No abstract

Recurrent Neural Networks (RNNs) with sophisticated units that implement a gating mechanism have emerged as powerful technique for modeling sequential signals such as speech or electroencephalography (EEG). The latter is the focus on this paper. A significant big data resource, known as the TUH EEG Corpus (TUEEG), has recently become available for EEG research, creating a unique opportunity to evaluate these recurrent units on the task of seizure detection. In this study, we compare two types of recurrent units: long short-term memory units (LSTM) and gated recurrent units (GRU). These are evaluated using a state of the art hybrid architecture that integrates Convolutional Neural Networks (CNNs) with RNNs. We also investigate a variety of initialization methods and show that initialization is crucial since poorly initialized networks cannot be trained. Furthermore, we explore regularization of these convolutional gated recurrent networks to address the problem of overfitting. Our experiments revealed that convolutional LSTM networks can achieve significantly better performance than convolutional GRU networks. The convolutional LSTM architecture with proper initialization and regularization delivers 30% sensitivity at 6 false alarms per 24 hours.

Interpretation of electroencephalogram (EEG) signals can be complicated by obfuscating artifacts. Artifact detection plays an important role in the observation and analysis of EEG signals. Spatial information contained in the placement of the electrodes can be exploited to accurately detect artifacts. However, when fewer electrodes are used, less spatial information is available, making it harder to detect artifacts. In this study, we investigate the performance of a deep learning algorithm, CNN-LSTM, on several channel configurations. Each configuration was designed to minimize the amount of spatial information lost compared to a standard 22-channel EEG. Systems using a reduced number of channels ranging from 8 to 20 achieved sensitivities between 33% and 37% with false alarms in the range of [38, 50] per 24 hours. False alarms increased dramatically (e.g., over 300 per 24 hours) when the number of channels was further reduced. Baseline performance of a system that used all 22 channels was 39% sensitivity with 23 false alarms. Since the 22-channel system was the only system that included referential channels, the rapid increase in the false alarm rate as the number of channels was reduced underscores the importance of retaining referential channels for artifact reduction. This cautionary result is important because one of the biggest differences between various types of EEGs administered is the type of referential channel used.

Abuse of synthetic psychostimulants like synthetic cathinones has risen in recent years. 3,4-Methylenedioxypyrovalerone (MDPV) is one such synthetic cathinone that demonstrates a mechanism of action similar to cocaine. Compared to cocaine, MDPV is more potent at blocking dopamine and norepinephrine reuptake and is readily self-administered by rodents. The present study compared the rewarding and reinforcing properties of MDPV and cocaine using systemic injection dose-response and self-administration models. Fifty kilohertz ultrasonic vocalizations (USVs) were recorded as an index of positive affect throughout experiments. In Experiment 1, MDPV and cocaine dose-dependently elicited 50-kHz USVs upon systemic injection, but MDPV increased USVs at greater rates and with greater persistence relative to cocaine. In Experiment 2, latency to begin MDPV self-administration was shorter than latency to begin cocaine self-administration, and self-administered MDPV elicited greater and more persistent rates of 50-kHz USVs versus cocaine. MDPV-elicited 50-kHz USVs were sustained over the course of drug load-up whereas cocaine-elicited USVs waned following initial infusions. Notably, we observed a robust presence of context-elicited 50-kHz USVs from both MDPV and cocaine self-administering rats. Collectively, these data suggest that MDPV has powerfully rewarding and reinforcing effects relative to cocaine at one-tenth doses. Consistent with prior work, we additionally interpret these data in supporting that MDPV has significant abuse risk based on its potency and subjectively positive effects. Future studies will be needed to better refine therapeutic strategies targeted at reducing the rewarding effects of cathinone analogs in efforts to ultimately reduce abuse liability.

Glycogen synthase kinase-3b (GSK3b) is a critical regulator of the balance between long-term depression and long-term potentiation which is essential for learning and memory. Our previous study demonstrated that GSK3b activity is highly induced during cocaine memory reactivation, and that reconsolidation of cocaine reward memory is attenuated by inhibition of GSK3b. NMDA receptors and protein phosphatase 1 (PP1) are activators of GSK3b. Thus, this study investigated the roles of NMDA receptor subtypes and PP1in the reconsolidation of cocaine contextual reward memory. Cocaine contextual memories were established and evaluated using cocaine conditioned place preference methods. The regulation of GSK3b activity in specific brain areas was assessed by measuring its phosphorylation state using immunoblot assays. Mice underwent cocaine place conditioning for 8 days and were tested for place preference on day 9. Twenty-four hours later, mice were briefly confined to the compartment previous paired with cocaine to reactivate cocaine-associated memories. Administration of the GluN2A-and GluN2B-NMDA receptor antagonists, NVP-AAM077 and ifenprodil, respectively, immediately following recall abrogated an established cocaine place preference, while preventing the activation of GSK3b in the amygdala, nucleus accumbens, and hippocampus during cocaine memory reactivation. PP1 inhibition with okadaic acid also blocked the activation of GSK3b and attenuated a previously established cocaine place preference. These findings suggest that the dephosphorylation of GSK3b that occurred upon activation of cocaine-associated reward memories may be initiated by the activation of PP1 during the induction of NMDA receptor-dependent reconsolidation of cocaine mnemonic traces. Moreover, the importance of NMDA receptors and PP1 in reconsolidation of cocaine memory makes them potential therapeutic targets in treatment of cocaine use disorder and prevention of relapse.