A cylindrical projection lithography-fabricated flexible on-catheter in situ integrated sensor for continuous in-artery blood pressure monitoring

Abstract

Continuous and accurate blood pressure monitoring is vital for the diagnosis and management of life-threatening cardiovascular diseases. Conventional fluid-filled catheter systems are fundamentally constrained by distortion, infection risks, and limited portability. In parallel, Micro-Electro-Mechanical System (MEMS) sensors assembled to catheters face miniaturization bottlenecks and packaging complexity. Here, we present a flexible on-catheter interventional pressure sensor (CIPS), leveraging suspended-graphene arrays for pressure sensing and dual polymer encapsulation layers for hermetic and waterproof encapsulation. The highly integrated CIPS with compact architecture is enabled by cylindrical projection lithography (CPL)-based 3D in situ MEMS fabrication process. By introducing interdigital electrodes (IDEs) in sensitive areas and optimizing structural parameters, CIPS delivers an ultrafast response (<0.36 s), exceptional sensitivity (3.5 × 10⁻⁶·mmHg⁻¹), and a broad detection range (6–380 mmHg), outperforming state-of-the-art suspended-graphene pressure sensors. The dual-layer polymer encapsulation strategy ensures the hermetic and waterproof integrity of CIPS. Crucially, CIPS enables real-time monitoring of arterial pressure in the rat abdominal aorta, marking the first demonstration of suspended graphene in animal-level bioelectronic sensing. This work has also underscored the translational potential for CPL-based 3D in situ MEMS fabrication strategy.

Introduction

Arterial blood pressure is a vital physiological parameter, and its persistent deviations from homeostatic ranges have been clinically correlated with life-threatening cardiovascular diseases, such as heart failure, coronary artery disease, and stroke, which are leading causes of mortality^1,2. Continuous and timely arterial pressure monitoring is essential for patient assessment and outcome improvement, especially during cardiac or neurosurgical procedures^3,4. Arterial puncture and catheterization is the gold standard for continuous artery blood pressure measurement. However, this paradigm relies on long fluid-filled catheters that introduce challenges such as waveform distortion, infection risk, limited mobility, and system complexity. Direct blood pressure monitoring inside blood vessels became feasible thanks to the emergence of the MEMS pressure transducer in the 1970s, offering more accurate and steady signals compared to conventional monitoring schemes.

The millimeter-scale vascular lumen poses an engineering challenge in developing miniaturized interventional and implantable pressure sensors. Conventional MEMS technologies have enabled millimeter-to-submillimeter pressure chips integrated with microcatheters^5–10, yet further miniaturization is necessary for relieving patient’s discomfort and inflammation. However, further miniaturization is hindered by substrate fragility and reduced sensitivity from smaller sensing units. Nanostructure-based approaches, such as silicon nanowires, improved sensitivity but introduced nonlinearity and complex circuitry^11–13. Polymer-based sensors offer flexibility, low modulus, and biocompatibility, and have demonstrated high-fidelity blood pressure detection^14–18. However, their reliance on metallic electrodes limits electromechanical sensitivity, necessitating urgent development of novel sensing materials to synergistically enhance performance.

Graphene has emerged as a promising candidate for next-generation MEMS/NEMS technology due to its exceptional electromechanical properties stemming from its unique two-dimensional structure and atomically precise crystal lattice^19–21. The material’s pronounced electromechanical performance positions it as an exceptional candidate for pressure sensors²². Notably, graphene sustains superior stress tolerance compared to silicon films despite having a thickness three orders of magnitude smaller than that of silicon. Furthermore, its fabrication protocols demonstrate full compatibility with existing MEMS processes. A representative “suspended graphene” structure serving as the basic sensing element for graphene-based pressure sensors was reported in 2016, in which the graphene was transferred to the surface of the SiO₂ layer with a cylindrical cavity formed by DRIE²³. Subsequent efforts optimized this configuration by constructing a large-scale suspended graphene array, achieving remarkable sensitivity via synchronous deformation of thousands of suspended graphene under pressure loading (Fig. 1c. (i))^23–26. However, during the removal of the supporting layer, some suspended graphene tends to collapse due to capillary effects, resulting in low productivity for such devices. Meanwhile, random gas pathways exist between graphene and the substrate that permit cavity gas leakage, leading to unstable device performance^27,28. Additionally, bare graphene directly exposed to ambient conditions leads to deteriorated sensor stability over time and precludes deployment in liquid environments. Some studies retained the supporting layer after transferring graphene to enhance the productivity of suspended graphene, but these devices were still not hermetic and waterproof (Fig. 1c. (ii))^29–31. Robust encapsulation thus becomes imperative for reliable blood pressure monitoring. Furthermore, none of these studies reported the sensor’s response characteristics to rapid pressure application (pressure surges <0.5 s), which is crucial for dynamic pressure monitoring (such as blood pressure changes).

Fig. 1. Schematic illustration of CIPS with performance promotion strategies for continuous intra-arterial blood pressure monitoring.

a Conceptual depiction of CIPS application scenario (Left, CIPS enters the blood vessel via an indwelling needle) and explosion view of its structural design (right, suspended graphene arrays provide sensing capabilities, interdigital electrodes enhance the sensor’s performance, and double-layer PDMS ensures hermetic and waterproof encapsulation). b Comparison of in situ integrated manufacturing on catheters with conventional assembly strategies for standalone sensor and catheter. c Comparison of encapsulation and design strategies for suspended-graphene pressure sensors. d, e Optical images of the fabricated CIPS showing its compact structure and flexibility. f CIPSs with different diameters

Beyond pursuing the limits of dimension and performance, methodologies for deploying intracorporeal pressure sensors are also critical. Intravascular catheters routinely incorporate pressure sensors for continuous blood pressure monitoring. Sensors are mounted on the end of the catheter at the early stage (Fig. 1b.(i)), yet suffer from the detection error induced by perpendicular impact between the sensor and the blood flow³². A paradigm shift thus emerged with sidewall-mounted sensors (Fig. 1b.(ii))^5,18,33,34. However, both configurations inherently compromise the cross-sectional profile through interconnection wiring and waterproof packaging requirements. This fundamental constraint necessitates innovative sensor architectures and packaging solutions for intravascular applications.

Our group pioneered MEMS-compatible cylindrical projection photolithography (CPL)^35,36 for enabling in situ fabrication of flow³⁷, temperature³⁵, and glucose sensors³⁸ on the polyimide (PI) catheter’s outer surfaces. In these sensors, the substrate’s geometrical feature of large surface area (in the magnitude of square millimeter upon a 300 μm diameter cylindrical substrate) is utilized to realize outstanding sensor performance. Therefore, the MEMS-compatible CPL technology provides an effective scheme to solve the “dimension-performance” trade-off of pressure sensors. Besides, this in situ fabrication strategy will save the packaging space, thus reducing the overall sensor’s dimension.

Here, we designed and manufactured a flexible catheter-based interventional pressure sensor by in situ integrating all sensing components directly onto the medical catheter’s peripheral surface. Electrode contacts and IDEs were directly fabricated around the catheter’s surface via CPL. Multi-layer graphene and the cavity array constructed the sensitive structure of the pressure sensor, which is sealed by two polydimethylsiloxane (PDMS) layers for hermetic and waterproof encapsulation, adding less than 15 μm to the catheter’s diameter, thus keeping a compact architecture. Moreover, the inner lumen of the catheter remained unoccupied in this design, preserving its original functionalities such as drug delivery and integration of optical fibers for imaging or laser therapy. Furthermore, all constituent materials (gold, graphene, PI, and PDMS) are biocompatible, ensuring compliance with medical device safety standards. The CIPS showed a fast response (<0.36 s) to rapid pressure surge, a wide detection range (6–380 mmHg), and an outstanding sensitivity (3.5 × 10⁻⁶·mmHg⁻¹) enabled by the introduction of IDEs and structure design. Moreover, in-vivo blood pressure monitoring within the abdominal aorta of rats was demonstrated by CIPS, validating the feasibility of implementing suspended-graphene sensors and highly integrated catheter-based devices in clinical applications.

Results and discussion

Design and simulation of CIPS

Intravascular detection of blood pressure provides more accurate signals compared to indirect measurement. The application paradigm of sensors within arteries is also critical, as it directly determines the validity of measurements and patient comfort. The use of an indwelling needle in radial artery puncture for intravascular blood pressure monitoring offers several advantages. It facilitates a convenient and reliable connection between artery and pressure sensing systems in a minimally invasive way while maintaining a closed and sterile pathway and eliminating the need for repeated punctures, minimizing vascular injury and infection risk, serving as an ideal interface for interventional MEMS pressure sensors for continuous hemodynamic monitoring. Following radial artery puncture, the CIPS is inserted into the radial artery via an indwelling needle, after which the interconnect cable is connected to an external readout system or wireless signal transmitter module (Fig. 1a).

The developed CIPS is architected with three functionally integrated components: the flexible PI catheter with cavity array, suspended multilayer films (SMF, graphene/1st PDMS/2nd PDMS), and IDEs. Detailed structure and parameters of the IDE and cavity array are presented in Fig. S1. The cavity (blind hole) array fabricated by the femtosecond laser is embedded in the PI substrate to accommodate the deformed SMFs. Circular cavity is selected in this research for its better contribution to the sensor’s performance than rectangular and square counterparts³⁹. Graphene is selected as the sensing material as the sensor’s transduction mechanism is based on its intrinsic piezoresistive effect. Multi-layer (6–8 layers) graphene is adopted in this research for better mechanical stability⁴⁰. PDMS is selected because of its ultra-low Young’s modulus, which can generate more deformation under pressure loading, increasing the sensor’s sensitivity. The multilayer films involving graphene and PDMS layers acting as the deformation layer, together with the cavity array, construct the SMF array structure, which is the basic sensing component of the sensor. The introduction of the IDEs can decrease the initial resistance (R₀) of the device, thus improving the sensor’s response speed and other performance, which will be analyzed in detail later. The designed architecture expands the effective sensing area to the millimeter scale while maintaining a sub-millimeter catheter profile through circumferential unit arrangement.

The in situ integration scheme minimizes radial expansion (~15 μm), achieving a compact architecture with obvious thickness reduction in cross-sectional profile versus conventional end-mounted and side wall-mounted configurations (Fig. 1b). This paradigm balanced the “dimension-performance” dilemma, elevating the sensor’s performance while keeping a low cross-sectional profile. This merit, together with the reasonable packaging strategy and the IDEs enhancing the CIPS’s performance, will contribute to reliable intravascular blood pressure detection.

In our encapsulation strategy, two layers of PDMS encapsulate the graphene and the whole sensor (Fig. 1c. (iii)). The 1^st PDMS acting as the supporting layer for the graphene can improve the productivity of the suspended structure after transferring graphene. The 2^nd PDMS layer can block gas pathways between graphene and the substrate, achieving hermetic and waterproof sealing simultaneously.

The CIPS with 500 μm outer diameter (OD) in the normal and bending state is shown in Fig. 1d (magnified graph shows the entire sensor structure), in which almost no extra space can be observed beyond the PI catheter. Meanwhile, the sensor shows excellent flexibility and can be inserted into the tortuous artificial transparent blood vessel (Fig. 1e). Besides, the sensor can be fabricated on catheters with different OD depending on various application purposes, showing the adaptability of the above design strategies and the process technologies (Fig. 1f).

The SMF serves as the basic sensing element of CIPS, and its structural parameters directly determine the sensor’s performance. Reaching a large strain of the graphene ensures a greater strain increment brought by applied pressure, thus improving the sensor’s sensitivity. Meanwhile, limiting the stress of PDMS within its elastic limit is also critical for avoiding mechanical creep or plastic deformation. In our design, the depth and the diameter of the cylindrical cavity are 15 μm and 30 μm, and the thickness of the dual-PDMS layers is 15 μm (Fig. S2). The cavity depth of 15 μm was specifically chosen to mitigate the ‘sealed gas effect’, ensuring that the internal pressure buildup does not dampen the membrane’s deflection²⁹. Mechanical simulation using COMSOL Multiphysics was conducted to quantitatively analyze the mechanical behaviors of the SMF undergoing different pressure loadings. The results show that the maximum deflection of the SMF occurs at the center of the sensitive area, and it is positively proportional to the applied pressure (ranging from 38 mmHg to 380 mmHg) (Fig. 2a, b), which is consistent with the mechanical relationship between the deflection and the applied pressure:

$$\frac{P}{d}=\frac{d^2}{3a^4}\cdot \frac{\mathit{Et}\left(7-\nu \right)}{1-\nu }+\frac{4\sigma t}{a^2}$$ 1

where P is applied pressure; d, E, $\sigma$, $\nu$, and t, are the deflection at the film’s center, Young’s modulus, stress, Poisson’s ratio, and thickness of the PDMS layer, respectively; $a$ is the cavity radius. Furthermore, the maximum stress and strain occur at the edge of the SMF and is linearly proportional to the applied pressure (Figs. S3, S4). The maximum Von Mises stress at 380 mmHg is 190 kPa, far from the PDMS’s elastic limit (Fig. 2c)⁴¹. The graphene’s strain should be equal to the PDMS’s strain and can be expressed as:

$$\epsilon =\frac{2d^2}{3a^2}$$ 2

Fig. 2. Mechanical simulation of the basic sensing unit under applied pressure.

a Simulation results of graphene’s deflection under different pressures. b Deflection distribution of the graphene along the central line X. c Maximum von Mises Stress and strain of the graphene layer under different pressures

The maximum strain of the SMF (21%) approaches the maximum elastic strain of graphene (Fig. 2c)⁴², indicating significant signal generation without irreversibly destroying the sensitive material. In summary, the simulation results validate the rationality of the designed parameters for the basic sensing element, establishing a reliable foundation for subsequent sensor fabrication.

The influence of the IDE on the sensor’s output signal was also investigated. The sensitivity S of the pressure sensor is:

$$S=\frac{\Delta R/R_0}{\Delta P}$$ 3

where R₀ is the device’s total initial resistance, and ΔR₀ is the resistance change under the pressure change ΔP. According to the above equation, decreasing the initial resistance R₀ is a viable way to improve S, which can be achieved by introducing IDEs. This effect will be further amplified if the Wheatstone bridge is utilized⁴³. The sensing region can be sliced into several parallel segments after introducing IDEs, and the number of segments equals to the number of IDE N_IDE. In one segment before introducing IDE (Fig. S5), its resistance R_s and R₀ can be expressed as:

$$R_{\mathrm{S}}=R_{\mathrm{g}1}+\frac{1}{\frac{1}{R_{\mathrm{g}3}}+\frac{1}{R_{\mathrm{g}}}+\frac{1}{R_{\mathrm{g}4}}}+R_{\mathrm{g}2}$$ 4

$$R_0=\frac{R_S}{N_{\mathit{IDE}}}$$ 5

where R_g is the resistance of the graphene in the sensing area, and R_g1, R_g2, R_g3, and R_g4 are the graphene’s resistance outside the sensing region. The relationship between the ΔR_S and ΔR_g is:

$$\Delta R_{\mathrm{S}}\approx \frac{\Delta R_{\mathrm{g}}}{{\left(1+\frac{R_{\mathrm{g}}}{R_{\mathrm{g}3}}+\frac{R_{\mathrm{g}}}{R_{\mathrm{g}4}}\right)}^2}$$ 6

Introducing IDEs constructs external conducting pathways under graphene, leading to a decline to R_g1, R_g2, R_g3, and R_g4, thus R_S and R_tot decrease. Besides, introducing more IDE (increase in N_IDE) can further decrease R_tot according to Eq. (5). Significant enhancement in the sensor’s sensitivity can be achieved through the proposed strategies according to Eq. (3). From a physical mechanism perspective, the total resistance (R₀) without IDEs of the device is dominated by long graphene segments on the substrate (non-suspended regions). These regions act as “parasitic resistors”—they contribute significantly to the base resistance but generate negligible resistance change (ΔR ≈ 0) under pressure, thereby diluting the overall sensitivity signal (ΔR/R₀). As derived in Eqs. 4–6, the introduction of IDEs creates parallel conductive pathways underneath the graphene. This effectively bypasses the high-resistance non-sensing regions. By minimizing the electrical weight of these parasitic regions, the device’s total resistance becomes dominated by the suspended, high-strain active regions. Consequently, the relative resistance change (ΔR/R0) is significantly amplified, resulting in higher sensitivity, highlighting the critical role of the incorporated IDEs in optimizing the sensor’s performance.

Sensor fabrication

The CIPS’s fabrication process mainly comprises CPL, lift-off, graphene wet transfer, and packaging process (Fig. 3a). The process was initiated by dip coating the positive photoresist (PR) on the outer surface of the PI catheter. Then, the CPL and development were conducted to define the pattern of IDEs and connecting electrodes. Sequential sputtering of chromium/gold (Cr/Au) followed by standard lift-off processing yielded metallic structures of IDEs and connecting electrodes. The femtosecond laser subsequently ablated the substrate surface between IDEs to form the cavity (blind hole) array. Although several advanced dry^26,44 and bubbling⁴⁵ strategies for transferring graphene onto object substrates were reported to obtain suspended graphene arrays with high yield, we planned to transfer the PDMS/graphene layer on the outer surface of the above substrate via the wet transfer for its outstanding shape adaptability for transferring ultra-thin polymer films on the irregular substrate^46,47. Concretely, the 1^st PDMS layer was spin-coated on the commercially available copper-based graphene sheet. Then the copper was etched to release the graphene/PDMS layer, which was then transferred to the PI catheter with electrode pattern and cavity arrays. Finally, two enameled wires were mounted to the connecting electrode by penetrating them from the inner space through laser-drilled via holes (Fig. S6) and a 2^nd layer of PDMS was spin-coated on a special custom-made tool to seal the whole device (Fig. S7). Complete fabrication details are provided in the Methods section.

Fig. 3. Fabrication process and images of CIPS.

a 3D and cross-sectional view of the 3D in situ fabrication process of the highly integrated catheter-based pressure sensor. SEM images of b IDEs and c cavity array on PI catheter (bar = 500 μm). d Optical images of the SMF structure (bar = 50 μm). e Cross-sectional view showing the material components of the fabricated pressure sensor (bar = 20 μm)

The scanning electron microscopy (SEM) images demonstrated that IDEs and cavity arrays around the PI catheter were successfully fabricated (Fig. 3b, c). The length and width of a single IDE are 2200 μm and 50 μm, respectively. The depth of the cavities was measured by an ultra-depth field microscope and was 14 μm on average (Figure S8). Unlike the cavities formed in an additional layer reported in most pressure sensors, the cavity array ablated by a femtosecond laser in this work is embedded in the substrate. It is well-known that the cavity is space-consuming for the pressure sensor because it offers a sealed space to form a pressure difference with the outside environment and accommodate the deformation of the sensitive layer. In our design, the cavity’s space is transferred to the substrate, largely reducing the overall dimension of the pressure sensor. The supporting layer is usually served as a sacrificial layer and removed in most suspended graphene sensors, which is prone to inducing structural failure due to graphene’s mechanical fragility. In our process, the removal operation was canceled, and the supporting layer of PDMS was reserved to protect the fragile graphene, thus almost all the SMF structures were intact, demonstrating the feasibility of our process for improving the yield and ensuring the high sensitivity and stability for the sensors (all SMFs participate in the pressure response) (Fig. 3e). The fabricated sample was cut by a triple ion beam cutter, and the SEM image of the exposed cross-section surface reveals that the thicknesses of the dual PDMS layers and the Cr/Au layer are 15 μm and 220 nm, respectively (Fig. 3f). Energy-dispersive X-ray spectroscopy (EDS) elemental mapping identifies distinct nitrogen- (green), silicon- (blue), and gold- (yellow) enriched regions, corresponding to PI substrates, PDMS, and Au/Cr electrodes, respectively (Fig. S9). Besides, the enameled wires were mounted to the connecting electrodes on the outer surface of the catheter by penetrating them from the inner surface through laser-drilled via holes. It would be less space-consuming and more robust to survive the mechanical destruction during the implantation or intervention process by this method compared to exposing the connecting wires outside the catheter. These collective results demonstrate the successful in situ fabrication of highly integrated CIPS via the 3D MEMS process based on the CPL technology.

Influence of IDE and sensor characterization

Introducing IDEs on the substrate’s surface creates additional conductive pathways, thereby reducing the device’s overall resistance R₀ (Eq. (5)). Lowering R₀ brings several merits: sensor sensitivity increases (Eq. (3)); the sensor’s RC time constant (τ = RC) is significantly reduced, decreasing response delay, and shortening response time (t_res); and sensor thermal noise (TN) is suppressed, enhancing the sensor's signal-to-noise ratio (SNR) (Fig. 4a). A systematic investigation was first conducted to evaluate the impact of introducing IDEs (Detailed testing procedures are provided in the Methods section, and schematic illustration and optical image of the testing system are shown in Fig. S10). The CIPSs’ performance with 0, 4, and 8 IDEs (Fig. S11) was systematically investigated. All three sensor configurations with varying numbers of IDEs maintained identical cavity counts (29 × 8 = 232), and the rotation speed for casting the 1^st and 2^nd PDMS were 3000 rpm and 6000 rpm. The R₀ of the CIPS sensors equipped with 0, 4, and 8 IDEs are 3301 Ω, 131 Ω, and 59 Ω, respectively. As the number of IDEs increases, the sensor resistance decreases by approximately two orders of magnitude (Fig. 4b). We then examined the response levels of CIPS sensors under varying air pressures. The response levels of all sensors increased linearly with rising pressure. Three groups of CIPSs exhibited linear piezoresistive responses (R² > 0.99) with excellent inter-device consistency, and the sensitivity increased as the quantity of IDE increases (the average sensitivity of CIPSs with 0, 4, and 8 IDEs were 5.53 × 10⁻⁷, 1.27 × 10⁻⁶, and 3.53 × 10⁻⁶mmHg^-1, respectively) (Fig. 4c). The proposed CIPS with PDMS encapsulation layer exhibited a high sensitivity of 3.53×10^-6mmHg⁻¹, which is even comparable to bare suspended-graphene pressure sensors without any encapsulation layer^25,26, demonstrating the synergetic effect of introducing IDE, material selection, and structural design for enhancing the sensor’s sensitivity. Real-time response curves of three types of sensors at 116 mmHg pressure (signals processed through a low-pass filter with a cutoff frequency of 12 Hz) show that the sensor response time shortens as the quantity of IDE increases—from 6.74 s for 0 IDE to 0.36 s for 8 IDEs (actual response times may be shorter since this response includes the pressure application process) (Fig. 4d). The accelerated response speed by introducing IDEs enhances the sensor’s dynamic performance, establishing a foundation for detecting rapidly changing pressures (such as blood pressure fluctuations). In addition to enhancing response speed, the real-time response curve of the sensor at 116 mmHg (unfiltered signals) demonstrates that introducing IDEs significantly improves signal quality (Fig. 4e). After introducing IDEs, the sensor’s root mean square noise (RMS_noise) decreased by a factor of 73 (from 0.146 Ω without IDE to 0.002 Ω with 8 IDE). The reduction in noise levels and increased sensitivity achieved by introducing IDEs directly resulted in a significant improvement in SNR (from 7.74 dB to 16.9 dB), thereby lowering the sensor’s minimum detection limit (from 69 mmHg to 6.38 mmHg) (Fig. 4f). A comprehensive comparison of sensors with different IDE quantities was conducted based on five metrics: sensitivity, response speed, signal-to-noise ratio (SNR), root mean square noise (RMS noise), and detection limit (the reciprocal transformation was applied to smaller-is-better metrics to facilitate intuitive evaluation through radar charts) (Fig. 4g). The introduction of IDE resulted in comprehensive improvements across all five metrics, with the sensor with 8 IDEs demonstrating the best overall performance.

Fig. 4. Performance enhancement by introducing IDEs.

a Principle of CIPS’s performance improvement by introducing IDEs. Comparison of b resistance and c resistance change rate under different applied pressures of CIPSs with different IDE quantities. d Real-time response of CIPSs with different IDE quantities to a pressure surge of 116 mmHg (signals processed through a low-pass filter with a cutoff frequency of 12 Hz). e Comparison of signal quality of CIPSs with different IDE quantities (unfiltered real-time response to a sustained pressure of 116 mmHg). f Comparison of SNR at 116 mmHg pressure level and detection limit of CIPSs with different IDE quantities. g Radar chart for comprehensive performance comparing of CIPSs with different IDE quantities. h Radar chart for comprehensive performance comparing between 8 IDEs CIPSs with 232 cavities and 4 IDEs CIPSs with 348 cavities

A comparative analysis between CIPSs configured with 8 IDEs (232 cavities) and those with 4 IDEs (348 cavities, Fig. S12) was then conducted. As for CIPSs with 4 IDEs, adding cavities elevated the sensor’s response level and sensitivity comparable to that of CIPSs with 8 IDEs (Fig. S13). A comprehensive comparison of sensors with different IDE quantities was conducted based on five metrics: sensitivity, response speed, signal-to-noise ratio (SNR), root mean square noise (RMS_noise), and non-linearity (NL). Although the samples with 4IDEs and 348 cavities demonstrated a slight advantage in both sensitivity and SNR, the samples with 8IDEs and 232 cavities exhibited superior overall performance (Fig. 4h). Especially, the former’s t_res was much longer (2.7 s) than the latter (0.41 s), which is unacceptable for continuous dynamic monitoring of blood pressure.

These results collectively indicate that introducing IDEs can significantly optimize the overall performance of the CIPS, which is crucial for achieving high sensitivity, fast response, and high reliability. CIPS with 8 IDEs and 232 cavities demonstrated the best overall performance and will be adopted in subsequent experiments.

We subsequently investigated the influence of the spin speed for casting the 1^st PDMS on the CIPSs’ performance. The samples with 3000 rpm exhibited the highest overall response level and sensitivity, followed by the samples with 6000 rpm, with the samples with 4500 rpm showing the lowest (Fig. 5a, b). This finding contradicts the established principle that thinner sensing layers (achieved through higher spin speeds) enhance response sensitivity. However, microscopic analysis revealed critical morphological evidence explaining this deviation (Fig. S14). Unlike the samples with 3000 rpm, the SMFs exhibited capillary action-induced abnormal deformation (sucked by the surface tension of water trapped in the cavity) for the samples with 4500 rpm and 6000 rpm, and were collapsed for the samples with 6000 rpm. Some air pathways for water evaporation are blocked by the better bonding between the PI substrate and the thinner multi-layer film, trapping the water in cavities. The high spin speed of 6000 rpm forms the ultra-thin PDMS layer, and its bending rigidity is too low to maintain a normal suspending state, rendering it collapses into the cavities. As a result, the samples with 3000 rpm presented the highest sensitivity (3.53 × 10⁻⁶ ± 8.41 × 10⁻⁸mmHg⁻¹), the best consistency across sensors (minimum standard error (SE)), and minimum NL (2.20 ± 0.23%) because most of the normal SMFs contributing to the pressure response (Fig. 5b). The samples with 4500 rpm presented the lowest sensitivity (2.03 × 10⁻⁶ ± 6.641 × 10⁻⁸mmHg⁻¹) and medium NL (3.68 ± 0.54%) because of the existence of randomly distributed water-sucked SMFs. The samples with 6000 rpm presented medium sensitivity (2.83 × 10⁻⁶ ± 1.69 × 10⁻⁷mmHg⁻¹), the worst consistency across sensors (highest SE), and the highest NL (4.52 ± 1.29%) because of the existence of the most randomly distributed abnormal (water-sucked and collapsed) SMFs, and its sensitivity was higher than the samples with 4500 rpm may thanks to the ultra-thin PDMS layer. Collectively, the 3000 rpm samples present the best overall performance and would be adopted in the following experiments.

Fig. 5. CIPSs’ (with 8 IDEs) performance under various pressure tests.

a Response level under different applied pressures and b sensitivity and non-linearity of CIPSs with different rotation speeds for casting 1^st PDMS layer. c Real-time response of a CIPS to different sustained elevated pressure (rotation speeds for casting 1st PDMS layer is 3000 rpm, the same applies below). d The real-time response of PCPS to step-like differential pressure during the pressure loading and unloading process. e Response levels of a CIPS under bending and vertical conditions at different pressures. f Performance comparison of a CIPS at bending and vertical state. g A CIPS undergoing small and large pressure cycles. h Real-time response of a CIPS undergoing 90 pressure cycles (50–150 mmHg). In a, b, e, and f, data are presented as mean ± standard error of the mean

We also studied the annealing condition after transferring the graphene/PDMS layer on the PI substrate for better sensor performance and found that samples that underwent two-step annealing (50 °C for 30 min, 90 °C for 60 min) exhibited lower signal drift than samples that underwent one-step annealing (90 °C for 60 min) (Fig. S15). One possible explanation is that faster heating in one-step annealing may generate excessive air pathways during the water evaporation process, forming trapped air gaps between the sensitive layer and PI substrate, which would move randomly when external pressure is applied, changing the contact resistance between the graphene layer and PI substrate. Thus, we adopted two-step annealing to the samples in the following experiments.

Several studies on graphene-based pressure sensors report the existence of gas pathways between the graphene and substrate, leading to the decline of signal level as the pressure is maintained, which is a critical limitation hindering practical applications of such sensors^24,27,48. We tested the sensor’s real-time pressure response to different sustained pressure levels for about 10 s, and the CIPS’s response signals were stable, rather than decreasing (Fig. 5c), indicating that the 2^nd PDMS encapsulation layer blocked the gas pathways. Then we tested the CIPS’s detection limit, and this value reached 6.38 mmHg (Fig. S16). Although this limit can be further optimized by introducing a low-pass filter and a low-noise amplifier, it is sufficient for the sensor to detect minute fluctuations in blood pressure. Next, we tested the sensor using step-like differential pressure in the pressure loading and unloading process. The sensor’s output remained steady at each step, and it rose and declined in the pressure loading and unloading process (Fig. 5d). The response level exhibited no difference at the same pressure value during the pressure loading and unloading pressure, suggesting a low hysteresis for the sensor. Given the tortuous geometries inherent in human vasculature, the pressure responses of the sensors were evaluated under controlled flexural deformation conditions. Almost no difference in sensors’ performance could be observed between the straight and bending state from both individual (Fig. S17) and statistical results (Fig. 5e, f), demonstrating that the bending operation poses no influence on its performance and the CIPSs could work properly in the tortuous human vasculature system. Then, three small pressure cycles (0–280 mmHg) and three large pressure cycles (0–380 mmHg) were performed, and the sensor response was recorded simultaneously. It responded linearly to the small and large pressure, and the response level remained consistent in the small and large pressure cycles (Fig. 5g). Furthermore, we evaluated the cycling stability of the CIPS by conducting 90 pressure cycles in a pressure interval covering the human blood pressure range (50–150 mmHg) and simultaneously recording the sensor’s response. A high consistency in response level over these 90 cycles was observed, indicating the sensor’s excellent reusability (Fig. 5h). Besides, pressure tests to CIPSs with substrates’ OD of 300 μm (OD 300) and 1000 μm (OD 1000) were also conducted, and they all responded linearly to different applied pressures, illustrating the adaptability of the proposed process for fabricating CIPSs with different ODs for multiple purposes (Fig. S18).

Altogether, the fabricated CIPSs demonstrated high sensitivity and linearity, fast response speed, low detection limit, minimal hysteresis, excellent cycle stability, and outstanding fabricated adaptability on substrates with various ODs, while maintaining unchanged performance in bending states, proving its application potential in multiple scenarios.

In vitro and in vivo demonstration

To verify the detection capability of CIPS in liquid environments, an in vitro study was conducted by immersing the CIPS beneath the water-filled pressure tank. The pressure that was applied to the sensor (P_s) equals to the pressure brought by the water (P_w) plus the air pressure in the tank (P_a, the value of the commercial pressure sensor). The CIPS in water responded linearly to the incremental applied air pressure (Fig. 6a). The decrease in CIPS’s sensitivity (1.03 × 10⁻⁶ ± 4.93 × 10⁻⁸mmHg^-1, three times lower than working in air) and increase in detection limit (11 mmHg, Fig. S19) in water may contribute to the low compressibility of liquids^49,50. The real-time signals of CIPS in water under both unpressurized and pressurized cycling (60 - 148 mmHg) conditions were compared (Fig. 6b). CIPS exhibited a distinct signal response to the pressure cycling, demonstrating its capability for working in a liquid environment and validating the effectiveness of the encapsulation strategy.

Fig. 6. In vitro and in vivo demonstration of CIPS.

a Response level to different pressures and b real-time response signal to pressure cycles (60–140 mmHg) of CIPS in static water. c Schematic illustration and d optical image of the in vivo experiment setup with the rat model. e Global and local magnified real-time blood pressure signal from CIPS implanted in a rat’s abdominal aorta

In vivo validation on rat models was subsequently conducted by implanting CIPS into a rat’s abdominal aorta (Fig. 6c, d). Adult SD rats (6–8 weeks) with thick blood vessel and strong heartbeat were selected for sensor implantation. A section of the surgically exposed abdominal aorta was clamped by two artery clamps for implantation. A CIPS sensor was inserted into the abdominal aorta through an indwelling needle. Then the upper artery clamp was removed to allow the sensor to encounter the blood pulse. After finishing the surgical procedure, sensor signal acquisition commenced. The sensors exhibited distinct signal responses to blood pressure fluctuations, with response levels comparable to those observed in water within the blood pressure range (slight differences in signal level stemming from the fluid mechanics parameter variations between blood and deionized water) (Fig. 6e). Although in vitro and acute in vivo experiments have confirmed the sensor’s reliability during short-term interventions, we note that the hydrophobicity of the PDMS surface may pose a risk of protein adsorption in long-term implant applications. Future work will focus on introducing anticoagulant surface coatings (such as heparinization or hydrophilic coatings) to further enhance its blood compatibility for long-term clinical monitoring. The results from both in vitro and in vivo experiments demonstrate that CIPS can operate stably in liquid environments and enable continuous blood pressure monitoring in animals. This represents the first reported work on in vivo pressure sensing using a suspended graphene pressure sensor, laying a crucial technical foundation for the application of graphene pressure sensors in the field of bioelectronics.

Conclusion

In summary, we have developed a catheter-integrated pressure sensor that represents a significant advance in intravascular continuous blood pressure monitoring. By conformally fabricating SMF arrays on the catheter surface and sealing them with ultrasoft biocompatible PDMS layers, the CIPS achieves hermetic and waterproof packaging while keeping a compact architecture. The introduction of IDEs and rational structural design reduces the initial resistance of the sensors and optimizes the mechanical behavior, thereby significantly enhancing CIPS’s response speed (<0.36 s), sensitivity (3.5 × 10⁻⁶ mmHg⁻¹), detection range (6–380 mmHg), and other merits. The CIPS operates reliably in liquid environments and has been successfully validated in vivo through real-time continuous blood pressure monitoring in the rat abdominal aorta. This work establishes a promising strategy for seamlessly integrating high-performance pressure sensors into interventional catheter platforms, paving the way for highly integrated sensing catheters in next-generation interventional medicine and setting a solid technical foundation for future MEMS/NEMS graphene-based bioelectronics.

Methods

Simulation settings

Structural Mechanics Module in COMSOL Multiphysics was employed to investigate the mechanical properties of the single sensing unit. Given that the deflection of the graphene film far exceeds its thickness, the inevitable geometric nonlinearity of the graphene film must also be accounted for. The boundary where graphene contacts the substrate was defined as a fixed boundary. A boundary load directed toward the cavity was applied from above the 2^nd PDMS layer. The material parameters used in modeling are shown in Table S1.

Fabrication process of CIPS

The flexible PI catheters (Vasculex Type-S, Furukawa Electric, Japan) with different OD were adopted as the substrates. A layer of positive photoresist (S1813, Resemi, China) was dip-coated onto the catheters’ surface and baked at 90 °C for 30 min and then exposed by CPL, and the designed patterns were obtained after development (AZ400K). Unlike static projection lithography which suffers from edge distortion on curved surfaces, the CPL system employs a synchronized scanning mode. The rotation of the catheter is coupled with the linear movement of the mask, ensuring that the exposure area is continuously maintained at the substrate’s apex within the optical Depth of Focus (DOF ± 45 μm). This mechanism ensures uniform resolution across the entire circumference without pattern degradation.” Then a Cr/Au layer was sputtered (MSP-400) to the above substrate and IDEs and connecting electrodes were formed after the lift-off process in acetone. Subsequently, the cavity array was ablated by the femtosecond laser (FM-uvfm10a, Fermi Laser Technology Co., Ltd., China). On the other hand, PDMS (DOWSILTM 184) was diluted (1:0.8 in hexane) and degassed in a vacuum oven. Then it was spin-coated (3000–6000 rpm, 50 s) on the commercial copper-based graphene (6–8 layers, XFNANO, China) and was cured at 80 °C for 6 h to form the supporting layer with ultra-thin film and ultra-low Young’s modulus. The copper-based graphene with a PDMS supporting layer was placed in copper etching solution (dissolving 20 g of CuSO₄ in 100 ml HCl and 100 ml deionized (DI) water) for 30 min to release the graphene/PDMS film, The graphene/PDMS was transferred into deionized water (DI water) to remove residual etchant and soaked for 2 h to enable surface tension-driven relaxation of the film to eliminate micro-wrinkles. Concurrently, the PI catheter with IDEs and cavity arrays was sequentially ultrasonicated in ethanol and deionized water (DI water) to thoroughly remove ablation debris and potential recast material, and then plasma-treated after drying to clean the contact surface and enhance the adhesion strength between the transfer film and substrate. Next, the graphene/PDMS film was then picked up by the above substrate and it was successfully transferred to the outer surface of the PI catheter. Subsequently, the sample underwent a two-step annealing process in an oven: first at 50 °C (30 min) for water evaporation, followed by 90 °C (1 h) for stress relaxation and complete shape adaptation. Two enameled wires (diameter: 80 μm) were mounted on the surface of the connecting electrodes by penetrating them from the inner surface to the outer surface through the via holes beside the electrodes. The above sample was sequentially ultrasonicated in ethanol and DI water and subjected to ultraviolet ozone (UV-O₃) treatment (BZS250GF-TC, Hwotech, China) for 10 min after drying to enhance the hydrophilicity of the PDMS supporting layer. Finally, the sample was mounted on a special custom-made tool, and the 2^nd layer of diluted PDMS (1:0.8 in hexane) was spin-coated (6000 rpm, 50 s) on it and cured at 80 °C for 6 h as the second encapsulation layer.

Characterization of the in situ fabrication

Images of SEM and EDS were captured using a super-resolution field-emission scanning electron microscope/energy spectrometer (JSM-7800F, Jeol, Japan). Optical images and depth of the cavity were captured and measured using an ultra-depth-of-field three-dimensional microscope (VHX-7000, Keyence, Japan). The sample with a cross-sectional surface was cut by a triple ion beam cutter (EM TIC 3X, Leica, Germany).

Testing procedure for CIPS

The pressure sensor was placed in a stainless steel pressure tank, and an air pump outside the tank was utilized to increase the gas pressure in the tank; a commercial pressure sensor was also placed outside the tank to monitor the real-time pressure level inside the pressure tank, and a high-precision digital source meter (Keysight B2902A) was used for collecting the pressure sensor’s output signals; a data terminal collected the data from the commercial pressure sensor and the high-precision digital source meter at the same time.

Root-mean-square noise (RMS_noise) and signal-to-noise ratio (SNR) calculation

The noise level was characterized by RMS_noise value of the baseline fluctuations in the absence of pressure input. Specifically, the RMS noise was calculated as:

$${\mathit{RMS}}_{\mathit{noise}}=\sqrt{{\frac{1}{N}\sum _1^N\left(R_i-\stackrel{\circledR }{R}\right)}^2}$$ 6

where R_i is the instantaneous resistance at the ith sampling point, $\stackrel{\circledR }{R}$ is the mean baseline resistance, and N is the total number of sample points.

The signal-to-noise ratio (SNR) of the CIPS was quantified by comparing the amplitude of the pressure-induced output signal to the intrinsic electrical noise level of the device. The output signal amplitude (ΔR) was determined as the difference between the peak response under a defined pressure stimulus and the baseline level without loading. The SNR was calculated as:

$$\mathit{SNR}\left(\mathit{dB}\right)=20{\log }_{10}\left(\frac{\Delta R}{{\mathit{RMS}}_{\mathit{noise}}}\right)$$

Author contributions

Fengming Ye conceived the concept, performed formal analysis, investigation, validation, wrote the original draft. Jianwei Hou, Xuanyu Li, Jie Ni, Mengqiu Li, Jiaxiang Wang, Yangtao Yu, and Zhiyuan Hu. performed formal analysis and investigation. Guifu Ding, Bin Sun, Kaiming Su, Xiaojun Guo, Fuyou Liang, and Zhuoqing Yang provided resources. Yahui Li and Faheng Zang edited the draft. Zhuoqing Yang conceived the concept, supervised the project, reviewed, and edited the draft.

Conflict of interest

The authors declare no competing interests.

Ethics declarations

In vivo experiment. All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Shanghai Jiao Tong University (approval number: A2025195). Adult Sprague–Dawley (SD) rats were purchased from Shanghai Jihui Laboratory Animal Co., Ltd. All surgical procedures were performed under anesthesia, and the animals were euthanized by cervical dislocation after the experiments.

Supplementary information

The online version contains supplementary material available at 10.1038/s41378-026-01242-z.