Ultrasound Channel Compression: Define Fidelity Before Saving Bandwidth

By ALEVA | California, USA

Engineering question: What must remain unchanged—or explicitly bounded—when ultrasound channel data are compressed before beamforming?

Two different promises

Lossless compression and task-preserving compression are different acceptance contracts. A lossless decoder must reproduce the agreed digital samples exactly. A lossy approach may instead trade reconstruction error against bitrate; a visually plausible B-mode image does not establish sample identity or suitability for every downstream task.

A 2018 peer-reviewed study by Kim and colleagues evaluated BL universal coding on pre-beamformed, beamformed and IQ ultrasound data. Its proposed FPGA/USB implementation was future work, not a demonstrated production FPGA result. The date matters: this is foundational work, not a new 2026 device announcement. [1]

Federici and colleagues' August 2026 journal article describes a learned channel-data codec using variational inference and entropy coding, with a scale hyperprior transmitted as side information. Its indexed abstract reports carotid-channel training and evaluation on in-vivo and phantom data. These findings motivate evaluation; they do not guarantee performance on a different probe, acquisition sequence or measurement task. [2]

Count the complete path

For a fixed acquisition, the uncompressed sample payload is R = C × N × b × E bits/s, where C is active channels, N samples per channel per event, b stored bits per sample and E acquisition events per second. This is a dimensional payload calculation, not a link-throughput measurement. Include headers, padding, codec metadata and transport overhead separately.

Our proposed OEM acceptance record has three parts:

• Fidelity: For lossless operation, compare every decoded sample with the reference. For lossy operation, predefine signal-error limits and task-specific checks, such as weak-target reconstruction and Doppler estimates; do not rely only on visual inspection.

• System cost: Measure encoding, transfer and decoding together, including memory use, latency variation, sustained power and fallback behavior. Fewer transmitted bits do not, by themselves, establish lower end-to-end latency.

• Reproducibility: Preserve probe, acquisition settings, sample format and codec/model versions. Test acquisition conditions outside the tuning set and retain an uncompressed reference path.

OEM/ODM decision

Choose the fidelity contract before selecting the codec. Keep a compression option only if its measured system benefit meets that contract on the intended data and hardware. This is a proposed evaluation framework—not an ALEVA codec implementation, hardware benchmark or clinical-validation claim.

Review limitation: The 2026 source was reviewed through its indexed abstract, not a full-paper methodological appraisal. No compression-ratio or clinical-equivalence claim is transferred to an ALEVA product.

References

1. Kim JH et al. Real-Time Lossless Compression Algorithm for Ultrasound Data Using BL Universal Code. Sensors 18(10):3314. Published October 2, 2018. DOI: 10.3390/s18103314. Open full text.

2. Federici B, Mischi M, van Sloun RJG. Variational Ultrasound Channel Data Compression. IEEE Transactions on Ultrasonics 73(8):927–938, August 2026. DOI: 10.1109/TUSON.2026.3693915. Indexed abstract.