Figure 1.

Combining long-term epileptiform records from an implanted device to smartphone behavioral data

(A) Eight subjects were implanted with the RNS neurostimulator and NeuroPace cortical strip and depth leads (NeuroPace, Inc.) (see also Table S1). The location of the strip and the depth lead varied substantially from one subject to another, and the figure is for illustrative purposes. Detector activity from subject #1 was recorded for 185 days. The 4 leads tracked distinct activity patterns measured by two leads.

(B) An illustration of the joint interval distribution that is used to capture the short-term behavioral dynamics on the smartphone by describing the relationship between an (say k) inter-touch interval (ITI) and the subsequent interval (k+1). Note that different parts of this feature space capture different forms of behavioral dynamics. For instance, the lower left corner represents rapid interactions and the diagonal represent rhythmic output where the subsequent intervals are the same. The probability distribution accumulated over the entire smartphone recording period is shown for each subject. The daily probability distributions are shown for subject #1.

(C) The periodograms of the neural detector activity for each active detector from all the 8 patients used for this study. The periodograms were generated using the wavelet transform method described in (Baud et al., 2018). “L” corresponds to the lead number, and “D” corresponds to the detector number.

(D) The power of 24 h cycles across the joint interval distribution. The non-significant power values according to shuffled bootstraps are masked (see also Videos S1, S2, S3, S4, S5, S6, S7, and S8 and Figure S1).