulse wave at the peripheral site travels a much shorter distance and thus instantly contributes to the waveform, resulting in a higher amplitude. In addition, the pulse wave at limbs tended to have a later dicrotic wave compared with the wave at the neck. The dicrotic wave is generated from the blood reflux flowing to the artery again due to the closure of the aortic valve. For the pulse wave at peripheral sites, the blood reflux travels a longer distance from the heart, thus contributing to the increasing time interval between the systolic peak and the dicrotic wave.

The PTTs of the three cases differed from each other, as shown in Fig. 4(h). The pulse wave propagating from the heart to the CA needed the shortest PTT of 5.7 ms in the average. In comparison, the average PTTs of the heart to the RA and the heart to the PA were 27.4 ms and 107.9 ms, respectively. The validation of PTT is due to the different artery lengths and pulse wave propagation velocity. The pulse wave at the aorta has a relatively slow PWV (~5 m/s) because of the large elastic modulus³⁹, and the pulse wave at the arteriole has a quick PWV (15–35 m/s)⁴⁰. Therefore, the PTT from the heart to the CA was very short, and the pulse wave propagating from the heart to the PA required a long time considering an artery length of more than 1 meter. Relying on the multichannel and time-synchronized operation capability, the soft μFBG group realized systemic PTT calculation. Moreover, it is capable of detecting the all-mechanical process of pulse wave propagation and achieving a more accurate propagation time of pulse waves than pulse arrival time (PAT) by eliminating the pre-ejection period (PEP) delay⁴¹. Based on the above advantages, the μFBG group demonstrates the potential for accurate clinical diagnosis and early screening of lesions in the cardiovascular system.

Under the load from exercise, there are many changes in hemodynamics due to increasing metabolism. Research on hemodynamic variation law during exercise can provide an early diagnosis of cardiovascular disease and comprehensively evaluate cardiovascular system function^{42, 43}. Unlike a commercial cuff-type sphygmomanometer, the soft μFBG group is capable of detecting physiological signals in real time and monitoring hemodynamics continuously. Supplementary information Fig. S6 shows an operational flowchart of hemodynamic monitoring using the soft μFBG group. One patch of the μFBG group was attached to the chest skin, and the other was attached to the left foot. After signal filtering, the near-end BCG signal, far-end pulse wave, and respiration signal were extracted with great temporal synchronization, as shown in Fig. 5(a). The I wave of the BCG signal and the B wave of the pulse wave were used to calculate the HR and PTT. The subject lay on the bed and exercised by pretending to ride a bicycle with the right foot alone. At the same time, a commercial cuff-type sphygmomanometer was employed on the left arm of the subject to provide the gold standard HR and BP. Figure 5(b) and 5(c) demonstrate typical waveforms of the BCG signal and pulse wave when the subject was in static and in exercise, respectively. As shown in Fig. 5(d), the HR was increased at the beginning of exercise and decreased to a normal level after 3–4 minutes at the end of exercise. The BCG signal and pulse wave at several typical time points are shown in Supplementary information Fig. S7. At the same time interval, there are more cardiac cycles in exercise than in static. The HR achieved by the soft μFBG group was consistent with the results of the sphygmomanometer, showing the accuracy of hemodynamic monitoring. During exercise, another interesting phenomenon was that the PWV (proportional to 1/PTT) significantly increased along with the systolic blood pressure, which is shown in Fig. 5(e). This phenomenon is due to the contraction of nonexercise muscle blood vessels caused by sympathetic stimulation in exercise. The contraction leads to increased arterial pressure and accelerated blood flow. The variation in HR and PTT achieved by the soft μFBG group was repeatable under multiple exercise periods, demonstrating the potential for dynamic hemodynamic monitoring.

PR reflects the unobstructed level of the cardiovascular system, and PR monitoring is significant for the diagnosis of cardiovascular diseases such as arteriosclerosis, thrombus, and hypertension^{44, 45}. Relying on the PTT, the dynamic variation in PR was successfully monitored by our soft μFBG group. Figure 6(a) shows a schematic diagram of the PR monitoring experiment. One patch of the μFBG group was attached to the chest skin to detect the BCG signal, and the other was attached to the left wrist to detect the pulse wave. An inflatable cuff was employed at the left upper arm to impose external pressure on the artery to simulate the variation in PR. The external pressure rapidly increased and slowly decreased in the process. The continuous BCG signal and pulse wave are shown in Fig. 6(b), and the green region indicates the compression period. The BCG signal was stable, but the pulse wave was disturbed and even interrupted because high external pressure blocked the blood flow. It is a notable feature that PTT increased with the external pressure when the pulse waveform was recognizable. When the external pressure increased, the transmural blood pressure defined by the differential value between the internal pressure (also known as BP) and external pressure of the artery decreased, resulting in slow PWV (for a detailed analytical model of the artery, see Supplementary information Materials and methods and Fig. S9). The response of PTT to the external pressure was reproducible, demonstrating the soft μFBG group’s potential for PR monitoring. Benefiting from the multichannel operation capability, the soft μFBG group can monitor the regional PR of the artery and have a potential application in rapid thrombus positioning.

We have demonstrated a soft μFBG group that offers continuous and noninvasive monitoring of spatiotemporal hemodynamics. Combining flexible material integration with precise micro/nanostructure design, the soft μFBG patch achieves both a great sensing response on stress/vibration and intimate coupling with human skin with suitable mechanical compliance. Thanks to its ultrafine diameter and ultrathin thickness, this sensor’s sensitivity has been increased by two orders of magnitude compared to commercial FBGs. Exploiting the advanced femtosecond laser fabrication technique and quasi-distributed optical fiber sensing technique, high-quality all-mechanical physiological signals such as the near-end BCG signal and far-end pulse wave are detected to reflect the actual mechanical process of blood circulation. Relying on the multichannel and time-synchronized operation capability, the soft μFBG group can be used to monitor systemic hemodynamic parameters and reflect regional artery statuses. The device satisfies clinical requirements for real-time and continuous hemodynamic monitoring and reconfigurable regional health evaluation of the cardiovascular system. In addition, electromagnetic immunity makes the μFBG group compatible with essential medical imaging, especially magnetic resonance imaging. Applications in systemic PTT calculation, dynamic hemodynamic monitoring during exercise, and regional PR monitoring demonstrate the great potential in the diagnosis of cardiovascular diseases such as arrhythmia, angiosclerosis, hypertension, and thrombosis and can facilitate precise clinical diagnosis, the fast screening of lesions, and daily health management.