Rationale:
Animal studies are an essential component of epilepsy research, wherein assessing efficacy of pharmaceutical and/or therapeutic strategies often requires quantification of seizure burden. Most commonly, such indices are determined though electroencephalogram (EEG) recordings, which is expensive and labor-intensive. Moreover, seizure yield can be very low and/or variable, so animals may need to be monitored for weeks on end to acquire an adequate sample of events, opening the door to compromised signal quality and animal mortality. Therefore, technologies that decrease user workload and facilitate high-throughput experimentation are of great value to the research community. Therefore, we sought to apply piezoelectric sensor technology to accomplish non-invasive seizure detection.
Method:
Non-invasive seizure detection was accomplished using piezoelectric motion sensors (Signal Solutions, LLC; Lexington, KY). Time-frequency analysis of the signal generated from these sensors can give insight into animal behavior (activity, breath rate, etc.). Here, measures of signal properties were used to detect epileptic events, which were verified through video and/or EEG records. Initially, mice (n=16) and rats (n=8) were treated with pilocarpine i.p. to induce status epilepticus and placed in recording cages, where a 12-week record of video and piezo data were collected. Data were processed and screened for seizures on a weekly basis. Thereafter, a subset of animals (6 mice, 2 rats) with the highest confirmed seizure yield were instrumented with EEG/EMG headmounts, and monitored for an additional four weeks, providing simultaneous EEG, EMG, piezo, and video recordings. Upon completion, accuracy of EEG- and piezo-based detection methods were compared and used to guide algorithm refinement.
Results:
Piezo-based seizure detection was able to identify 90-100% of EEG-verified seizures, but with low precision. However, secondary sifting of detections based on templates of detection feature dynamics at seizure onset resulted in the retention of a modest proportion (80%) of seizures with improved (1-in-10) positive detection rate. This translates to a reduction in the time required to screen a week of data to a mere 5 hours (96% savings). Nearly half of all seizures can be detected though 1 hour of review.
Conclusion:
This piezo approach shows great promise as a means for high-throughput screening of epilepsy models, with onset of detection closely correlated with EEG seizure onset. While these results focus on tonic-clonic seizures, we have also applied this method to other seizure models including absence seizures. Further development will include a more thorough investigation of these other cohorts and algorithm refinement to improve detection specificity. Overall, this approach could fill a great need in the research community and provide a convenient method of screening seizures in rodent models in a more cost- and resource-efficient manner.
Funding:
:This work was supported by NIH Grant No. R41 NS107148 and by a seed grant from EpiC, the University of Kentucky Epilepsy Research Center.
Disclosure: K. Donohue and B. O'Hara have ownership stake in Signal Solutions, LLC, the manufacturer of the sensors used in this study.
FIGURES
Figure 1