Bioelectronics Laboratory
  • Preclinical
  • Published 2026

Real-time wireless telemetry for miniature implants

A 2.4 GHz radio protocol that streams neural data from millimeter-scale implants without gaps, on commercial hardware.

Block diagram of the implant (data acquisition of biopotential, inertial, temperature and humidity signals; main system-on-chip with a 2.4 GHz antenna; power conversion and regulation; stimulation engine), a mouse in a wireless power transfer cage linked to a USB transceiver and computer, and photographs of the implant unhoused on a fingertip and housed, with an 18.05 mm scale bar
The implant system used to test NRTP, with the implant shown before and after housing. Full size Fig. 1 from Elgohary et al., Bioelectronic Medicine 12, 8 (2026), CC BY-NC-ND 4.0.

The problem

Neural recordings need sampling rates and resolution comparable to high-quality audio, streamed through tissue from an implant with a tiny power budget and an inefficient antenna. Bluetooth Low Energy is mature, but its throughput and robustness fall short under weak or noisy signal conditions.

Our approach

The Neural Real-Time Telemetry Protocol (NRTP): fixed-length packets, immediate acknowledgments, bounded retransmissions and a single radio channel, running on commercial 2.4 GHz hardware without custom chips.

NRTP kept zero data loss down to −75 dBm

Using identical hardware for both protocols, we swept received signal strength and compared throughput, data loss and current draw. NRTP sustained zero data loss down to −75 dBm, while Bluetooth Low Energy degraded below −55 dBm under interference, because its retries are deferred.

The link-margin advantage, up to about 23 dB at first loss and 11 dB at 0.5% loss, corresponds to roughly 3.2 times the range in air, about 2.5 cm more implant depth in tissue, or a lower transmit power for the same performance.

Interleaving turns lost packets into brief half-rate segments, not gaps

Splitting samples across alternating packets means a lost packet lowers the sampling rate briefly instead of leaving a gap in the recording. Overlapping data between packets improved robustness further, but required double the packet rate, which drew too much power for an implant.

The link is deliberately lopsided: our patent application describes configuring the implant to spend more than 90% of its airtime transmitting, because data flow mostly from the animal to the computer.

Diagram of numbered samples packed into radio packets: interleaving odd and even samples into separate packets; overlapping neighboring packets so each sample is sent twice; and interleaving followed by overlapping
How samples are packed into packets under each strategy tested: interleaving (A), overlapping neighboring packets (B), and interleaving followed by overlapping (C). Full size Fig. 3 from Elgohary et al., Bioelectronic Medicine 12, 8 (2026), CC BY-NC-ND 4.0.