Channel order reduced the mean size of independent 100 ms zlib packets for stored float32 SEEG in 35 stimulation records

We tested whether channel order reduces the size of lossless SEEG packets. We used 84 two-second windows from two fixed OpenNeuro datasets. The stimulation data had 35 records from 21 people. The rest data had seven records from seven people. Each independent 100 ms packet used zlib level 6, a 32-byte header, and raw words if compression increased size. In the stimulation data, channel order reduced the person mean from 18.568 to 17.565 bits per stored float32 word. The reduction was 5.40%. The paired 95% interval for the difference was [-1.137, -0.859] bits per word. All 35 record means were lower. In the rest data, the difference was -0.014 bits per word, with interval [-0.032, 0.002]. The rest interval includes zero. Channel order also gave smaller stimulation packets than the prior ten-mode codec. Median encode time on one host was 2.17 ms, compared with 27.24 ms for that codec. All 98,280 tested packets recovered exact source bytes. The package preserves data, source pins, code, tables, packet examples, and three figures. The result concerns transport of the released source words. It does not establish neural decoding accuracy, live stimulation control, or device energy savings.

SEEG · lossless compression · brain-computer interface · zlib · packet encoding · float32