Synchronization enhanced force microscopy for clinical tumor tissue mechanical characterization Authors: Jiahao Song, Yanlong Zheng, Xifeng Sun, Yutao Xu, Xianming Qin, Xueyong Wei Published: 2026 ABSTRACT AbstractThe mechanical properties of biological tissues and cells are novel marks reflecting their physiological and pathological states. Conventional techniques for mechanical measurements are often limited by bulkiness or ex vivo constraints. In this paper, we developed an all-silicon resonant MEMS force sensor. The sensor uses electromagnetic excitation to achieve stable operation in atmospheric conditions, and integrates a high-Q resonator synchronization enhanced technique. This design improves the resolution of the sensor and reduces detection noise without sensitivity attenuation. Surface Young’s modulus tests were conducted on human gastric tissues. We revealed the mechanical differences among normal, tumor and fibrotic tissues which offers mechanical indicators for distinguishing these morphologically similar tissues in clinical scenarios. Notably, the measurement process is gentle and non-destructive, ensuring the integrity of tissue samples. This sensor provides a high-precision, miniaturized tool for tissue mechanical characterization, which can be used for navigation in minimally invasive gastric cancer surgery. Subject terms: Electrical and electronic engineering, Engineering Introduction Biological tissues’ mechanical properties (e.g., stiffness, elasticity, viscosity) are key intrinsic indicators reflecting their physiological functions and pathological states1,2. The cytoskeletal remodeling at the cellular level, structural reorganization at the tissue level and disease progression often accompanied by significant changes in mechanical parameters3–7. This close correlation between mechanical characteristics and pathological status endows tissue biomechanics with broad application potential in clinical diagnosis and surgical guidance8–10. Several established techniques exist for measuring tissue mechanical properties, including micropipette aspiration and atomic force microscopy (AFM). Micropipette aspiration quantifies stiffness by measuring tissue deformation under negative pressure11, while AFM enables nanoscale mechanical mapping using a microcantilever probe12. However, these methods have limitations that hinder their clinical application. Their systems are bulky and operationally complex, requiring tissues to be excised from the body for ex vivo measurement. Additionally, ex vivo conditions alter the in-situ mechanical properties of tissues, meaning measurement results do not reflect the true intraoperative state13. Therefore, a miniaturized, high-precision sensor is necessary that can be integrated with surgical tools to achieve real-time in-situ mechanical detection of tissues. Microelectromechanical systems (MEMS) sensors provide small size and fast response, making them an effective tool for detecting micro forces. However, traditional MEMS force sensors have design flaws that limit their applicability in surgery. For instance, capacitive MEMS sensors require larger capacitor plate areas to improve resolution14,15, which conflicts with the miniaturization and integration requirements of in vivo medical machinery. Similarly, piezoresistive MEMS sensors, which detect force via resistance changes16–19. Their sensitivity depends on stress changes in the piezoresistive layer. which requires significant deformation of the sensitive structure to improve resolution, making it difficult to achieve high-precision and wide range measurements. In contrast, optical MEMS sensors rely on optical path transmission for displacement/force detection, but the narrow operating space of surgery often blocks or distorts these paths, reducing measurement stability20–22. Among MEMS force sensors, resonant devices hold inherent advantages of high resolution and compact structure. Their measurement principle, based on resonant frequency shifts induced by external forces, avoids the structural deformation constraints of piezoresistive sensors and the large-size requirement of capacitive sensors, making them promising for miniaturized in-situ detection. However, a fundamental challenge persists for resonant MEMS sensors operating in air, as air damping significantly reduces the quality factor (Q factor) of the resonator, resulting in compromised frequency stability and low measurement accuracy. Previous attempts to address this issue by using piezoelectric excitation23,24, bias-optimized piezoresistive readout25,26 and mechanically coupled dual-resonator platforms27. Although effective for specific damping issues, these methods are still limited by the intrinsic low Q in air and lack active regulation for frequency noise suppression. The mutual synchronization of dual resonators inspired us to leverage the synchronization effect for improving the frequency stability of low-Q sensors28. Although unidirectional injection synchronization has already been applied in devices like accelerometers29, the mutual synchronized force sensors have not been previously investigated. In this paper, we developed an all-silicon resonant MEMS force sensor with synchronization enhancement technology. The sensor uses electromagnetic excitation to operate stably in atmospheric conditions, eliminating the problem of silicon-based resonant sensors being unable to vibrate in atmospheric environments due to excessive damping. Additionally, to overcome the low-Q of resonant sensors in air, which is the primary issue causing reduced resolution. The MEMS sensor is coupled with a high-Q resonator in a vacuum chamber, a synchronization strategy that suppresses frequency noise to improve resolution by 2.5-fold without reducing sensitivity, enabling the measurement capability of soft tissues or cells. To validate its clinical utility, we characterized the mechanical properties of human gastric tissues. The measurement process is gentle and non-destructive. This work not only provides a miniaturized mechanical sensor but also offers a new technical approach to enhance the precision of improving basic research, such as cell mechanics. Results and discussion Setup and working principal overview The experimental setup for the biomechanical measurement system is depicted in Fig. 1a, where the resonant force sensor is fixed on a three-axis nano-positioning stage to ensure high-precision position control. The cancer tissue sample is placed on a manual three-axis positioning stage. The overall experimental arrangement is illustrated in Fig. S1. This dual-stage configuration is designed to achieve precise alignment between the sample surface and the sensor input end. Setup and working principle of force microscopy system. a A photograph of the force microscopy system setup, consist of a MEMS force sensor installed on a 3D nano positioner and a manually controlled 3D specimen position. b A scheme for the overall package of the force sensor. c SEM image of the force sensor. d Schematic diagram of testing circuit of sensors. e Amplitude-frequency response at a diving voltage of 1.25 V. f The sensor natural frequency offset and amplitude enhancement caused by increased driving voltage The sensor’s effective operation in atmospheric environments relies on its specialized packaging design, as illustrated in Fig. 1b. Fabricated via MEMS processing technology (see Fig. S2), the sensor chip is adhesively bonded to a test printed circuit board for signal transmission. Permanent magnets are mounted above and below the force sensor chip to provide a uniform magnetic field, which is essential for the electromagnetic excitation of the resonator. The aluminum housing is fully grounded to achieve effective electromagnetic shielding, reducing external electromagnetic interference. The detailed structure of the sensor is characterized by scanning electron microscopy (SEM), with Fig. 1c displaying its overall layout. The sensor integrates a sample input end, a probe support structure, a mechanical amplification structure (micro-lever) and a sensitive resonator. As the core sensitive unit of the sensor, the resonator features an H-shaped beam structure, which is shown in Fig. 1d. The left beam serves as the driving beam. The driving voltage is generated by NCO (2Vdcos(Ωt)) and converted into driving signals (Vdcos(Ωt) and -Vdcos(Ωt)) through a single ended to differential amplifier. The driving signals are applied at the driving beam upper and lower ends. Owing to the piezoresistive effect of silicon, an alternating current is generated in the driving beam. In the uniform magnetic field, the driving beam produces an alternating electromagnetic force, triggering vibration of the driving beam. This vibration is transmitted through the middle connecting beam to drive the right beam which serves as sensitive beam into synchronous vibration. When the sensor is subjected to an external load, the sensitive beam undergoes axial stress variation. This stress variation changes the stiffness of the sensitive beam, which in turn alters the natural frequency of the resonator. The working principle and finite element analysis are shown in Fig. S3. The sensitive beam is designed with detection plates. The vibration causes a change in capacitance between the plates and the detection electrodes, generating a detection current. This current is converted to a voltage via a differential amplifier and read out by a lock-in amplifier (HF2LI). Figure 1e presents the open-loop amplitude-frequency curve of the resonator measured by the lock-in amplifier. Under a driving voltage Vd of 1.25 V, the resonator exhibits a natural frequency of 59.5 kHz, an amplitude of approximately 0.016 V and a quality factor of 21. Notably, the resonator’s amplitude and frequency can be adjusted via the driving voltage. An increase in driving voltage raises the current in the driving beam, which in turn enhances the driving force. This enhanced driving force ultimately leads to a linear increase in the resonator’s amplitude. In contrast, the effect of driving voltage on frequency stems from the Joule heating effect30, which heats the resonator to reduce the Young’s modulus of silicon material and thereby leads to a gradual decrease in the resonator’s natural frequency as the driving voltage increases. As shown in Fig. 1f, the natural frequency of the sensor is adjusted by 1 kHz with the driving voltage increases from 0.5 V to 2 V. Construction of synchronized force sensor by the electrical signals Synchronization has been proven to significantly reduce frequency noise in vibration systems and has been successfully applied in resonant accelerometers28,31,32. To address the frequency output fluctuation issue of the sensor (with a Q-factor of 21) operating in atmospheric environments, we propose a mutual synchronization technology to reduce the sensor’s frequency noise by coupling with a high-Q resonator. Figure 2a shows the schematic of the synchronization system constructed by coupling the sensor with a resonator inside a vacuum chamber. The vacuum-chamber resonator has a Q-value of approximately 1800 and adopts piezoresistive detection and electrostatic driving. Its electrostatic driving signal is twice the natural frequency, and the detection signal is demodulated at the natural frequency to form a parametric excitation oscillator. The sensor is built into a closed-loop oscillation at its natural frequency. Synchronization relies on electrical signal coupling rather than mechanical connection. The sensor’s oscillation signal is first regulated to 60 mV by an amplitude controller (labeled A in Fig. 2a), then injected into the parametric excitation oscillator via an adder. Similarly, the parametrically excited oscillator’s driving signal is adjusted to an amplitude of 800 mV by an amplitude controller prior to its injection into the sensor. And the phase is tuned by a phase controller (labeled ϕ in Fig. 2a) to match the sensor’s operating phase point, before being injected to the sensor via an adder. Implementation of synchronized force sensor. a A schematic diagram of synchronous state between the sensor and a parametrically excited oscillator. b The measured synchronization bandwidth of the force sensor, the perturbation frequency amplitude is 800 mV. c The measured synchronization bandwidth of the parametrically excited oscillator, with perturbation frequency amplitude of 60 mV. d Time-domain response of synchronization process by regulating the frequency of parametric oscillator using Joule heating effect To realize stable synchronization, the synchronization bandwidth of the sensor and the parametric excited oscillator must first be determined. A perturbation signal generated by a lock-in amplifier and injected into the force sensor, frequency sweeping (up and down) was then performed to determine the synchronization bandwidth. As shown in Fig. 2b, the sensor exhibits a synchronization bandwidth of approximately 9.5 Hz under perturbation frequencies with a second-harmonic amplitude of 800 mV. The parametric excitation can generate an ultra-broad synchronization bandwidth, which was verified in our previous work33. Under the injection of a perturbation signal of 60 mV, a bandwidth of 8972.2 Hz is probed here as depicted in Fig. 2c. Subsequently, the Joule heating effect is used to reduce the detection voltage of the parametric excitation oscillator, adjusting the natural frequency of the oscillator to approach the oscillation frequency of the sensor. Figure 2d shows the time-domain response of the synchronization process. When the frequency signals of the two systems gradually approach each other, beat frequency occurs due to signal coupling. The closer the frequencies, the larger the frequency fluctuation caused by the beat frequency. When the frequency difference between the oscillator and the sensor is smaller than the sensor’s synchronization bandwidth, synchronization is achieved, and beat frequency disappears. After synchronization, the frequency change of the sensor drives the frequency change of the parametric excitation oscillator. When an external load induces a frequency shift in the sensor, the oscillator’s frequency tracks that of the sensor. This will invariably result in the oscillator frequency falling within the sensor’s synchronization bandwidth. Following the establishment of initial synchronization, the sole condition for system desynchronization is that the sensor’s frequency deviates beyond the synchronization bandwidth of the parametric excitation oscillator. This method achieves synchronization via electrical signals, eliminating environmental constraints and allowing the sensor to operate independently in atmospheric conditions. Meanwhile, although the sensor working in atmospheric conditions operates in a linear regime with a relatively narrow synchronization bandwidth, this synchronization scheme shifts the limitation of synchronization bandwidth to a high-Q resonator operating in a vacuum, which can be easily driven into a nonlinear regime. Thereby allowing the synchronization bandwidth of the entire system to be readily tuned through the modulation of effective nonlinear stiffness34. Calibration of the synchronization enhanced force sensor After completing the sensor synchronization construction, it is necessary to verify its mechanical measurement performance through systematic calibration. Figure 3a shows a photo of the sensor’s displacement calibration. During the calibration, the sensor is brought close to the surface of a silicon wafer, and displacement is applied in a stepwise manner to simulate the contact interaction between the sensor and the sample. Figure 3b presents the full-process frequency output of the sensor during contact and separation with the silicon surface under 20 nm step loading. The sensor maintains good consistency between the loading and unloading processes. The frequency during unloading exceeds the sensor’s natural frequency, which indicates the adhesion between the sensor and the silicon surface. As displacement unloading continues, the sensor separates from the silicon surface, and its frequency returns to the natural frequency. Characterization of the force sensor. a A photograph of contact between sensor input end and silicon wafer surface. b Real time output of sensor with the contact and separation process between the sensor and the silicon wafer. c Cyclic displacement loading and unloading testing of the force sensor. d Sensor displacement sensitivity obtained from 10 cycles of testing e Force sensitivity calibration. The force is referenced by electrostatic force. f Force sensitivity with synchronization and unsynchronization, compensated by displacement sensitivity. g Measurement range, continuously loaded in 20 nm steps Figure 3c displays the real-time output of the sensor during cyclic loading and unloading. Fitting analysis was performed on the displacement and output frequency. In 10 cycles of testing, the displacement sensitivity during loading and unloading remained consistent, both being 1.4862 ± 0.0005 Hz/nm. The specific fitting data results can be seen in Fig. 3d. To further verify the sensor’s force measurement performance and the effect of synchronization, force sensitivity calibration was conducted using electrostatic loading35,36. Figure 3e shows the real-time output of the sensor during force sensitivity calibration, in which DC voltage is applied to the electrostatic force input electrode to generate electrostatic loads. During the test, sensitivity data were collected in both synchronized and unsynchronized states. Figure 3f presents the sensor’s force sensitivity fitting results in the two states, with detailed force calculation and force sensitivity calculation process available in Supplementary information S5. The calibration results show that the sensor’s force sensitivity is 0.1431 ± 0.0007 Hz/nN, which confirms that the synchronization scheme in Fig. 2a does not cause attenuation of the sensor’s sensitivity and ensures the accuracy of force measurement. The sensor’s measurement range was calibrated through displacement loading, and the results are shown in Fig. 3g. Under the condition of 0.9997 linearity, the sensor’s maximum measurable force is 40.7 μN, and the maximum measurable displacement is approximately 3900 nm. The enhancement of frequency stability by synchronization is mainly reflected in the improvement of sensor resolution. Figure 4a presents the comparison of the sensor’s output fluctuation before and after synchronization, and it can be intuitively observed that the output fluctuation is significantly reduced after synchronization. Figure 4b quantifies this improvement by showing that the root mean square (RMS) of the sensor’s output decreases from 7.4 nN to 2.9 nN after synchronization, corresponding to an approximately 2.5-fold enhancement. Figure 4c shows the Allan variance of the sensor output, indicating that synchronization exerts a significant suppression effect on the sensor’s short-term output noise. At the integration time of 0.1 s, the Allan variance decreases from 1.472 nN to 0.743 nN. When the integration time exceeds 1 s, the output noise is mainly dominated by drift from environmental factors such as temperature, suggesting that the synchronization scheme has no significant effect on long-term frequency drift. The measured output frequency noise of the sensor is displayed in Fig. 4d. At 1 Hz, the sensor’s output noise decreases from 646 pN/Hz1/2 to 150 pN/Hz1/2 after synchronization. At 10 Hz, the output noise further decreases from 439 pN/Hz1/2 to 40 pN/Hz1/2. These results demonstrate that the proposed synchronization strategy effectively suppresses the output noise of low-Q resonators. Importantly, the synchronization strategy significantly improves sensor resolution without compromising core performance such as sensitivity. This strategy is not limited to structural design and application scenarios, but provides a universal reference solution for optimizing the performance of low-Q resonant sensors working in air/water. A detailed summary of the sensor’s performance indicators can be found in Table S1 of the supplementary materials. Comparison of resolution between synchronized and unsynchronized sensor. a Real time frequency fluctuation of sensor output, with a sampling rate of 1800sa/s. b. Statistical distribution map of sensor output within five minutes. c Allan variance of sensor output between 0.1 s and 100 s with 10 Hz target bandwidth of lock-in amplifier. d Noise density of sensor output Tumor tissue mechanical characterization Biological tissue’s mechanical properties are key markers reflecting its physiological and pathological states, and diseases such as cancer are often accompanied by significant changes in mechanical parameters like tissue stiffness and elasticity. This provides an important non-invasive basis for disease diagnosis and intraoperative tissue identification. After completing the sensor calibration and confirming its stable performance, we measured the surface Young’s modulus of cancerous tissues of human stomach. Figure 5a shows the sequential process of tumor tissue measurement. The first image depicts coarse alignment via the manual translation stage. The tumor tissue was positioned beneath the MEMS sensor under microscope visualization, ensuring the sensor’s sensitive axis aligned with the target region on the tissue surface. The second image illustrates the tissue being gradually moved closer to the sensor using the manual stage to avoid abrupt contact. The third image captures initial contact between the tissue and sensor. During the test, we collected data from sensors that showed stable step outputs to avoid the influence of liquid on the tissue surface, see details in Fig. S6. Measurement of mechanical properties of tumor tissue. a Measurement of tumor tissue processes. b A photograph of test human stomach normal tissue, tumor tissue, and fibrotic tissue samples. c. Real time sensor output when using 2-micron step testing. d Cyclic displacement loading and unloading testing of the force sensor. e Young’s modulus of the tissue surface measured at 4 × 4 points on the sample, seven independent experiments are conducted at each measurement location. f Normal gastric mucosa reveals the typical architecture of gastric pits (indicated by a black arrow) and individual gastric glands (red arrow), with uniform epithelial cells maintaining clear polarity. g Fibrotic tissue adjacent to a cancerous nest demonstrates dense, collagen-rich stromal bands (deep blue arrow) surrounding malignant cells, indicative of a desmoplastic reaction. h Tumor tissue exhibits marked architectural distortion, loss of glandular organization, and cellular pleomorphism. A representative tumor cell (yellow arrow) shows nuclear hyperchromasia, irregular contour, and an increased nuclear-to-cytoplasmic ratio. Images in (f–h) were obtained from formalin-fixed, paraffin-embedded tissue sections stained with hematoxylin and eosin Figure 5b shows a photograph of the tissues used for measurement, including normal human gastric tissue, gastric tumor tissue, and gastric fibrotic tissue. To ensure the representativeness of measurement results, a 4×4 grid of measurement points was set on the normal tissue area. Figure 5c displays the real-time output of the sensor during the measurement process, with curves corresponding to normal tissue, tumor tissue, and fibrotic tissue respectively. The measurement was conducted using a 2 μm step displacement loading method to simulate the gentle contact between the sensor and the fragile tissue surface. Figure 5d presents the force-displacement curves measured at points AE, BE, CE, and DE. From the results, it can be observed that the stiffness of fibrotic tissue is significantly higher than that of normal tissue, approximately 4–5 times that of normal tissue. In contrast, the stiffness of tumor tissue shows a softening trend compared with normal tissue, approximately 1/3 that of normal tissue. Since the measured force-displacement curves are affected by factors such as the contact area between the sensor and the tissue, and the thickness of the tissue, the Hertz contact model was adopted to characterize the surface Young’s modulus of the tissues37. The specific calculation process of Young’s modulus is available in Supplementary information S7. To further visualize the spatial distribution of tissue mechanical properties and ensure measurement accuracy, each measurement point shown in Fig. 5b was tested, and 3D images of mechanical properties were generated as show in Fig. 5e. Seven independent experiments were conducted for each measurement point to ensure the reliability of the data. The results show that fibrotic tissue, which forms in the late stage of tumor progression, undergoes significant hardening, with a Young’s modulus of approximately 3.98 kPa; the Young’s modulus of normal tissue is approximately 1.32 kPa, and that of tumor tissue is approximately 0.5 kPa. Notably, the surface stiffness of normal tissue shows good uniformity in the measurement area, with little difference in Young’s modulus between different points. In contrast, fibrotic tissue and tumor tissue exhibit large differences in Young’s modulus within the measurement area. We compared the histopathological characteristics of normal, fibrotic, and cancerous tumor tissues in specimens stained with hematoxylin and eosin (40 times magnification). The difference in Young’s modulus of tissues may be caused by uneven growth of cancer tissue. As shown in Fig. 5g, fibrotic tissue shows marked hardening due to excessive collagen fiber deposition which is a typical fibrotic pathological change, and uneven collagen distribution causes local density differences and mechanical heterogeneity. Tumor tissue’s softening trend stems from disordered tumor cell proliferation, as shown in Fig. 5h. Compared to normal cells, tumor cells have lower cytoskeletal density and looser intercellular connections, reducing overall tissue stiffness. And its highly heterogeneous internal structure, featuring areas of cancerous cell proliferation, necrotic zones, and residual normal tissue, further leads to significant spatial variations in hardness. In contrast, normal tissue has a relatively regular and uniform organizational structure, see Fig. 5f, therefore its Young’s modulus show good consistency throughout the entire measurement area. The mechanical characterization results not only confirm the sensor’s ability to distinguish different tissue types, but also link tissue mechanical properties to clinical needs. The synchronization-enhanced force resolution improves the detection of softer materials and the measurement of mechanical heterogeneity in early-stage tumors. For instance, as shown in Fig. 5e, the Young’s modulus varies by as little as ~3 Pa between some measurement points in tumor tissue. This higher resolution is crucial for accurately resolving subtle force differences in such heterogeneous or softer pathological tissues38. Conclusion In summary, we designed a resonant MEMS force sensor that can be used for characterizing the mechanical properties of biological tissues. The electromagnetic excitation design enables all-silicon sensors to operate stably in atmospheric conditions, avoiding the vacuum dependency of conventional resonant sensors. The cross-environment electrical coupling synchronization strategy provides a universal precision enhancement solution for low-Q resonant MEMS force sensing by decoupling environmental adaptability. Experimental validation confirms the sensor’s gentle, non-destructive measurement mode and reliable data output. The sensor can stably distinguish the Young’s modulus of normal, tumor, and fibrotic gastric tissues, which verifies its practical applicability for intraoperative tissue mechanical characterization. Benefiting from its compact footprint and atmospheric stability operation capability, the sensor is compatible with the working channel of clinical laparoscopic instruments (typical inner diameter ≥5 mm). Despite these achievements, the current system still has room for further optimization. In terms of integration, the sensor is currently in preliminary adaptation with laparoscopic instruments, and future work can focus on optimizing miniaturized packaging. This includes optimizing the sensor’s flexible lead arrangement and on-chip excitation, integrating the high-Q parametric oscillator into the detection circuit via vacuum-packaged headers to simplify the overall system while retaining noise suppression advantages, and adopting custom ASICs or miniaturized oscillation circuits (e.g., self-excited oscillation circuit39 or digital phase-locked loop circuit40) to reduce the readout system’s size for better clinical adaptability. Moreover, the exploration of Q-value optimization41 of the resonator for synchronization can be investigated to further decrease the detection noise floor. Furthermore, by achieving force noise floor in the piconewton level through synchronization, the sensor opens up possibilities for fundamental biomedical research42, such as the quantitative assessment of cell adhesion forces and the measurement of tension within individual protein filaments. Materials and methods Preparation of tissues In the present work, a total of 9 tissue samples were collected, including 3 cases of gastric tumor tissue, 3 cases of normal gastric tissue, and 3 cases of gastric fibrotic tissue. After resection by surgeons, each sample was immediately sent to the Department of Pathology without any pretreatment. A portion of tissue with a thickness of approximately 2 mm was excised from each sample, and then placed on glass slides in an antibiotic solution environment for subsequent biomechanical characterization. Experimental setup The experimental system was constructed for precise, stable mechanical measurement of gastric tissues (see Fig. S1). The displacement loading unit was a piezoelectric nano-positioning stage (Thorlabs MAX 312D, positional accuracy: 10 nm, travel range: 20 μm). A manual positioning stage was used for coarse positioning (positional accuracy: 2 μm, travel range: 20 mm). The piezoelectric stage was mounted on a manual rotation stage to align the MEMS force sensor’s sensitive axis with the normal vector of the sample stage. All system components were fixed on a rigid optical platform to suppress external vibrations. Temperature stability is critical for resonant MEMS force sensors. Benefiting from the all-silicon design, our sensor exhibits a favorable temperature coefficient (see Fig. S4). However, to ensure measurement accuracy for ultra-low force detection, the entire experiment was conducted in a constant temperature environment (25 ± 1 °C). This setup ensures thermal equilibrium of the sensor, avoiding temperature-induced frequency drift. Additionally, a CCD camera (GP 300 C) was integrated to provide real-time visualization of the sensor-sample contact interface. The sensor and the parametric oscillator were powered by a precision DC power supply (Keysight E3631A), and its open-loop electromagnetic excitation and closed-loop resonant frequency shift detection were both implemented via an HF2LI lock-in amplifier. The parametric oscillator is placed in a vacuum chamber with a vacuum degree of less than 0.1 Pa. Calculation of Young’s modulus Young’s modulus of gastric tissues was derived using Hertzian contact theory, which describes the elastic deformation of a soft material under indentation by a rigid indenter. Considering the tissue thickness was significantly larger than the indentation depth, the gastric tissue was treated as an isotropic elastic half-space. The sensor indenter was considered rigid. The Young’s modulus (E) was calculated using the formula: Where v is the Poisson’s ratio of the tissue which taken as 0.5, P is the contact force measured by the MEMS sensor and h is the indentation depth. The detailed calculation process can be found in Supplementary Information S7. Data processing Raw data from the HF2LI lock-in amplifier (Sample rate 1800Sa/s, Phase lock loop bandwidth 10 Hz). The Allan variance σA was calculated as where τ is the integration time, N is number of segments, fi® is the average frequency in ith segment and f0 is the natural frequency of the sensor. 1-σ standard deviation was computed to quantify short-term fluctuations and it can be obtained by where xi is the ith individual measurement value, μ is the mean of the n measurements, and n is the sample number. Power spectral density (PSD) analysis characterized noise distribution across frequencies. The frequency dataset was processed with a Hanning window (overlap: 50%) and converted to PSD using fast Fourier transform (FFT). Stepwise data extraction was applied to handle the stepwise pattern of measurement signals. For each step, the stable output segment was extracted, and its arithmetic mean was calculated as the representative sensor output for that load condition. Linear fitting was performed using Origin software. Least-squares linear regression was adopted, and the goodness of fit was evaluated using the coefficient of determination (R²). Supplementary information supplement document