The signal is useful, but not uniform
An uncomplicated microwave setup registered changes in a simulated chest and in a lab phantom as the modeled lung's dielectric loading, the electrical properties assigned to it, changed. Yet the signal did not behave as one simple scale across the two anatomies and the tested ranges. The work used torso models and a controlled water phantom, not in-vivo measurements, and the authors describe it as methodological feasibility only.
The researchers were asking how little measurement complexity could preserve useful sensitivity to lung dielectric loading while remaining robust to antenna-positioning variability. Their setup used an opposed antenna pair, a nominal 6 cm antenna-to-chest distance, 1 mW of transmitted power and a 1 to 2 GHz operating band. The main readout, called S21, was the transmitted response from one antenna to the other, reported in decibels. Reflected S11 served as a check on antenna matching and geometric stability rather than as a direct fluid marker.
What the first simulation found
For Gustav, the adult-male torso model, the lung assignment took five relative-permittivity states, a numerical description of the model's electrical response: 20, 30, 45, 55 and 75. Electrical conductivity was scaled from 0.1 to 0.5 S/m. A 3 by 3 grid tested lateral and head-to-foot shifts of minus 10, zero and plus 10 mm, using relative permittivity 31.6 and conductivity 0.306 S/m as the positioning reference.
In Gustav, band-averaged S21 did not move steadily across the full modeled range. It was minus 68.4 dB at relative permittivity 20, rose to minus 57.1 dB at 45, and then declined to minus 60.2 dB at 75. No frequency point across the full 1 to 2 GHz sweep was strictly monotonic. A narrower window from 1.194 to 1.348 GHz did show a monotonic decrease for the 45, 55 and 75 states.
Placement changes the readout
Positioning changed the result as well. Gustav's baseline in-band S21 was minus 62.6 dB, while shifted positions ranged from minus 65.2 to minus 60.2 dB. The maximum departure was 2.57 dB and the mean absolute departure was 1.70 dB. Head-to-foot shifts of 10 mm produced 4.4 dB of peak-to-peak variation, compared with 0.5 dB for lateral shifts.
To compare the two sources of change, the study divided the dielectric-response span by the largest positioning change. In Gustav, that ratio was 4.4 for states 20 to 75, 2.6 for 30 to 75, and 1.2 for 45 to 75. The corresponding dielectric spans were 11.3, 6.6 and 3.1 dB, against a 2.57 dB maximum positioning change. Because the ratio changed with the chosen range, it is best read as a range-specific comparison, not a universal robustness factor.
A second anatomy shifts the picture
The researchers repeated the protocol with Emma, the adult-female torso model, using the same antenna geometry and the same 1 to 2 GHz range, 6 cm spacing and 3 by 3 grid of minus 10, zero and plus 10 mm shifts. Its baseline intermediate-loading S21 was minus 49.9 dB, compared with minus 62.6 dB in Gustav. In both anatomies, the mean in-band response was non-monotonic from 20 to 75. The apparent maximum was near 45 in Gustav but near 30 in Emma. The predefined 1.15 to 1.35 GHz window therefore remained informative in Emma, but the results did not support one universal scalar threshold.
Emma was less affected by the simulated translations, with a maximum positioning-induced change of 1.4 dB. Its dielectric-to-positioning ratios were 5.1 for 20 to 75, 1.6 for 30 to 75, and 2.8 for 45 to 75. Cranio-caudal, or head-to-foot, displacement was stronger than lateral displacement. The response spans also differed by anatomy: Gustav showed 11.3, 6.6 and 3.1 dB across the three ranges, while Emma showed 7.1, 2.2 and 3.9 dB.
The phantom points in the same direction
The physical check used a simplified planar phantom at the same nominal 6 cm antenna distance. Water was added in 75 mL increments from 375 to 1,050 mL, with three consecutive sweeps at each loading and no disassembly between sweeps. Across the meaningful 375 to 900 mL range, mean band-averaged S21 fell from minus 37.7 dB to minus 50.6 dB. Six of seven increments moved downward, producing a total 12.9 dB drop and supporting the direction of the simulated upper-permittivity response.
Still a feasibility study
The study's limits are substantial. It used only two voxel models, and the main positioning analysis covered shifts of only plus or minus 10 mm at a fixed antenna distance. The phantom was planar, with foam and uncontrolled dielectric loss. The authors say those constraints limit generalization. They also say clinical deployment would require in-vivo reproducibility under repeated repositioning, validation against quantitative reference imaging and prospective patient evaluation.
The research was supported by the Innovation Incubator 4.0 program and the Polish Mother's Memorial Hospital Research Institute under grant PMMHRI-BCO.75/2020. M.Ś., I.Z., A.BD and W.O. are named inventors on Polish Patent No. 245498 and PCT/IB2023/055570 related to the system; the remaining authors declare no competing interests.
Paper data and sources
Original title: Positioning-robust broadband microwave transmission as a minimal-complexity approach to thoracic dielectric sensing: simulation and phantom analysis.
Authors: Maciej Ślot, Ilona Zasada, Wielisław Olejniczak et al.
Journal/Repository: Scientific reports
Status: Peer-reviewed
First online: 2026-08-20
DOI: 10.1038/s41598-026-60247-w
Original paper · Full text