Rice run coding reduced mean framed two-bit rank bytes by 60.3 percent on later pulses in eight SEEG sessions

We tested lossless transport of two-bit SEEG rank symbols in OpenNeuro ds004696, version 1.0.1. We used eight people, 86 selected sites, and 860 later response and control buffers. Each buffer had 983 samples. Early pulses selected one Rice parameter per site. Each frame included 11 bytes for fields and an error check. The encoder used fixed packing when run coding was not smaller. Mean framed bytes fell from 257.00 to 102.11 across people, a reduction of 60.27 percent. The descriptive 95 percent person bootstrap interval was 55.67 to 64.97 percent. All eight people had savings. Every decoded rank symbol and integer response score was equal to its input. Mean AUC remained 0.91951. Eight shorter frames per buffer retained 44.44 percent savings. Sample permutations removed the complete-frame savings. The package preserves the R1 frame specification and proof, compact voltage inputs, all later frames, figures, source hashes, and verification records. The GPL reference codec and analysis programs are excluded. A separate decoder and a full rerun passed. One new source audit matched exactly. Lossless recovery applies to rank symbols. Voltage precision was already removed. Ranking still needs the complete buffer through 500 ms. This offline test does not establish device energy savings or clinical gain.

SEEG · lossless compression · Rice coding · rank symbols · neural signal transport · intracranial EEG