As one of the ten most common cancers worldwide, kidney cancer is a global concern [1]. In this study, radiomics and artificial intelligence (AI) techniques are employed to diagnose kidney lesions. Notably, relevant statistical, geometric, and tissue-related features can be easily identified through radiomics with the help of neural networks. Subsequently, the extracted features can be analyzed to differentiate malignant lesions from benign ones and classify renal cell carcinoma (RCC) subtypes, as well as to predict Fuhrman grade, genetic mutations, and response to immunotherapy. In this context, the findings of a comprehensive review of the literature up to July 2022, which was conducted to examine the diagnostic value of radiomics, grading, biomarkers, and ongoing clinical trials, suggest that standardizing imaging protocols along with the use of AI and radiomics can improve sensitivity, specificity, and accuracy in diagnosing kidney tumors [2]. Currently, physical examination, blood and urine tests, biopsies, and imaging techniques are primarily used to detect cancer. While these tests cannot conclusively determine whether a person has kidney cancer, they sometimes provide the first clue pointing to a possible kidney problem. The imaging techniques adopted for this purpose include computed tomography (CT) scans, X-rays, magnetic resonance imaging (MRI), and cystoscopy [3].

Given this background, a previous article examined the role of microwave imaging techniques in the early detection of lung cancer based on the dielectric properties of normal and cancerous tissues [4]. Identifying dielectric contrasts between normal and cancerous lung tissues proved crucial for image reconstruction in radar-based or tomographic methods. Furthermore, this article drew on microwave measurement data from previous studies to identify important parameters, such as the optimal distance of the antenna from the human thorax and the detectable tumor size. Overall, this study presents a safe, rapid, and cost-effective method for the early detection of kidney cancer. Another study employed a circularly polarized antenna array operating at 2.4 GHz, designed using CST software [5]. The performance of the system, consisting of four FR4-based patch antennas, was tested using a human kidney phantom model. Shifts in S₁₁ and resonance frequency were analyzed considering multiple tumor stages. Notably, in the early stages, frequency shifts were observed to be minimal, while increases in S₁₁ were significant. Meanwhile, in the advanced stages, increases in both frequency shifts and S₁₁ were more pronounced. Moreover, this method guaranteed safety owing to its low specific absorption rate (SAR). Overall, this technique represents a non-invasive, radiation-free, comfortable, and affordable screening approach. These findings suggest that microwave- based methods can be considered an alternative approach for the detection of kidney cancer by leveraging differences in the dielectric properties of healthy and malignant kidney tissues. In this regard, the existing literature has established that circularly polarized patch antenna arrays enable the identification of kidney tumors based on changes in S₁₁ and resonance frequency [6]. To improve detection performance, previous studies have proposed using cavity-based slotted waveguide antennas and antenna structures manufactured using three-dimensional (3D) printers. The applicability of these structures in kidney cancer detection has also been experimentally and numerically investigated [7]. In addition, it has been reported that microwave antennas operating in the ultra high frequency (UHF) band can detect early-stage renal cell carcinoma based on differences in the electromagnetic responses of tumorous and normal kidney tissues [8]. With regard to proximity- based sensing approaches, implantable and miniature antenna designs have been numerically evaluated based on the effects of malignant kidney tissue on antenna resonance characteristics. The literature also consists of a number of studies that have suggested different antenna array methodologies for the early detection of kidney tumors. However, simulation-based investigations supported by antenna measurements have revealed that the presence of malignant kidney tissue affects the resonance behavior of antennas through changes in the reflection coefficient and related parameters [9–11]. Nonetheless, in recent years, microwave imaging (MWI) has emerged as a non-ionizing and cost-effective method in healthcare, particularly in the field of medical imaging. Simultaneously, advances in AI have significantly enhanced the capabilities of medical imaging tools. In light of these developments, researchers chose to explore the intersection of these two fields by examining the integration of AI algorithms into MWI techniques and investigating their impact on accuracy and overall performance [12]. In addition to highlighting the latest developments and historical context of MWI technology, they undertook the chronological classification of AI-assisted MWI, traced the evolution of intelligent systems in this domain, and provided a critical review of prominent studies while also offering a detailed understanding of the progress made and challenges faced in this area. Yet another study proposed a circularly polarized patch antenna array operating at 2.4 GHz to detect kidney cancer using microwave techniques. The researchers employed a kidney phantom model for analysis, accounting for four cancer stages. Tumors were found to cause increased S₁₁ and frequency shifts, thereby aiding detection, especially in advanced stages. Moreover, the SAR values remained within the safety limits. This method is safe, compact, fast, low-cost, non-invasive, and free from ionizing radiation. Therefore, it can be suitably employed in the quick detection and assessment of patients showing signs of bladder and kidney cancer.

Recently, wearable antennas have attracted significant interest due to their potential applications. They are an essential component of the wireless body area network (WBAN) systems used in sports, healthcare, the military, and identification. Unlike traditional antennas, these devices maintain direct contact with human tissue, which affects parameters such as gain, radiation pattern, bandwidth, reflection loss, directivity, efficiency, and SAR. With regard to detection, the goal of such devices is to identify tumors by understanding how microwave signals interact with body tissues, given that cancerous tissues have different dielectric properties than healthy ones. More importantly, microwave techniques help extract diagnostic information without the need for invasive surgery. In this context, a previous study thoroughly analyzed the requirements and performance metrics of wearable antennas, focusing on SAR evaluations and microwave-based antenna research, to provide a comprehensive and up-to-date overview of the latest developments in the medical application of microwaves, with a particular focus on near-field systems operating in the 1–15 GHz frequency range. Another study discussed the technical implementations and algorithmic aspects of the prototypes and devices developed for detecting certain diseases, including transmitter-receiver systems, radiating elements (such as antennas and sensors), and imaging algorithms. This article also noted the challenges associated with the clinical usability of microwave-based technologies and presented future perspectives [13].

Against this backdrop, the current study presents a microwave-based approach for renal tumor detection by (1) characterizing, designing, and fabricating a 2.4 GHz rectangular microstrip patch on PF-4 (S₁₁ was measured using a vector network analyzer [VNA], and the radiation pattern was measured in an anechoic chamber); (2) building a four-layer (skin-fat-muscle-kidney) CST model with 10.2-mm-deep ellipsoidal inclusions (x-radius 2 mm, y-radius 1 mm, height 6 mm), positioned at a distance of 122.84 mm from the antenna; and (3) conducting a 25–100 mm distance sweep at 0.5 W to arrive at the SAR-compliant placement value of 87.84 mm (accepted power approximately 0.492 W; max 10-g SAR approximately 1.71 W/kg; total SAR approximately 1.22 W/kg). Furthermore, departing from wide-spectrum S-parameter heuristics, the detection procedure adopted in this study relies only on the first local maximum of the receiver-port time-domain signal o21(t), estimating the normalized phase lead Δϕ and amplitude rise ΔA. According to the CST results, the tumor-free → one-tumor transition yielded Δϕ approximately 125° with ΔA approximately 15%, while adding a second and third tumor introduced additional approximately 99° and approximately 81° phase steps, respectively, along with approximately 11%–16% amplitude increases, thus enabling the graded separation of 1–3 tumors.

Moreover, a practical limit was observed: reducing the dimensions of the inclusion to 1.9 mm × 0.9 mm × 5 mm rendered the o21(t) indistinguishable from the baseline, indicating that tumors ≥2 mm × 1 mm × 6 mm can be reliably detected using the proposed method. Overall, the proposed method requires only one receiver-port observation and a single peak, thereby enabling fast, non-ionizing, and low-cost screening.

A two-antenna microwave system was designed to detect elliptical tumors inside a kidney (Fig. 1). In particular, Tx-Rx microstrip antennas and a kidney system model were simulated in CST Studio Suite. A rectangular microstrip patch operating at 2.4 GHz was designed and then implemented. Notably, the radiation pattern of the antenna was measured in an anechoic chamber, and its S₁₁ was measured using VNA. Meanwhile, the kidney system model employed in this study comprised four layers (skin-muscle-fat-kidney) characterized by frequency-dependent permittivity and conductivity. Elliptical tumors, which exhibited lower permittivity and higher conductivity than kidney tissue, were inserted into the kidney model. The nominal tumor size was x-radius = 2 mm, y-radius = 1 mm, and height = 6 mm. The tumors were placed 10.2 mm deep below the kidney surface and 122.84 mm away from the antenna. Notably, when the tumor dimensions were reduced to 1.9 mm × 0.9 mm × 5 mm at the same location, the receiver-port o21(t) signal became indistinguishable from its tumor-free counterpart. Therefore, a detection threshold of ≥2 mm × 1 mm × 6 mm was established for the present setup. Furthermore, a SAR sensitivity sweep (0.5 W stimulus) was performed to ensure accurate antenna-system separation considering distances of 25, 50, 75, 87.84, and 100 mm. Ultimately, 87.84 mm was selected as the appropriate distance (maximum 10 g SAR, approximately 1.71 W/kg; total SAR, approximately 1.22 W/kg).

In the simulation, a pulse launched by the Tx antenna propagated through the layers and interacted with the tumor, following which the scattered field returned to the Rx port. At this point, the time-domain receiver-port o21(t) signal was recorded. For detection/staging, only the first local maximum the first peak phase lead Δϕ (normalized by the source period) and the first peak amplitude increase ΔA (%) was used, enabling fast and robust detection and staging of 1–3 tumors from a single receive-port observation.

Fig. 2(a) shows the final tumor locations corresponding to the maximum o21(t) signal values. As shown in Fig. 2(b), an elliptically shaped tumor with dimensions of 2 mm × 1 mm × 6 mm was positioned inside the kidney, starting from the uppermost region and progressing toward the lowermost region from left to right in 5 mm increments, and the corresponding o21(t) value was recorded at each location. When the end of a row was reached, the tumor was shifted 5 mm downward and returned to the far-left position of the next row, and this procedure was repeated until the entire kidney had been scanned. After recording the o21(t) value at every position, Tumor 1 was permanently placed at the location where the maximum o21(t) value was obtained. Subsequently, while Tumor 1 was kept fixed at its permanent location, the same procedure was repeated for Tumor 2 of the same size, and Tumor 2 was permanently placed at the location yielding the maximum o21(t) value. Finally, with Tumors 1 and 2 maintained at their permanent locations, the same procedure was applied to Tumor 3, which was then permanently positioned at the location corresponding to the maximum o21(t) value. Before permanent placement, an elliptical volume matching the tumor dimensions was removed from the kidney to create a cavity for each tumor. Each tumor was embedded at a depth of 10.2 mm within the kidney and had an x-radius of 2 mm, a y-radius of 1 mm, and a height of 6 mm.

When tissues in the human body are exposed to an electromagnetic field, their electrical permeability and electrical conductivity values change with the frequency of the applied field. In this study, modeling of the tumor(s) in the kidney was conducted by assuming that the kidneys with and without a tumor exhibit different electrical properties. The kidney model used in this study was designed in a CST environment and comprised four layers characterized by different dielectric and electrical conductivity values. An ellipse-shaped tumor was placed inside the kidney. Notably, the dielectric constant value of the tumor was lower than that of the kidney, while its electrical conductivity value was higher. The designed kidney model is illustrated in Fig. 1, and the properties of the four different layers of the kidney model are listed in Table 1 [14].

A rectangular patch microstrip antenna operating in the 1–10 GHz frequency range at 2.4 GHz was designed for further analysis. The properties of the insulator and conductor employed to construct the antenna are shown in Table 2.

The proposed antenna was fed by a transmitter microwave source via the inset feed method. The dimensions of the designed antenna are illustrated in Fig. 3. To conduct further analysis, the rectangular microstrip antenna designed in the CST environment was fabricated.

The simulation results of an antenna serve as important criteria for validating its design quality. In this context, the frequency- dependent variation of S₁₁ in a system indicates the operating band and frequency of the antenna. The frequency-dependent S₁₁ variations of the kidney models with and without a tumor are presented in Fig. 4. Meanwhile, Fig. 5 traces the voltage standing wave ratio (VSWR), showing negligible differences between the tumor-free and three-tumor loadings. It is observed that the resonance remains at around 2.4 GHz, with nearly identical depth and bandwidth. This indicates that the presence of tumors did not substantially alter antenna matching, thereby supporting the proposed detection approach where the first peak phase Δϕ and first peak amplitude ΔA rely on propagation/scattering changes at the receiver port, and not on matching shifts.

In Fig. 6, the 3D change in antenna directivity with respect to the theta and phi angles is illustrated. Furthermore, the two-dimensional horizontal coverage of the directivity of the designed and fabricated antennas is depicted in Fig. 7. It is evident that the directivity of the proposed antenna’s main lobe is better than that of its back lobe. In contrast, in Fig. 8, the antenna radiation patterns demonstrate better results for both the main lobe and the back lobe in vertical coverage.

The curve depicted in Fig. 9 exhibits a stable plateau of approximately 4.5–5.3 dB over 2–6 GHz. Furthermore, around the target band of 2.4 GHz, the gain is approximately 5–5.3 dB, which is adequate for the link budget. A distinct dip in gain (approximately 2 dB) appears near 7–8 GHz, likely due to higher- order modes/mismatch. Thereafter, above 9 GHz, the gain rises back to approximately 5 dB and then declines to approximately 3 dB at 10 GHz. In summary, within the band of interest (2.4 GHz), the antenna provides approximately 5 dB realized gain, supporting SNR at the receiver port and strengthening the reliability of detection based on the first peak phase Δϕ and first peak amplitude ΔA.

Fig. 10(a) presents the setup for measuring changes in the directivity of the proposed antenna with regard to frequency in an anechoic chamber, while Fig. 10(b) shows the measurement setup for calculating the S₁₁ of the antenna using a VNA, as well as the change in S₁₁ in terms of frequency. Furthermore, Fig. 11 shows a 3D representation of changes in the system as a result of SAR analysis.

Table 3 reports the computed SAR for five antenna–system separations at 2.4 GHz, considering a 0.5 W input. Among the multiple distances simulated to examine the trend, the configuration chosen for the kidney model in this study was 87.84 mm, since it yielded a maximum 10 g SAR of 1.71 W/kg, a maximum 1 g SAR of 5.39 W/kg, and a total SAR of 1.22 W/kg. As expected, localized exposure declined with an increase in separation. For instance, Table 3 shows that the 10 g metric drops from 4.6 W/kg (25 mm) to 1.39 W/kg (100 mm), while the 1 g metric declines from 20.41 W/kg to 4.38 W/kg with an increase in separation [15].

Fig. 12 illustrates the expected ordering: 3-tumor leads, followed by 2-tumor, 1-tumor, and tumor-free, consistent with the first peak time. The numerical first-peak results are as follows:

• 0 → 1: t1 approximately 0.912 ns, t2 approximately 0.849 ns → Δϕ approximately 125.22°, with the amplitude being 0.0453 → 0.0525 → ΔA approximately 15.85%.

• 1 → 2: 0.8488 → 0.7984 ns → Δϕ approximately 99.13°, with the amplitude being 0.05249 → 0.05827 → ΔA approximately 11.03%.

• 2 → 3: 0.7984 → 0.7586 ns → Δϕ approximately 80.60°, with the amplitude being 0.05827 → 0.06754 → ΔA approximately 15.90%.

The above results imply that the approximately 125° first peak phase in 0→1 serves as a strong detector of tumor presence, while the approximately 80°–100° phase steps and approximately 11%–16% amplitude growth across 1→2→3 imply reliable staging. Table 4 summarizes a “first-peak-only” transition analysis of the receiver-port response [16]. Notably, while many studies rely on S₁₁ shifts, the proposed port-based time-domain metric (first peak only) delivered a strong 0→1 separation (Δϕ approximately 125°, ΔA approximately 15 %), along with consistent approximately 80°–100° phase steps with approximately 11%–16% amplitude growth for 1→2→3 staging. In addition, the proposed antenna offers SAR safety, a clear detection threshold (≥2 mm × 1 mm × 6 mm), and the potential for construction using off-the-shelf materials, thereby facilitating practical, non-ionizing, and low-cost screening.

In the first row of Table 4, the o21(t) signal corresponding to the non-tumorous kidney was considered the first signal, while the signal corresponding to the kidney with one tumor was considered the second signal. In the second row, the signal corresponding to the kidney with one tumor was taken as the first signal, and the signal corresponding to the kidney with two tumors was taken as the second signal. In the third row, the signal corresponding to the kidney with two tumors was taken as the first signal, whereas the 021(t) signal corresponding to the kidney with three tumors was taken as the second signal. t1 (ns) is the time at which the first signal reaches its first peak value, t2 (ns) is the time at which the second signal reaches its first peak value. A1 is the maximum value of the first signal, A2 is the maximum value of the second signal, Δϕ is the phase (°) difference between the first and second signals. ΔA is the percentage increase in amplitude from the maximum of the first signal to the maximum of the second signal written as

In this paper, a microwave-based approach for renal tumor detection is proposed. A kidney system model and a 2.4 GHz rectangular microstrip patch antenna were designed in CST. The antenna was fabricated, its radiation pattern was measured in an anechoic chamber, and its S₁₁ was measured using a VNA. The detection process involved only the first local maximum of the receiver-port time-domain o21(t) signal, estimated using two features: the first peak phase lead Δϕ and the first peak amplitude increase ΔA (%).

Detection performance: For the tumor-free kidney → one tumor kidney transition, Δϕ approximately 125° with ΔA 15% was achieved. For 1 tumor kidney → 2 tumor kidney and 2 tumor kidney → 3 tumor kidney, the phase steps were observed to be approximately 99° and approximately 81°, respectively, accompanied by an amplitude increase approximately 11%–16%. This enabled the graded separation of 1–3 tumors using only the first peak, with Δϕ being the primary discriminator and ΔA being the supportive metric.

Geometry and location: Elliptical tumors with an x-radius of 2 mm, y-radius of 1 mm, and height of 6 mm were placed 10.2 mm deep inside the kidney and 122.84 mm away from the antenna. These dimensions denote the system’s detection threshold. Notably, for a tumor of dimensions x = 1.9 mm, y = 0.9 mm, and z = 5 mm, the o21 phase and amplitude at the receiver were indistinguishable from the tumor-free case.

Thus, tumors smaller than an ellipse of dimensions x = 2 mm, y = 1 mm, and z = 6 mm cannot be reliably separated using the proposed setup.

SAR and placement: Considering a 0.5 W stimulus, SAR sensitivity was evaluated at 25, 50, 75, 87.84, and 100 mm antenna-system separations. At 87.84 mm, the accepted power was calculated to be approximately 0.492 W, the maximum 10 g SAR approximately 1.71 W/kg, and the total SAR approximately 1.22 W/kg. Hence, this distance was selected for further analysis. Moreover, the proposed antenna operates at 2.4 GHz, which aligns with commonly used wireless communication bands. Furthermore, its components PF-4, which is easily available at a low cost, and single-sided adhesive copper tape facilitated the realization of a lightweight wearable antenna.

The proposed system successfully distinguished tumor-free cases from 1–3 tumor cases with high accuracy. It also managed to detect tumors as small as 2 mm × 1 mm × 6 mm, even at a 122.84 mm antenna standoff, underscoring its practical significance. Furthermore, unlike much of the prior literature, detection is carried out based only on the first local maximum of the receiver-port o21(t) signal by leveraging its amplitude and phase via Δϕ and ΔA, indicating another advantage offered by the proposed approach. Nonetheless, it must be noted that the proposed method currently has a limitation pertaining to the observed detection threshold: Tumors smaller than 2 mm × 1 mm × 6 mm could not be reliably separated in the present setup. Therefore, improving this sensitivity must be focused on in future work.