An ultra-compact circularly polarized implantable antenna with 3D head validation for wireless deep brain stimulation (DBS) applications is presented in this work. Operating in the 2.45 GHz ISM band, the antenna features a miniaturized meandered structure and high-permittivity substrate to achieve an ultra-small size of 4.1× 5.2 × 0.5 mm3, equivalent to $0.125{{{\rm{\lambda }}}_{\mathrm{g}}} \times 0.159{{{\rm{\lambda }}}_{\mathrm{g}}} \times 0.015{{{\rm{\lambda }}}_{\mathrm{g}}}$ where ${{{\rm{\lambda }}}_{\mathrm{g}}}$ is guided wavelength. Circular polarization is realized using an H-shaped slot, improving communication reliability in lossy and multipath-prone biological environments. The antenna is designed and analysed using multilayer human tissue models and a realistic 3D head phantom in Ansys Electronics Desktop 2023 R1. Experimental validation is carried out through implantation in pork tissue and a fabricated 3D head model. The measured results show a wide impedance bandwidth of up to 550 MHz and a peak gain of −29.14 dBi. SAR analysis confirms safe operation within IEEE limits at typical implantable device power levels. A comprehensive link budget analysis demonstrates the proposed system supports reliable communication up to 2.2 m at a data rate of 7 Mbps with sufficient link margin and validates DBS telemetry. The antenna exhibits directional radiation with a high front-to-back ratio, making it suitable for body-centric wireless applications. The proposed design offers a promising solution for compact and efficient implantable communication systems.
Objectives: This study evaluates human exposure to radiofrequency electromagnetic fields in an office environment equipped with Reconfigurable Intelligent Surfaces (RIS), a key technology for 6G networks. Method: Using a two-steps methodology combining raytracing for electric field estimation and Specific Absorption Rate (SAR) calculations, we assessed whole-body (wbSAR) and brain (brainSAR) exposure levels in female and male models under multiple beam configurations at 3.5 GHz. Two scenarios were investigated: RIS beam is intended to be i) directed toward workstations and ii) toward a hypothetical moving user. The analysis included spatial distribution mapping of exposure levels, peak SAR value assessment, and statistical analysis of anthropometric and posture-related differences across various beam configurations. Results: Results reveal that brainSAR peak values in most cases exceeds wbSAR peak values, with exposure levels being highly dependent on beam configuration, subject posture, and its anatomical characteristics. The present study revealed that, in the simulated indoor scenario, the anatomy-related and posture-related differences in terms of exposure levels vary within the environment depending on the relative position of the human body and the RIS. Notably, non-users may experience higher exposure than intended users due to main beam interception at different heights. However, all the exposure levels found are well below the limits recommended by the ICNIRP guidelines. Clinical or Biological Impact: The exposure assessment demonstrates that EMF absorption from RIS technology varies based on individual anatomy and posture, with important implications for designing indoor 6G environments depending on diverse user profiles.
Metasurfaces have emerged as wireless resonators to locally enhance MRI receive sensitivity, yet most designs rely on modes with their magnetic fields oriented mainly in surface-normal direction. These become ineffective when placed in the transversal plane in closed-bore magnetic resonance imaging (MRI) systems. Here, we present a butterfly-based wireless metasurface that exploits two orthogonal tangentially oriented eigenmodes to enable receive enhancement in this otherwise unfavorable configuration, thereby boosting the sensitivity of standard scanner receive arrays. In phantom measurements at 3 T, the metasurface yields signal-to-noise ratio (SNR) enhancements of up to 14.4-fold with an effective penetration depth of 73 mm. The tangential operating principle focuses the enhancement close to the metasurface and therefore provides high local SNR with a reduced penetration depth compared to surface-normal mode metasurfaces. An in-vivo validation in the ankle demonstrates feasibility and shows improved central and dorsal signal coverage compared to the spine coils alone. Overall, the proposed tangentially enhancing metasurface class constitutes a complementary and cost-effective resonator concept for targeted SNR enhancement in placements where conventional coils are unavailable and previously proposed metasurfaces are not practical.
Designing antennas for implantable devices remains challenging due to the trade-off between miniaturization, bandwidth, and safety compliance in biological tissues. This paper presents a compact wideband implantable antenna designed for head-implantable devices. The miniaturization is achieved through shorting via loading, meandered geometry, and ground-slot perturbations that elongate the current path, while the loading by the high-permittivity tissue environment further reduces the effective wavelength. The proposed antenna employs a meander-line patch and a slotted ground plane fed by a 50-$\Omega$ coaxial probe, and it is implemented on a Rogers Duroid RT5880 substrate with a thickness of 0.254 mm. With an overall size of 4.5 × 4.5 × 0.254 mm$^{3}$ (5.14 mm$^{3}$ volume), the antenna achieves circular polarization and an impedance bandwidth of 0.703–1.738 GHz, corresponding to an absolute bandwidth of 1.035 GHz and fractional bandwidth of 84.8%. The design exhibits a peak gain of −37.6 dBi at the resonance frequency. To validate the design, the antenna is fabricated and experimentally tested, with measured results showing good agreement with simulations. In addition, extensive parametric studies are conducted to optimize performance. The simulated Specific Absorption Rate (SAR) for 1 g of tissue is 80.3 W/kg for a reference input power of 0.5 W. When normalized to the intended operating power of −16 dBm, the SAR is reduced to approximately 0.004 W/kg, which is significantly below the FCC safety limit.Link-budget analysis demonstrates reliable communication up to 6.8 m, 3.9 m, and 3.1 m for 1 Mbps, 10 Mbps, and 25 Mbps data rates, respectively. The achieved miniaturization, circular polarization, and validated performance make the proposed antenna a good candidate for next-generation biomedical implant applications.
We present a planar microstrip structure in a stamp-like configuration that enables stable two-port measurements without requiring cutting or reshaping of the material under test (MUT). In addition, a machine-learning (ML)- and deep-learning (DL)-assisted framework is proposed for broadband permittivity retrieval of biological tissues, leveraging the full-spectrum electromagnetic (EM) response. A diverse dataset comprising 321 frequency points over the 1–17 GHz band was generated for lossy biological materials using a full-wave EM simulator, which closely replicates real-world measurement response and eliminates the need for physical measurements when creating the ML/DL dataset. Four supervised regression models were systematically developed with optimized hyperparameters, while robustness was enhanced through air-gap augmentation to emulate practical measurement conditions. Among the models, the random forest (RF) model demonstrated superior accuracy and stability, achieving an average error of 1.38 ± 1.02% on the unseen synthetic test dataset. Experimental validation on porcine tissue samples confirmed the effectiveness of the proposed approach, yielding mean ± standard deviation relative errors of 3.62 ± 2.24% for skin, 3.09 ± 1.70% for muscle, and 10.34 ± 4.33% for fat. The proposed framework provides a practical, scalable, and cost-effective solution for real-time retrieval of frequency-dependent complex permittivity, moving beyond single-frequency analysis and offering strong potential for biomedical applications.
Radar-based motion estimation for non-contact vital sign sensing relies on phase modulation of the backscattered signal, causing in-phase and quadrature (I/Q) samples to trace a circle in the complex plane. Noise and hardware imperfections may cause a deviation of the circle center from the origin. Typical compensation approaches calibrate such direct current (DC) offsets by fitting a circle to the I/Q data and use it for scaling to a unit circle representation for phase stability. However, when the target motion is small relative to the radar signal’s wavelength, circle fitting becomes a highly ill-posed problem, which is typically not accounted for in existing approaches. Therefore, we propose NesMom, a novel lightweight, momentum-guided circle-fit correction that adaptively smooths consecutive circle estimates over time. The temporal smoothing and gradual fit updating allows our method to stabilize amplitude variation over time, resulting in higher-quality displacement extraction. The proposed method is evaluated against nine benchmarking approaches on a 55-subject dataset acquired with a 60 GHz radar measuring micrometer-scale skin expansion for blood pressure monitoring. Its effectiveness is demonstrated by superior performance across multiple amplitude stability metrics (e.g., reduction of standard deviation by 88.8%) and improved skin expansion recovery through an increased number of high-quality waveforms. The results indicate broad applicability, enhancing accuracy and reliability in radar-based small-scale motion extraction for various vital sign monitoring applications.
Incorporating spatial priors derived from high-resolution imaging modalities within a microwave imaging algorithm can significantly improve the resolution and the effectiveness of the imaging results. However, differences arising in positioning the patient inside the microwave system and the high-resolution diagnostic machine can significantly affect the imaging results. To alleviate such a risk, a new a-posteriori method for registration is proposed. The method uses qualitative inverse scattering techniques to derive a preliminary and reference image that is used to guide the registration of the high-resolution image at hand via the moment-based registration technique. To evaluate the performance, real data collected for human forearms, including both magnetic resonance images and microwave data, are considered.
Monitoring biochemical changes within hydrated and physically enclosed environments—such as wound beds beneath dressings, tissue scaffolds, and implantable matrices—demands sensing modalities that operate without embedded electrodes, optical windows, or wired connections. Here we demonstrate that a passive planar resonator loaded with a stimulus-responsive hydrogel converts analyte-driven permittivity changes into spectral signatures that a commercial 60-GHz frequency-modulated continuous-wave (FMCW) radar extracts non-contactly in the near field. A closed-form perturbation law, governed by a single geometry-defined electric-energy participation factor, provides a predictive rather than purely empirical framework for sensing design. Using pH-responsive hydrogels (pH 2–10) as a controlled dielectric stimulus, we obtain monotonic radar responses with high linearity ($R^{2} \approx 0.99$), strong simulation–experiment agreement ($\rho \approx 0.995$), and robust repeatability with low hysteresis across handling cycles, with interrogation performance quantified in terms of sensitivity and detection limit. Two complementary observables—FFT-peak magnitude and an RF-equivalent sweep-domain spectral-peak shift—are extracted from a single interrogation and benchmarked against a commercial Dielectric Assessment Kit to ground the radar observables in quantitative dielectric physics. Because the transduction mechanism is dielectric rather than chemically specific, the platform is compatible with hydrogels engineered for glucose, lactate, cardiovascular-relevant analytes, and other clinically relevant biomarkers, establishing a general route toward non-contact biochemical sensing at sealed or covered interfaces.
Real-time wireless capsule endoscopy (WCE) requires a stable, high-throughput in-body-to-out-body link. However, the procedure itself is inherently dynamic, and peristaltic movement drives continuous capsule translations and rotations that can cause strong channel variability and intermittent fades. While many WCE telemetry studies focus on transmitter design and static path-loss characterization, the receiver-side architecture is often treated as a fixed component, despite its critical role in preventing outages during the exam. In this work, we propose a dynamic singleinput multiple-output (SIMO) system that switches among external receiver antennas based on the capsule's 3D alignment to maximize the expected link margin at each instant. We validate the approach in a tissue-mimicking phantom while emulating capsule rotations and compare it with a single-antenna receiver baseline. Under the evaluated in-vitro conditions, the proposed SIMO system prevents connection losses and increases instantaneous throughput from 0.96/1.19Mbps (mean/median) to 1.55/1.64Mbps, corresponding to similar to 61% and similar to 37% improvements, while also improving RSSI stability. These results demonstrate the feasibility of dynamic receiver diversity for mitigating orientation fades, showing the approach's potential before clinical translation.
This study presents a novel immunoassay technique based on a highly sensitive microwave Whispering Gallery Mode (WGM) dielectric resonator fabricated from high-resistivity silicon and illuminated with 870 nm near-infrared (NIR) light. The dielectric resonator coupled to a microstrip transmission line, which is working between 23 GHz and 30 GHz, serves as the primary sensor for detecting antigen-antibody interactions. Antigen-antibody complexes are immobilized on the bottom surface of a transparent polystyrene plate, enabling NIR light from an LED to pass through the sample before reaching the resonator. The NIR illumination modulates the resonator's conductivity and induces measurable shifts in the transmission coefficient (S21) at the resonant frequency. These S(21 )variations correlate directly with binding events, yielding distinct signatures for antibody-only versus antibody-antigen samples. To demonstrate the quantitative capability of this WGM-based immunoassay, 25-hydroxyvitamin D (25(OH)D, calcidiol) was selected as a representative small-molecule biomarker. The sample under test (SUT) is positioned between the LED and the resonator, alters the resonator's coupling as NIR light passes through it, producing pronounced S-21 changes at the resonant modes. With its simplicity, low cost, and portability, this technique offers strong potential for practical applications, including viral detection, blood-based cancer diagnostics, and food safety monitoring.
This study characterized time-dependent changes in the dielectric properties of rat brain tissue during the first three hours following global ischemia. Adult male Sprague–Dawley rats (n = 10) were euthanized by decapitation to induce global ischemia, and each brain was immediately immersed in artificial cerebrospinal fluid maintained at 37 $^\circ$C.Using an open-ended coaxial probe connected to a vector network analyzer, we continuously measured the complex permittivity and conductivity of gray matter at discrete 5-minute intervals across the 0.5–10 GHz frequency range over a 3-hour window, while white matter properties were evaluated at the conclusion of the experiment. A detailed uncertainty analysis quantified both systematic and random errors, confirming that observed property changes exceed instrumental variability. In gray matter, relative permittivity decreased by approximately 5–8% over three hours, while conductivity showed a frequency-dependent reduction ranging from 18% at 0.5 GHz to 12% at 10 GHz, with the largest shifts occurring within the first 30 minutes post-ischemia. White matter exhibited smaller permittivity changes ($< 10$%) and less pronounced conductivity alterations. These findings demonstrate that ischemic injury produces measurable, dynamic dielectric contrasts between gray and white matter. The results lay a foundational framework for microwave-based imaging and real-time monitoring of ischemic events and may inform future development of portable dielectric-sensing devices for stroke diagnosis and treatment assessment.
Radiofrequency (RF) coil development for magnetic resonance imaging (MRI) relies on iterative bench-top adjustments in which tuning, impedance matching, preamplifier decoupling, and active detuning are optimized through repeated refinement of circuit component values and network analyzer measurements. Existing analytical descriptions used to guide this process commonly employ equivalent-circuit transformations that merge magnetic interactions into shared impedance elements, obscuring the contribution of individual components and not naturally accommodating the double-probe transmission measurement used during coil characterization. This work presents an alternative formulation in which electromagnetic coupling is introduced exclusively through dependent voltage sources, allowing each loop, whether belonging to the coil or to the double-probe, to retain its own circuit identity within a unified linear network. The scattering parameters observed on the bench follow directly from the mesh-current solution while maintaining a transparent one-to-one correspondence between lumped elements and their physical counterparts. Experimental validation on a receive coil at 123.25 MHz demonstrates close agreement between predicted and measured scattering parameters across all relevant operating states, with accurate reproduction of resonance frequencies, bandwidths, and spectral characteristics. The resulting network description provides a practical and extensible basis for predictive component selection and for the design and optimization of RF coils in clinical and research MRI.
Brain tumors are among the most lethal cancers globally. The varying types of brain tumors based on location, texture, and shape contribute to significant difficulty in detection, particularly from the computer vision perspective. Accurate early detection of the tumor's type and grade is critical for selecting appropriate treatment plans. In this work The author's present a novel convolutional neural network learning model for brain tumor classification. The author's build an optimized deep network, with the aim to achieve efficiency in terms of both computing time and classification accuracy. The proposed architecture incorporates a self-attention block with l(2)-normalization and a skip connection. The proposed model achieves a forward-pass inference time of less than 20 ms per image while maintaining a high classification accuracy of 97.15% on the "2D T1-weighted CE-MRI" dataset and 96.98% on the BRISC 2025 dataset, considering only three tumor classes (glioma, meningioma, and pituitary) and excluding the non-tumor class, as well as 96.79% on the four-class brain tumor dataset. This work presents a novel brain tumor MRI image classification technique utilizing a self-attention block with l(2)-normalization and a skip connection. Experimental results demonstrate the model's suitability for real-time clinical use with high accuracy.
In this work, we present and experimentally validate a fully implantable radio frequency system capable of inducing mild heating as a response to a received electromagnetic wave. The heating process is initiated by a transmitter antenna located outside the human body. It is received by a compact meandered miniaturized loop antenna integrated through a differential matching network with a planar inductive resonant LC heating circuit. The design process incorporates the complexities of the human body propagation channel, and the system is validated through phantom-based experiments and ex-vivo testing. Experimental results demonstrate strong agreement with simulation data, and thermal assessments reveal a consistent and controlled temperature increase, confirming the effectiveness of the proposed system. The experimental results provide a 0.01 degrees C increase in temperature per minute in phantom testing, whereas ex-vivo testing reveals an average change in temperature of around 0.06 degrees C per minute when relying on stacked organs from sacrificed mice. Such subtle temperature increases are in the range reported to modulate metabolic activity in biological tissues. This work establishes a foundational platform for future medical interventions utilizing continuous, non-invasive wireless heating within the human body.
Microwave breast imaging is a promising non-ionizing modality, but lesion assessment remains difficult because linearized dielectric reconstructions are often dominated by artifacts, noise, and model mismatch. This work proposes a physics-informed post-reconstruction anomaly-support framework for generating interpretable spatial support maps from multi-frequency microwave data. The method combines deviation from expected benign dielectric ranges, broadband consistency of reconstructed permittivity, and local spatial coherence to suppress diffuse artifacts and highlight physically plausible support regions. The framework was evaluated on 500 anatomically variable three-dimensional breast phantoms, comprising 250 reference and 250 malignant cases. Evaluation included receiver operating characteristic analysis, precision-recall analysis, localization-oriented metrics, region-wise reconstructed permittivity-frequency profiles, support-map baselines, and parameter sensitivity analysis. Across malignant cases, the proposed support map substantially increased tumor-to-background support contrast, with a median ratio of 1.88 [1.68, 2.07], compared with 0.16 [0.12, 0.21] for raw adjoint energy and 0.17 [0.13, 0.22] for broadband energy-only mapping. However, precise localization remained limited, with a median peak-to-tumor distance of 24.38 mm and a 5-millimeter lesion-hit rate of 6.8%. Full-dataset case-level discrimination remained near chance, with receiver operating characteristic area-under-the-curve values of 0.474 for the proposed score and 0.479 for the contrast-based baseline. Region-wise profile analysis showed that Born-adjoint reconstruction compresses reconstructed permittivity trends; therefore, broadband consistency should be interpreted as a reconstructed consistency cue rather than quantitative recovery of tissue dispersion. The framework is best viewed as a physics-guided support-contrast enhancement tool, not a stand-alone classifier, precise localizer, or segmentation method.
This paper presents a highly miniaturized two-channel flexible epidermal antenna to host the passive wireless bilateral monitoring of nasal breathing based on temperature measurement. The device fits the nose as a septum piercing and is made of highly flexible and soft biocompatible materials. It comprises two coupled loop antennas whose Ultra-High Frequency (UHF) Radio Frequency IDentification (RAIN RFID) temperature-sensing microchips are placed just inside the nostrils to provide both sensing and transmission of breath data, with stable performance thanks to auto-tuning mechanism. To support the electromagnetic design of such a compact dual-channel architecture, the device is analyzed through a dedicated model that accounts for the mutual coupling between the loops and the variable-impedance behavior of the auto-tuning ICs. Compared to previous epidermal solutions for bilateral nasal monitoring, the proposed dual antenna reduces the footprint by nearly two orders of magnitude. The experimental evaluation, involving ten subjects, statistically demonstrates a maximum reading distance above 40 cm, and up to 80 cm, for different interrogation conditions, taking into account both intra-person variability and reading orientation. Overall, the system demonstrates excellent channel symmetry (less than 1 dB in over 85% of cases) and stability over eventual frequency shifts (bandwidth wider than 100 MHz in over 90% of cases).
This study aimed to design, fabricate, and evaluate a large metasurface to significantly enhance the signal-to-noise ratio (SNR) of a commercial rat head coil across a relatively large region within the coil’s field of view, as well as a novel device called a frequency response compensator (FRC) to compensate for the coil detuning induced by metasurface integration. The geometries of both the metasurface and FRC were analyzed and optimized using full-wave electromagnetic simulations in CST Studio Suite. Both components were fabricated from capacitively loaded copper loops printed on a flexible dielectric substrate. The metasurface, configured as a rectangular loop, was positioned between the rat coil and the phantom, while the FRC—comprising two rectangular loops with tunable capacitors—was placed above the coil and wirelessly coupled to it. The enhancement in SNR was then simulated and experimentally measured to validate the effectiveness of these designs. Benchtop measurements showed that integrating the metasurface with the rat coil increased the SNR by up to 77% at the mouse-brain location (MBL), but required manual coil retuning due to the altered frequency response. Incorporating the FRC subsequently improved the SNR by up to 65% at the MBL and restored the coil’s frequency response. These consistent enhancements, confirmed by both simulation and experimental data, indicate that the developed metasurface and FRC provide a promising approach for improving the RF coil’s SNR without altering its frequency response.
Conductive Intracardiac Communication (CIC) provides a low-power data transmission approach for cardiac resynchronization therapy (CRT). However, quantitative characterization of CIC channel dynamics remains challenging because physiologically informed models that capture cardiac deformation and anatomical variability remain limited. This work establishes a closed-loop bidirectionally coupled electromechanical model in a multiphysics finite-element framework for a biventricular leadless pacemaker (BiV-LP) system. The model integrates electrophysiological activation (Aliev-Panfilov), nonlinear myocardial mechanics (Holzapfel-Gasser-Ogden), and an electro-quasi-static electromagnetic formulation to simulate CIC transmission over 100 kHz-100 MHz in a thoracic computational domain. Model plausibility was evaluated against electrophysiological and deformation data. Simulations show that CIC channel gain increases with frequency and exhibits periodic modulation over the cardiac cycle, with the largest peak-to-peak fluctuation in the MHz range. Gain variation is phase-locked to ventricular mechanics, with reduced fluctuation near end-systole and increased variation near end-diastole. Implantation location significantly influences channel stability: the right ventricular apex presents the largest fluctuation, whereas septal and outflow-tract sites exhibit improved robustness. Correlation analysis indicates that septal and outflow-tract modulation is primarily governed by local myocardial deformation, while apical variation is more sensitive to global ventricular geometry. Ex-vivo pump-driven ventricular deformation experiments support the deformation-induced CIC modulation mechanism. Static measurements confirm stable myocardial conductive coupling, and sandwich-controlled dynamic measurements demonstrate repeatable gain modulation consistent with simulation trends. These results clarify how cardiac motion and implantation position shape CIC channel dynamics and provide a physics-consistent framework for analyzing implantation-dependent CIC behavior under dynamic physiological conditions.
This paper presents a compact dual-band implantable antenna designed for subcutaneous temperature monitoring in bovine udder tissue. The antenna exhibits two -10 dB impedance bands of 0.36-0.81 GHz and 2.07-3.72 GHz, covering the 430 MHz telemetry band and the 2.45 GHz ISM band. Wide impedance bandwidth in both bands is achieved by combining a slot-loaded patch with an asymmetric spiral defected ground structure, which extends the low-frequency current path and introduces multiple resonances in the upper band. The antenna occupies only 5.8 & times; 6.0 & times; 0.254 mm(3) (8.84 mm(3)) on a Rogers RO3003 substrate. A prototype was fabricated and measured in minced bovine tissue. The measured -10 dB impedance bands are 0.39-0.63 GHz and 2.33-4.47 GHz, showing good agreement with the simulated results. The measured peak gains are -15.01 dBi at 430 MHz and -20.22 dBi at 2.45 GHz. When the input power is scaled to the implant transmit limit of -16 dBm (25 mu W), the 1-g averaged SAR values are 0.0084 W/kg and 0.0124 W/kg, respectively. The link-budget analysis indicates that the 430 MHz link provides robust low-rate telemetry over a representative 15 m readout range, whereas the 2.45 GHz link supports short-range higher-rate data upload.
This study proposes an unsupervised deep learning framework for early skin cancer detection using millimeter-wave imaging in the 25-45 GHz frequency range. Time-domain electromagnetic signals were acquired from tissue-mimicking phantoms and preprocessed using a combination of Principal Component Analysis (PCA), Self-Organizing Maps (SOM), and autoencoders. Notably, the preprocessing pipeline requires only a single scan, without the need to subtract signals from tumor-free and tumor-containing regions, making the method more practical and realistic for clinical applications. Tumor localization and imaging were performed using the Delay-and-Sum algorithm, resulting in accurate spatial reconstruction of tumor regions. Experimental results demonstrated that the integrated SOM autoencoder approach consistently detected tumors across varying positions and quantities, indicating strong robustness and generalization capability. Importantly, the system achieved accurate tumor detection without the need for labeled training data, offering a significant advantage over supervised models that rely on large, annotated datasets. These findings highlight the potential of the proposed framework as a practical, non-invasive, and scalable tool for early-stage skin cancer diagnosis.
