Polydimethylsiloxane (PDMS) (Sylgard 184) was purchased from Dow Corning Co., Ltd. NdFeB particles (5μm) were commercially obtained from Magnequench Co., Ltd. CNTs were purchased from XFNano Co., Ltd. Ecoflex 0030 and mold release agent (Release 200) were purchased from Smooth On Co., Ltd. Novec 7000 and Novec 649 were commercially obtained from 3M Co., Ltd. Hexane and anhydrous ethanol were purchased from Hushi Co., Ltd. Medical-grade polyvinylidene fluoride (PVDF) heat-shrinkable tubing was purchased from Zhongxin Co., Ltd. The superelastic nitinol wire (nitinol core) was commercially obtained from Mingfeng Co., Ltd. The anticoagulated porcine whole blood was purchased from Yuechi Co., Ltd. All the chemicals were used without further purification.

The experimental design for this study aimed to evaluate the biological response of L-929 cells to the MBF. The test samples (magnetic tip and phase-change balloon) were subjected to an extract preparation process under sterile conditions. Specifically, the samples were immersed in 75.6 mL of MEM medium containing 10% fetal bovine serum at a ratio of 3 cm² of sample surface area to 1 mL of extractant. The immersion was conducted in a sterile, inert container at 37 °C for 48 h with oscillation. After immersion, the extract was checked for any changes and stored at 20-30 °C, ready for use within 24 h without further filtration, centrifugation, dilution, or pH adjustment. Simultaneously, blank, negative (high-density polyethylene), and positive (ZDEC) controls were prepared under the same conditions. The L-929 cells, sourced from the American Type Culture Collection (CCL1, NCTC clone 929), were cultured in MEM medium supplemented with 10% fetal bovine serum and antibiotics (penicillin and streptomycin) at 37 °C in a 5% CO₂ incubator. Following cell growth to confluence, the original medium was aspirated, and 100 μL of different concentrations of the test sample extract (100%, 75%, 50%, 25%), blank control, positive control (100%), and negative control (100%) were added to the cells. The cells were then incubated for an additional 24 h at 37 °C in a 5% CO₂ atmosphere.

The evaluation criteria for assessing cell viability and potential cytotoxicity were based on the optical density (OD) measurements at 570 nm using a microplate reader, with a reference wavelength of 650 nm. Cell viability was calculated as the percentage of the mean OD value of the test sample/negative control/positive control relative to the mean OD value of the blank control.

$$ \text { Cell viability }(\%)=\frac{100 \times O D_{570 e}}{O D_{570 b}} \\ $$

OD_570e - Average optical density of test sample/negative control/positive control; OD_570b - Average optical density of blank control.

A lower viability percentage indicated higher potential cytotoxicity of the test sample. Specifically, if the viability decreased to less than 70% of the blank control, the test sample was considered to have potential cytotoxicity.

Quantitative data are expressed as the mean ± standard deviation (SD). Unless otherwise stated, experimental data were obtained from three independent replicates (n = 3). For the branching-channel navigation tests, ten repeated trials were conducted for each representative path. For comparisons between two groups, statistical significance was evaluated using a two-tailed independent Student’s t-test. A P-value of less than 0.05 (P < 0.05) was considered statistically significant.

Incorporating the navigation of magnetic guidewires and the volumetric expansion of the phase-change balloon, we developed a MBF with a diameter of 1 mm. The MBF consists of three main components: a magnetic tip at the distal end (length 2.5 cm), a phase-change balloon (length 2 cm), and a non-magnetized fiber (Polydimethylsiloxane, PDMS; length 1.5 m) at the proximal end. These three components are connected using a single, continuous nitinol core (diameter: 0.15 mm), which acts as a backbone, with their interfaces secured by cured silicone [Figure 1B and Supplementary Figure 2]. Notably, the core consists of a superelastic nitinol core with a phase transition temperature of approximately -10 °C. This ensures the wire remains in a stable austenite phase across operational temperatures. Consequently, microwave actuation does not induce any undesired shape changes or bending via the shape memory effect, allowing the design to maintain its intended flexibility and configuration. The magnetic tip is fabricated by uniformly dispersing NdFeB microparticles in the PDMS matrix. This composite is then injected into a tubular mold and cured, and subsequently axially magnetized by an impulse magnetic field (~4 T), enabling the tip to bend when an external actuation field B is applied, thus facilitating navigation through tortuous vasculature.

The phase-change balloon is constructed by affixing a pre-stretched Ecoflex tube to the anchor points on the nitinol core. It is then injected and sealed with 15 μL Novec engineered fluid, which serves as the inflation medium. This liquid has an intrinsic liquid-to-gas phase transition temperature (boiling point) of 42 °C [Supplementary Figure 3]. Consequently, when the actual internal temperature of the fluid reaches this 42 °C threshold, it triggers the phase transition required for balloon expansion. This specific thermal property ensures that the balloon remains unexpanded at normal body temperature [Supplementary Figure 4]. For potential biomedical use, the balloon was operated within a safety-guided temperature window of 42-50 °C, where the lower bound corresponds to the onset of phase transition and the upper bound serves as a conservative control limit^[35-38]. To enable wireless heating, the balloon is uniformly coated with CNTs via a spraying technique [Supplementary Figure 5]. A scanning electron microscope (SEM) image of the cross-section is provided in Figure 1C. When exposed to microwave energy, the CNTs absorb the energy, triggering the liquid-to-gas phase transition that inflates the balloon [Figure 1D].

The inflation process was further visualized through optical and infrared imaging [Figure 1E and Supplementary Video 1]. In the absence of biological tissue attenuation, the MBF reaches a baseline temperature of approximately 80 °C when positioned 15 cm from a 50 W microwave irradiator (Baoxing, China). We note that this value was obtained under a barrier-free in-air condition and is reported only to characterize the maximum wireless heating capability of the device, rather than its intended operating temperature in biological applications. Under this condition, the balloon region remains the primary functional heating zone, while limited heating beyond the balloon may also occur, likely owing to microwave coupling and heat conduction associated with the embedded nitinol core. The microwave operates at 2.45 GHz (wavelength ~12.4 cm), a frequency widely used in microwave thermotherapy-related applications^[39,40].

The balloon’s superior wireless heating efficiency is attributed to the excellent heat absorption capability of CNTs and the electromagnetic resonance effect of the nitinol core (detailed in the next section). This efficiency is quantified by the temperature rise per heating power, ∆T / P, and plotted against the heating distance in Figure 1F. The data in Figure 1F were compared under broadly similar barrier-free, in-air conditions, thereby reducing variability introduced by media attenuation. Under these conditions, the MBF shows improved wireless heating performance relative to previous techniques utilizing RF heating^[41-46] and microwaves^{[33,34,47,48]} (see Comparison Supplementary Table 2). Note that, the data in Figure 1F compares the wireless heating performance under barrier-free, in-air conditions, ensuring a fair comparison across different approaches without interference from biological tissues. This comparison is intended as a representative reference, since ∆T / P may still show some dependence on factors such as ambient temperature.

CNTs exhibit excellent electrical conductivity and electromagnetic wave absorption properties, making them effective for microwave-induced wireless heating^[49-52]. When exposed to a microwave electromagnetic field, free electrons in CNTs oscillate rapidly, converting the absorbed energy into thermal energy through lattice collisions. To evaluate the impact of CNT content on the heating performance, we tested balloons coated with varying CNT volume fractions of 10%, 20%, and 30%. As shown in Figure 2A, higher CNT content leads to a more pronounced temperature increase, with the temperature rising by 15 °C within 30 s at a 15 cm heating distance and a 50 W power.

To further enhance heating efficiency, we introduced a nitinol core into the balloon. As seen in Figure 2B, under the same conditions, the balloon with the nitinol core achieved a significantly higher temperature (P < 0.05, compared to the balloon without the core), and the one with 30% CNT content underwent a temperature increase of ∆T = 55.8 °C. We then investigated the effect of nitinol core length on heating performance. The optimal temperature enhancement occurred when the length of the nitinol core was close to approximately half of the free-space microwave wavelength [~6 cm; Figure 2C] under the present configuration. This enhancement is attributed to the electromagnetic resonance effect that maximizes the core’s ability to absorb and convert microwave energy into thermal energy, remarkably amplifying the heating efficiency [Figure 2D]. Electromagnetic thermal simulations conducted in COMSOL Multiphysics further validate this effect [Figure 2E]. Introduction of the nitinol core dramatically enhances the temperature within the balloon, demonstrating the critical role of resonance in optimizing wireless heating performance. We note that this ~6 cm value was identified as an empirical optimum in the current experimental setup, and the corresponding behavior under tissue-loaded conditions should be further examined in more physiologically relevant environments.

When considering small-vessel intervention scenarios such as cerebrovascular applications, the microwaves need to penetrate biological tissues and the skull to heat the balloon. To approximate this tissue-penetration scenario, we examined the effectiveness of microwave heating on the MBF after passing through porcine tissues with bone. The porcine tissues used in this experiment had a total thickness of approximately 15 mm (including both bone and attached soft tissue). This fan-bone specimen was selected because it provided a relatively large and intact tissue-bone area, which facilitated stable placement and repeatable microwave-penetration testing. The porcine tissues, placed 7 cm away from the MBF, approximate controlled ex vivo tissue-barrier condition [Supplementary Figure 6]. As shown in Figure 2F, the balloon’s steady-state temperature reached 49.3 °C at a 15 cm distance and 50 W power after penetrating porcine tissue with bone. This temperature is sufficient to induce the liquid-to-gas phase transition for effective balloon expansion. Specifically, we have defined the operational temperature window of 42-50 °C as a safety-guided protocol. The lower bound of 42 °C is set to correspond with the initiation threshold of the phase-change liquid, while the upper bound of 50 °C is established as a conservative guideline to mitigate the risk of thermal damage to adjacent vascular and neural tissues. Similar heating effects can be achieved by reducing the heating distance or lowering the power output [Figure 2G]. As summarized in Figure 2H, by synergistically adjusting the microwave power and heating distance, the MBF can be regulated to operate within this predefined target heating temperature range under the tested condition.

To examine whether microwave heating is preferentially concentrated in the balloon, we compared the thermal response of the CNT-coated balloon with that of adjacent porcine tissue under identical irradiation conditions [Supplementary Figure 7A]. Considering that blood and ion-containing biological fluids may affect microwave absorption, we further repeated the comparison using porcine tissue immersed in fresh anticoagulated porcine whole blood [Supplementary Figure 7B]. In both tissue-only and blood-containing conditions, the dominant hot spot remained localized at the balloon region, while the surrounding tissue or blood-containing medium showed limited heating, supporting the preferential microwave-heating behavior of the CNT-coated balloon under the tested static ex vivo conditions.

We further evaluated the thermal influence of the heated balloon on contacting porcine tissue [Supplementary Figures 8 and 9]. Infrared imaging showed that, although the balloon itself reached the actuation temperature, the temperature rise in the contacting tissue remained much smaller and no obvious extensive overheating zone was observed [Supplementary Figure 8]. Tissue-level thermal influence was further assessed by comparing the macroscopic appearance and microscopic surface morphology of porcine tissue after different thermal treatments [Supplementary Figure 9]. The tissue contacted with the MBF during 3 min of microwave heating retained a normal reddish appearance and showed no obvious microscopic structural alteration compared with the room-temperature control. Together, these results suggest that the balloon temperature should not be directly interpreted as the temperature of adjacent tissue, and support limited local tissue-level thermal influence under the tested static ex vivo conditions.

Overall, these results show that the MBF can maintain effective balloon actuation after tissue penetration while limiting excessive heating of adjacent tissue under the tested conditions. A limited local temperature increase may still appear on the porcine tissue surface, likely due to partial microwave absorption by the heterogeneous tissue sample and local variations in field distribution. Nevertheless, tissue attenuation, together with active modulation of power and distance, helps confine the heating effect and supports operation within the target 42-50 °C actuation window.

For effective angioplasty, the output pressure to reopen the blocked artery mainly depends on the radial expansion, while axial expansion is ineffective and undesired. Actually, minimizing axial expansion confines the balloon’s volume, increasing its internal pressure during the liquid-to-gas phase change. To achieve this, we implemented an axial pre-stretching strategy to control the balloon’s expansion behavior. As illustrated in Figure 3A, the Ecoflex tube was stretched before being affixed to the anchor points, enabling desired radial expansion while suppressing axial elongation. Let L₀, L, and L₁ denote the length of the unstretched Ecoflex tube, stretched balloon, and inflated balloon, respectively, and the d₀, d, and d₁ denote the corresponding diameters (Note d = 1 mm for the MBF). The pre-stretch, radial expansion, and axial expansion are defined as L / L₀, d₁ / d, and L₁ / L, respectively.

To assess the impact of the pre-stretch, we investigated both radial and axial expansion in free space under various internal pressures (denoted as p_free) which were controlled by a pump. As shown in Figure 3B, increasing the pre-stretch increases radial expansion while simultaneously reducing axial expansion. This is because axial pre-stretching pre-consumes the axial elongation that would otherwise occur during inflation, thereby shifting the subsequent deformation mainly to the radial direction. In addition, the reduced initial diameter after pre-stretching further increases the apparent radial expansion ratio. Specifically, with a pre-stretch L / L₀ = 1.6, the radial expansion reaches up to d₁ / d = 300% at 20 kPa, whereas axial expansion remains nearly unchanged, i.e., L₁ / L = 1. Compared to the unstretched case, the radial expansion at 20 kPa increases by 1.6 times while the axial expansion is eliminated [Figure 3C]. Based on these findings, the pre-stretch of 1.6 is adopted for the MBF design to ensure optimal expansion performance.

Since it is difficult to directly measure the internal pressure of the balloon under microwave heating due to its small dimensions, we adopt an indirect method by correlating the radial expansion of the heated balloon with the pumped one in free space. It was found that the maximum radial expansion of the heated balloon also reaches 300% after 40 s of heating at 50 W power and a 15 cm distance [Figure 3D]. This result indicates that the maximum internal pressure of the heated balloon is 20 kPa according to Figure 3B (denoted as $$ p_{\text {free }}^{\max } \\ $$ = 20 kPa). Upon turning off the power, the balloon begins to deflate immediately and gradually returns toward its initial dimension. In the present static experimental setup, the final stage of recovery occurs on the order of min, reflecting passive cooling of the sealed phase-change medium. It is also worth noting the balloon’s mass remains nearly constant throughout the expansion and deflation cycle, indicating a well-sealed assembly with minimal leakage.

When the balloon is microwave-heated in small vessels, it first expands freely until it contacts the vessel wall, at which point the internal pressure equals p_free from Figure 3B. Further inflation will exert output pressure (denoted as p_out) on the vessel. The output pressure can be calculated as

(1) $$ p_{\text {out }}=p_{\text {confined }}-p_{\text {free }} \\ $$

where p_confined is the internal pressure of the balloon in its confined state. Assuming the temperature remains unchanged during inflating, the ideal gas law relates the confined balloon pressure and freely-expanded balloon at maximum radial expansion of 300% via

(2) $$ p_{\text {confined }} V_{\text {confined }}=p_{\text {free }}^{\max } V_{\text {free }}^{\max } \\ $$

where V_confined and $$ V_{\text {free }}^{\max } \\ $$ are the balloon volumes in the confined and freely expanded states, respectively. Considering the unchanged axial length during inflation, the volume ratio can be readily expressed as $$ V_{\text {free }}^{\max } / V_{\text {confined }}=\left(3 d / d_{\mathrm{v}}\right)^{2} $$ where coefficient 3 represents the maximum radial expansion of 300% and d_v is the vessel diameter. Using these relationships, the output pressure as a function of the normalized vessel diameter $$ d_{\mathrm{v}} / d $$ is calculated and presented in Figure 3E.

Results in Figure 3E clearly suggest that the output pressure increases substantially up to 180 kPa as the vessel’s diameter decreases. This highlights the MBF’s ability to generate sufficiently high pressure to dilate highly narrowed vessels. To further validate this ability, we fabricated a simulated vessel with a diameter of 2.5 mm using silicone rubber, with Young’s modulus of 250 kPa close to that of real blood vessels^[53,54] [Supplementary Figure 10]. The MBF was inserted into this simulated vessel and the phase-change balloon successfully expanded under microwave heating, generating sufficient pressure to widen the vessel [Figure 3F].

The magnetic deflection capability of the MBF plays a critical role in its navigation through complex vasculature. To evaluate its deflection performance, we measured the magnetic deflection angle (θ) under varying magnetic field strengths (B) and angles (α). As shown in Figure 4A, both experimental results and finite element simulations demonstrate a consistent increase in the deflection angle with increasing magnetic field strength. The deflection performance was further validated at a specific magnetic field with α = 90°, showing excellent agreement between experimental and simulated results [Figure 4B]. Importantly, the integration of the phase-change balloon does not compromise the MBF’s navigational performance. As shown in Figure 4C, the deflection angle remains nearly identical with and without the balloon, confirming that the integration of the balloon does not hinder its navigational capability. Nevertheless, navigation is inherently design-dependent. FE simulations [Supplementary Figure 11] reveal that longer balloons increase the magnetic moment arm, yielding larger deflections. We selected a 20-mm length to optimally balance steering leverage with spatial maneuverability in tortuous vessels. Additionally, the nitinol core’s stiffness matches commercial guidewires of the same size, ensuring essential pushability without excessive bending resistance that impedes magnetic deflection.

To assess the MBF’s ability to navigate through complex branches of small arteries, we conducted accessibility tests in a planar model with four 2-mm-wide channels (labeled as channels 1, 2, 3, 4). As illustrated in Figure 4D, the MBF successfully accessed all four channels under the guidance of an external magnetic field, demonstrating its ability to navigate complex and narrow pathways. To further quantify this navigation performance, we conducted 10 repeated trials for representative branching paths and summarized the operation time, success rate in Supplementary Figure 12. In stark contrast, commercial guidewires with a straight tip [Figure 4E] or a J-shaped tip [Figure 4F] exhibited limited accessibility. The straight tip guidewire was stuck at the entrance and unable to navigate through any channels, while the J-shaped tip guidewire showed partial accessibility to channels 1 and 3 but failed to navigate to channels 2 and 4. These results underscore the MBF’s superior navigation capabilities compared to conventional guidewires, particularly in environments with narrow and tortuous pathways, as detailed in Supplementary Video 2.

To comprehensively evaluate the magnetic navigation performance of the MBF, we conducted a demonstration using a 3D-printed cerebrovascular phantom that replicates the intricate human cerebral arteries [Supplementary Figure 13], including the Circle of Willis (CW), the right internal carotid artery (ICA), and the right middle cerebral artery (MCA). Guided by external magnetic fields, the MBF successfully navigated through complex and tortuous branches. The navigation sequence, shown in Figure 5A, highlights the MBF’s ability to access small and narrow arterial branches. At 00:09, the MBF passed the curved ICA with a diameter of 3 mm. At 00:22, it successfully entered CW. Remarkably, at 00:38, the MBF navigated through an extremely narrow segment with a diameter of 1.6 mm. After that, MBF reached the M2 segment of the MCA at 00:47. The MBF continued to navigate seamlessly into even smaller arteries of MCA thereafter. The time-sequenced images and Supplementary Video 3 highlight the MBF’s smooth and consistent movement, even in highly curved and confined segments of the phantom. This demonstration underscores the MBF’s excellent navigation capabilities in small and tortuous blood vessels effectively, offering a promising solution for angioplasty in small arteries.

Following the demonstration of the MBF’s magnetic navigation capabilities in a 3D cerebrovascular phantom, we further performed biocompatibility testing and an ex vivo angioplasty demonstration in a porcine placenta model to assess its feasibility in a biologically relevant vascular environment. Cytotoxicity tests using L-929 fibroblast cells exposed to the MBF’s 100% sample extract, as well as diluted concentrations (25%, 50%, and 75%) are first performed. As shown in Supplementary Figure 14, the cells maintained over 90% viability across all extract concentrations, with no obvious differences compared to the negative control group. In contrast, the positive control group exhibited substantial cell damage, confirming the cytotoxicity of harmful substances. These results demonstrate that the MBF possesses excellent biocompatibility, making it suitable for vascular systems.

To validate the MBF’s angioplasty capability in a biologically relevant environment, we utilized a porcine placenta model for ex vivo testing. The porcine placenta is well-recognized for its vascular anatomy, which closely resembles the small arteries of the human brain, making it an ideal model for such demonstrations^[55]. At 00:08, the MBF was steered by a repulsive magnetic field, achieving a 75° counterclockwise deflection to enter a 2 mm diameter vessel branch. After withdrawing from this branch at 00:31, the MBF was directed by an attractive magnetic field, producing a 15° clockwise deflection into a narrower branch with a diameter of 1.1 mm. Upon withdrawal from this branch at 00:38, the MBF was repulsed again, with a 15° counterclockwise deflection, successfully entering a 1.8 mm diameter vessel branch [Figure 5B and Supplementary Video 4]. Finally, the MBF’s phase-change balloon was activated via microwave heating to perform angioplasty within the porcine placenta’s vessels. As shown in Figure 5C and Supplementary Video 5, the balloon expanded radially and widened the vascular lumen, demonstrating the feasibility of microwave-triggered balloon expansion in small and tortuous vascular pathways.