In five SEEG subjects, 12-bit quantization retained 88% high-gamma channel fidelity, while 10 bits retained 19%
We simulated analog-to-digital converter quantization on partial recordings from five subjects in OpenNeuro ds004752. At a +/-5 mV full scale, 12 bits (2.4 uV step) preserved 70-150 Hz log-power fidelity (Pearson r >= 0.9) in 151/171 bipolar channels (88%); 10 bits (9.8 uV step) did so in 33/171 (19%). At 10 bits, all-band decoding AUC fell by 0.093 and 0.123 in two low-amplitude subjects. At 8 bits, median 70-150 Hz AUC change was -0.098. A white quantization-noise model gave median required depths of 11.6 bits for 70-150 Hz and 9.8 bits for 30-70 Hz. These descriptive results from five partial sessions suggest a 1-2.5 uV step (12-13 bits at +/-5 mV) for high-gamma SEEG band-power features; bands below 30 Hz tolerated 8-9 bits. The study used recorded signals that already contained amplifier noise, tested one decoder and task contrast, and did not make inferential tests across subjects.
SEEG · intracranial EEG · analog-to-digital conversion · quantization · band power · high gamma · brain-computer interface · working memory · signal fidelity · decoding
References
3- usesFour SEEG subjects decoded task phase better than memory load or probe match across reference schemes sci:sha256:341751ce85385bec3d9952e1c6f69998b53f06f3664b87da2c409f44289ec276
We use the same dataset, trial timing (fixation 0-1 s, maintenance 3-6 s) and fixation-versus-maintenance decoding design, and add ADC quantization as the tested variable.
- usesdoi:10.18112/openneuro.ds004752.v1.0.1
The analyzed EDF prefixes came from OpenNeuro ds004752; the archived provenance records the source URL, ETags, and SHA-256 hashes.
- doi:10.7554/eLife.78677
Cited by
2 publicationsPublic catalog entries on Sciscape.
- supportsA simulated SEEG ripple trigger retained held-out F1 at 12 bits and 1000 Hz in 20 patients
Also finds 12 bits at +/-5 mV sufficient and 10 bits not sufficient, here for ripple detection instead of band-power task decoding.
- supersedesIn five SEEG subjects, 12-bit quantization largely preserved high-gamma features, while 10-bit quantization reduced them
Corrects one wrong sentence: the subjects that lost decoding at 10 bits did not have the smallest signals. Data, code and all other results are unchanged.