Abstracts

Novel antiseizure therapies or delivery modalities are being actively developed and tested. One of the most desirable outcomes of these efforts would be the development of a fully implantable device for automated detection, quantitative analysis and blockage of seizures. We developed a new Percentile Tracking Filter (PTF) that replaces the digital order-statistic filter in the original the Osorio-Frei Algorithm (OFA) in an analog implementation, demonstrate the feasibility of its implementation for real-time quantitative seizure detection and analysis, and assess its performance. The analog implementation of the order statistic filter, a key operation of the OFA, facilitates full device implantation due to its low power consumption, which decreases battery size requirements or frequency of recharging. The original OFA comprises an FIR band pass filter, a squaring operation and an order statistic filter that are performed digitally. These operations were carried out in analog by an analog band pass filter matching the pass band and sharpness of the FIR, an absolute value circuit in place of the squaring operation, and the PTF circuit to perform the order statistic filtering. The performance of the analog implementation was evaluated using 20 ECoG segments, each about 500 seconds long containing a seizure beginning after about 300s. The output of the analog circuit was digitized, normalized and compared with the output of the validated digital OFA. The analog implementation detected all the seizures without false positive detections. Sample-to-sample comparison of the performance of the analog circuit and the digital SDA are in Table 1. Sensitivity, specificity, and kappa were obtained by classifying each output sample during a seizure or a non-seizure segment as above or below a threshold, which determined TP, FP, TN and FN. The speed of detection was determined at the generic threshold and duration constraint of the validated SDA. [table1] The results showed that the performance of the analog implementation is nearly identical to the validated SDA. This analog implementation can be realized into an application specific integrated circuit (ASIC) that may be used in a fully implantable device for seizure monitoring, warning, and treatment, increasing acceptability. Preliminary analysis showed that this ASIC consumes very little power compared to a digital device and can last for several years without the need for recharging or replacing its battery. The savings in power result from eliminating data sorting and storing, which are essential operations in a digital order statistic filter. (Supported by Flint Hills Scientific, LLC)