Gut microbiome diagnostics today work by collecting stool, shipping it to a lab, and sequencing it. That pipeline takes hours to days, and it only sees what reaches the end of the digestive tract. An ingestible capsule that measures a few microbial markers in place and radios them out would remove both problems. The obstacle is the radio link, not the chemistry. The capsule has a tiny battery and a strict tissue-heating limit. Its signal passes through centimeters of lossy, uneven tissue. As it travels roughly nine meters of gut, it keeps moving between well-placed and badly placed positions relative to any fixed receiver on the skin.
This note proposes putting all mechanical flexibility on the outside. A wearable belt carries receive antennas that physically slide to follow the capsule. These are "movable antennas," radiators repositioned by small motors, and "pinching antennas," clips that slide along a plastic waveguide and turn a chosen point on it into a radiator. The capsule itself stays simple and fixed. The design goal also changes. Wireless papers usually try to maximize data rate. A microbiome reading needs only a few bytes, and it changes over hours. The right target is the capsule's energy per reliably delivered reading, within a deadline set by how fast the microbial community changes. The central question is whether moving the receiver buys anything that cheaper tricks, such as waiting for a good moment or adding more fixed antennas, cannot.
Core Claim
The capsule link loses signal in two ways, and they call for different fixes. The first is small-scale fading: fast dips caused by the capsule tumbling and by signal echoes. Time fixes these for free, because microbial state changes slowly and the capsule can simply buffer a reading and send it seconds later. The second is large-scale loss. This is persistent extra attenuation set by where the capsule sits in the abdomen relative to the receiver: tissue depth, bone and gas in the path. It can last for the whole time the capsule stays in one gut segment, so waiting does not help. Our claim is that repositioning the receiver on the belt reduces this persistent, position-dependent loss. That lowers capsule transmit energy per delivered reading beyond what opportunistic waiting and fixed-array antenna selection achieve together. If the claim holds, the gain should concentrate in deep, slow-transit segments and disappear where waiting already suffices. That prediction is sharp and could fail.
No paper in the file connects these fields. The gap analysis names the asymmetry we build on: actuation energy that would be unaffordable on an implant costs nothing to the capsule if it sits in the external receiver. No one has examined that asymmetry.
Relation to Prior Work
The antenna literature supplies the methods and the motivating evidence, but all of it is terrestrial.
- Multiuser Communications with Movable-Antenna Base Station Via Antenna Position Optimization optimizes positions of 16 base-station antennas with particle swarm search. It reports a 2.36 bps/Hz minimum user rate versus 1.44 for fixed antennas.
- Movable Antenna Enabled Integrated Sensing and Communication contributes a three-stage grid-search, gradient-ascent and projection positioning method (SPGA). We reuse it as our belt controller.
- Joint Beamforming and Antenna Design for Near-Field Fluid Antenna System shows that port selection under spherical wavefronts, the situation a few centimeters from a body, beats fixed antennas by 49.07% in rate.
- The pinching-antenna work, Pinching-Antenna Systems (PASS): A Tutorial and Flexible-Antenna Systems: A Pinching-Antenna Perspective, argues that relocating a radiation point over meters removes large-scale loss. That is exactly the loss component we target, but only in 20 m rooms.
- Delay Minimization in Pinching-Antenna-Enabled NOMA-MEC Networks co-optimizes pinch positions with power under a delay objective. That formulation is the closest to our energy-under-deadline objective.
All of these model free-space fading and maximize rate or throughput. We replace the channel with a tissue model and the objective with capsule energy under a biological deadline.
Flexible-Position MIMO for Wireless Communications: Fundamentals, Challenges, and Future Directions plays a double role. It motivates the proposal with its equivalence result: 4 flexible antennas over 30 positions matched 20 fixed ones. It also constrains the proposal: it flags mechanical latency and form factor as open problems. That is why we keep actuation off the capsule and exploit slow biological timescales so latency does not matter.
On the diagnostic side, Multi-cohort analysis of colorectal cancer metagenome identified altered bacteria across populations and universal bacterial markers matters for its payload size. A seven-species marker set reached AUC 0.80. A clinically useful payload may therefore be tens of bytes, which is why we argue rate is the wrong objective. Information Transmission in Quorum Sensing for Gut Microbiome models gut signaling as a low-pass channel dominated by environmental noise. This supports treating microbial readings as delay-tolerant. We do not claim the capsule reads quorum-sensing molecules; that remains the sensor designer's problem.
Research Program
Channel and trajectory simulator. We will build a simulator that models the capsule's path through the gut as a sequence of segments with dwell times. Each segment gets a depth- and tissue-dependent path-loss term and a tumbling-driven small-scale fading term. Receivers sit on a belt around the torso.
No in-body channel dataset or tissue model appears in the file, so this simulator is our own construction. We will calibrate it in a tissue-mimicking liquid phantom before trusting any comparison.
Optimization problem. The controller chooses three things: belt antenna positions, which are continuous for movable elements and discrete pinch locations on a waveguide; receive combining; and the capsule's transmit schedule and power. The objective is expected capsule energy per delivered reading. The constraints are a delivery deadline $T$, the time within which a reading must arrive, and a per-transmission power cap standing in for the tissue-heating limit. We set $T$ from minutes to hours to span plausible microbial change rates.
Channel estimation and actuation. Receiver positions are updated only from uplink pilots already present in the capsule's packets, so the capsule spends no extra energy on channel estimation. Actuation delay is charged against the deadline.
Hardware choice. We treat pinching antennas as the long-range option and movable elements as the short-range option. We state openly that a dielectric waveguide at the low-GHz and sub-GHz bands typical of in-body links may be impractically thick. Part of the program is therefore to determine which mechanism, if either, fits a wearable belt.
Diagnostic value. To connect the link to diagnostic value, we pass simulated delivery failures and quantization through real microbiome classifiers. A reading that misses its deadline is treated as missing.
Baselines
Naive reference. A single fixed on-body receive patch with a capsule that transmits every reading immediately at the power needed to close the link. This sets the energy scale every other scheme is measured against.
Strongest competitor. A fixed on-body array with the same number of receive chains and the same belt area as the movable design. It uses antenna selection and maximal-ratio combining, plus opportunistic scheduling: the capsule buffers readings and sends them when pilot-estimated channel quality crosses a threshold, subject to the same deadline $T$. This combines the two strongest cheap answers, spatial diversity from many elements and time diversity from delay tolerance. It gets the identical simulator, trajectories, power cap and deadline. Its selection threshold is tuned with the same search budget as the movable design's controller. We also run a denser fixed array with as many elements as the movable belt has candidate positions. That version corresponds to the "switching antenna" class in the flexible-position MIMO taxonomy and tests whether movement is just a cheaper way to get selection.
Reported numbers exist only for terrestrial settings: 2.36 versus 1.44 bps/Hz minimum user rate in the movable-antenna base-station paper, and 49.07% rate and 62.69% energy-efficiency gains over fixed antennas in the near-field fluid antenna paper. There is no reported result found on Lacuna for any in-body or on-body capsule link, fixed or movable. Our first deliverable is therefore the baseline number itself.
What counts as winning. Beating the strongest competitor means at least a 3 dB median reduction in capsule energy per delivered reading at matched deadline and outage. The gain must also concentrate in deep segments. A uniform gain would suggest a simulator artifact rather than the large-scale mechanism.
Ablations. - Movement without delay tolerance, which asks whether the two effects add up. - Movement with oracle versus pilot-estimated channel knowledge. - Movable belt with only sub-wavelength travel versus tens-of-centimeters travel. This isolates the claim that large-scale repositioning, not fade escape, drives the gain.
Diagnostic baselines. The reference points are the published numbers from the full data: AUC 0.80 for the seven-species support vector machine on the 526-sample colorectal cancer cohort, and AUC 0.805 for the ensemble on the Halfvarson inflammatory bowel disease cohort from Learning a deep language model for microbiomes: The power of large scale unlabeled microbiome data. We report how far each link scheme degrades AUC at a fixed capsule energy budget.
Why not just…?
Why not just keep stool sequencing? For population diagnosis it works: AUC 0.80, rising to 0.88 with age and BMI, in the multi-cohort colorectal study. This proposal does not claim to beat that accuracy. The capsule is justified only for questions stool cannot answer: location along the gut and changes within hours. If the diagnostic experiment shows that tolerable readings require payloads or sampling rates the link cannot deliver, the capsule premise weakens regardless of antennas, and we would say so.
Why not just add more fixed antennas and wait for a good moment? This is the real threat. It is why the strongest baseline combines exactly these two ideas rather than being a strawman. Delay tolerance removes much of the case for escaping fast fades, which is what most movable-antenna papers sell. Our case rests only on persistent, location-dependent loss. The travel-range ablation and the segment-wise breakdown settle it. If the dense fixed array plus scheduling comes within 1 dB of the movable belt, we conclude that movement is not worth its mechanics on the body and report a negative result.
Why not just raise capsule power or use digital beamforming? Raising power is capped by tissue heating and battery volume, so it is exactly the quantity we minimize rather than a free knob. Receive-side digital combining is already inside the strongest baseline. The pinching-antenna perspective paper found that repositioning nearly doubled sum rate over zero-forcing in a room. We do not assume this transfers to the body; the baseline comparison tests it directly.
Proposed Experiments
Experiment 1: simulation study. We sweep deadline, belt size, number of receive chains and gut-segment depth. For each scheme we report median and 95th-percentile capsule energy per delivered reading and outage fraction. We also plot the gain against segment depth. The decisive check is whether that gain rises with depth and survives the scheduling baseline.
Experiment 2: phantom validation. We build a liquid tissue phantom with a tumbling transmitter on a programmable path and a prototype belt with one motorized slider. We measure whether the simulator's ranking of schemes holds on hardware. A ranking reversal would stop the program at this stage.
Experiment 3: diagnostic study. We replay the colorectal cancer and Halfvarson feature vectors through each simulated link at equal capsule energy and report AUC relative to 0.80 and 0.805. This answers the question clinicians care about: at the same battery, which receiver design loses the least diagnostic accuracy.
References
- Multiuser Communications with Movable-Antenna Base Station Via Antenna Position Optimization
- Movable Antenna Enabled Integrated Sensing and Communication
- Joint Beamforming and Antenna Design for Near-Field Fluid Antenna System
- Pinching-Antenna Systems (PASS): A Tutorial
- Flexible-Antenna Systems: A Pinching-Antenna Perspective
- Delay Minimization in Pinching-Antenna-Enabled NOMA-MEC Networks
- Flexible-Position MIMO for Wireless Communications: Fundamentals, Challenges, and Future Directions
- Multi-cohort analysis of colorectal cancer metagenome identified altered bacteria across populations and universal bacterial markers
- Information Transmission in Quorum Sensing for Gut Microbiome
- Learning a deep language model for microbiomes: The power of large scale unlabeled microbiome data