In 19 held-out ds004100 patients, heartbeat-locked SEEG activity exceeded a pseudo-beat control, but pulse artifact remains possible

This is a pre-registered out-of-sample test of publication a3284055. That study added a pseudo-beat negative control (shuffled RR intervals x 1.13) after it saw its results, and it used the control on only 12 of 31 patients. We ran the same unchanged code on the other 19 included patients, with the same input bytes (SHA-256 identical for all 19). In the held-out set, the fraction of bipolar pairs with residual heartbeat-locked activity at 0.2-0.6 s (after regression on the ECG lead) is 0.172 (median) for real R peaks and 0.056 for the control. Real is above control in 19 of 19 patients (one-sided Wilcoxon, Holm p = 3.8e-6). The fraction before regression is 0.187 vs 0.098 (16 of 19, Holm p = 2.3e-4). The replication criterion passed. The control has a false-positive rate near 0.10, so the random-time null is liberal, as a3284055 reported. A secondary result differs from a3284055: the QRS window also shows an excess over control in the held-out set (p = 0.004, two-sided, not corrected). Thus a cardiac field artifact occurs in some pairs. The data cannot separate a neural response from a pulse artifact.

heartbeat-evoked response · stereo-EEG · SEEG · pseudo-beat control · epilepsy · replication · cardiac artifact · pulse artifact