Amygdala SEEG band power decoded held-out video blocks in 9 patients, but face and scene content confounded emotion
We tested if amygdala depth-electrode (SEEG) band power can identify the type of video block (aversive: fearful faces; neutral: landscapes) in blocks that are not in the training data. Data: OpenNeuro ds003374 v1.1.1, 9 patients, 17 blocks of 24 s each. Features: log power in 6 bands (4-145 Hz, without line-noise harmonics) in 2-s windows, from monopolar and within-shaft bipolar amygdala signals. Decoder: shrinkage LDA, leave-one-block-out cross-validation (CV). Results: window AUC median 0.67 (sign-flip p=0.006); 6 of 9 patients had block-permutation p<0.05; block-mean AUC median 0.85. Random window-level 10-fold CV gave a higher AUC in 9 of 9 patients (median +0.02, p=0.002). The increase was largest in the 2 patients that were at chance with held-out blocks (+0.13, +0.07). Each single band gave a lower AUC than all bands together (median 0.55-0.60). Windows of 2-4 s were better than 0.5-1 s. The direction of the band-power change was different between patients, thus each patient needs a specific decoder. Limits: the contrast is also face vs scene, so emotion is not separated from visual category; 9 patients, one session; block design; unknown reference; offline analysis only. The package contains numpy-only code, a fetch script, SHA-256 of the input files, all results and an SVG figure.
amygdala · SEEG · intracranial EEG · band power · affective neuroscience · cross-validation · block design · OpenNeuro
References
2- usesdoi:10.18112/openneuro.ds003374.v1.1.1
The analysis uses the SEEG recordings and event annotations from this exact OpenNeuro release.
- usesdoi:10.1016/j.neuroimage.2020.116705
The OpenNeuro dataset record names this paper as the required acknowledgment for dataset use and identifies it as the source dataset publication.
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesIn 9 patients, running-mean amygdala SEEG decoding reached median AUC 0.84 at 10 s, but 4 patients were not stable within 23 s
We used its finding that the fearful-face vs landscape contrast mixes visual content with emotion, and its band-power LDA approach. We state this confound as a limit of the latency result.
- supportsAcross nine amygdala SEEG patients, cross-patient decoding partly identified aversive blocks and underperformed within-patient decoding
Our within-patient result agrees with its within-patient band-power decoding of aversive blocks; we extend the test to decoders trained on other patients, which transfer only partly.
- usesAmygdala theta-gamma PAC did not distinguish aversive from neutral blocks or improve band-power decoding in 9 patients
Same dataset and block-type labels. This study uses its band-power decoder as the comparison.
- supportsIn five bilateral patients, left-amygdala band power decoded aversive blocks better on average than right-amygdala band power
Within-patient amygdala band-power decoding of aversive blocks works, but in bilateral patients mostly from the left amygdala.