Deepak Madhavan

Classification of patterns of EEG synchronization for seizure prediction

OBJECTIVE Research in seizure prediction from intracranial EEG has highlighted the usefulness of bivariate measures of brainwave synchronization. Spatio-temporal bivariate features are very high-dimensional and cannot be analyzed with conventional statistical methods. Hence, we propose state-of-the-art machine learning methods that handle high-dimensional inputs. METHODS We computed bivariate features of EEG synchronization (cross-correlation, nonlinear interdependence, dynamical entrainment or wavelet synchrony) on the 21-patient Freiburg dataset. Features from all channel pairs and frequencies were aggregated over consecutive time points, to form patterns. Patient-specific machine learning-based classifiers (support vector machines, logistic regression or convolutional neural networks) were trained to discriminate interictal from preictal patterns of features. In this explorative study, we evaluated out-of-sample seizure prediction performance, and compared each combination of feature type and classifier. RESULTS Among the evaluated methods, convolutional networks combined with wavelet coherence successfully predicted all out-of-sample seizures, without false alarms, on 15 patients, yielding 71% sensitivity and 0 false positives. CONCLUSIONS Our best machine learning technique applied to spatio-temporal patterns of EEG synchronization outperformed previous seizure prediction methods on the Freiburg dataset. SIGNIFICANCE By learning spatio-temporal dynamics of EEG synchronization, pattern recognition could capture patient-specific seizure precursors. Further investigation on additional datasets should include the seizure prediction horizon.

Comparing SVM and convolutional networks for epileptic seizure prediction from intracranial EEG

Recent research suggests that electrophysiological changes develop minutes to hours before the actual clinical onset in focal epileptic seizures. Seizure prediction is a major field of neurological research, enabled by statistical analysis methods applied to features derived from intracranial Electroencephalographic (EEG) recordings of brain activity. However, no reliable seizure prediction method is ready for clinical applications. In this study, we use modern machine learning techniques to predict seizures from a number of features proposed in the literature. We concentrate on aggregated features that encode the relationship between pairs of EEG channels, such as cross-correlation, nonlinear interdependence, difference of Lyapunov exponents and wavelet analysis-based synchrony such as phase locking. We compare L1-regularized logistic regression, convolutional networks, and support vector machines. Results are reported on the standard Freiburg EEG dataset which contains data from 21 patients suffering from medically intractable focal epilepsy. For each patient, at least one method predicts 100% of the seizures on average 60 minutes before the onset, with no false alarm. Possible future applications include implantable devices capable of warning the patient of an upcoming seizure as well as implanted drug-delivery devices.

Expression of receptor protein tyrosine phosphatase alpha mRNA in human prostate cancer cell lines.

Receptor protein tyrosine phosphatase α (RPTPα) is transmembrane protein phosphatases, and has been proposed to be involved in the differentiation of the neuronal system. In the present study, we demonstrated the expression of RPTPα mRNA in several normal human tissues. We further investigated the regulation of expression of RPTPα mRNA in epithelial cells utilizing three commercially available human prostate cancer cell lines LNCaP, PC-3 and DU145. This is because these cells exhibit different levels of differentiation, defined by the expression of a tissue-specific differentiation antigen, prostatic acid phosphatase (PAcP), and their androgen sensitivity. LNCaP cells express PAcP and are androgen-sensitive cells, while PC-3 and DU145 cells do not express PAcP and are androgen-insensitive cells. Northern blot analyses revealed that, in LNCaP cells, fetal bovine serum (FBS) and 5α-dihydrotestosterone (DHT) down-regulates RPTPα mRNA expression, similar to the effect on PAcP. Contrarily, FBS up-regulated the RPTPα mRNA level in PC-3 and DU145 cells. In LNCaP cells, sodium butyrate inhibited cell growth and up-regulated RPTPα as well as PAcP mRNA expression. Although, sodium butyrate also inhibited the growth of PC-3 and DU145 cells, the level of RPTPα mRNA was decreased in PC-3, while increased in DU145 cells. Thus, data taken together indicate that the expression of RPTPα is apparently regulated by a similar mechanism to that of PAcP in LNCaP cells.

CD40 Ligand Deficiency: Neurologic Sequelae With Radiographic Correlation

Patients with CD40 ligand deficiency are susceptible to central nervous system infections, but to date the neurologic progression or long-term outcome of central nervous system complications have not been reported in detail. Characterizing the central nervous system complications of immune deficiencies can lead to the identification of new pathogens. For this study, clinical data were reviewed on patients with both CD40 ligand deficiency and neurodegeneration, identified from a larger cohort of 31 patients. Five patients had progressive neurologic and cognitive decline in the absence of clinical signs of acute fulminant encephalitis, with anatomic brain abnormalities and high mortality (60%). Despite multiple evaluations, no pathogens were identified in four patients, all of whom were on standard intravenous immunoglobulin therapy at illness presentation. This clinical phenotype of progressive decline without acute fulminant encephalitis is similar to chronic enteroviral encephalitis in X-linked agammaglobulinemia, another condition with severe humoral immune defects. Whether infection secondary to subtherapeutic levels of central nervous system immunoglobulin G (IgG), inadequately protective levels of serum IgG, or impaired CD40 ligand-dependent IgG-independent antiviral responses contributed remains undetermined. Emerging gene-chip techniques applied in patients with primary immune deficiencies may identify heretofore unknown viruses. Prospective neurocognitive and evaluation of patients with CD40 ligand deficiency may identify affected patients before overt clinical signs appear.

Adenosine has two faces: Regionally dichotomous adenosine tone in a model of epilepsy with comorbid sleep disorders

OBJECTIVE Adenosine participates in maintaining the excitatory/inhibitory balance in neuronal circuits. Studies indicate that adenosine levels in the cortex and hippocampus increase and exert sleep pressure in sleep-deprived and control animals, whereas in epilepsy reduced adenosine tone promotes hyperexcitability. To date, the role of adenosine in pathological conditions that result in both seizures and sleep disorders is unknown. Here, we determined adenosine tone in sleep and seizure regulating brain regions of Kv1.1 knockout (KO) mice, a model of temporal epilepsy with comorbid sleep disorders. METHODS 1) Reverse phase-high performance liquid chromatography (RP-HPLC) was performed on brain tissue to determine levels of adenosine and adenine nucleotides. 2) Multi-electrode array extracellular electrophysiology was used to determine adenosine tone in the hippocampal CA1 region and the lateral hypothalamus (LH). RESULTS RP-HPLC indicated a non-significant decrease in adenosine (~50%, p = 0.23) in whole brain homogenates of KO mice. Regional examination of relative levels of adenine nucleotides indicated decreased ATP and increased AMP in the cortex and hippocampus and increased adenosine in cortical tissue. Using electrophysiological and pharmacological techniques, estimated adenosine levels were ~35% lower in the KO hippocampal CA1 region, and 1-2 fold higher in the KO LH. Moreover, the increased adenosine in KO LH contributed to lower spontaneous firing rates of putative wake-promoting orexin/hypocretin neurons. INTERPRETATION This is the first study to demonstrate a direct correlation of regionally distinct dichotomous adenosine levels in a single model with both epilepsy and comorbid sleep disorders. The weaker inhibitory tone in the dorsal hippocampus is consistent with lower seizure threshold, whereas increased adenosine in the LH is consistent with chronic partial sleep deprivation. This work furthers our understanding of how adenosine may contribute to pathological conditions that underlie sleep disorders within the epileptic brain.

Anterior corpus callosotomy in patients with drug-resistant epilepsy: Invasive EEG findings and seizure outcomes

Corpus callosotomy (CC) is used in patients with drug-resistant seizures who are not candidates for excisional surgery and failed neurostimulation. We examined ictal scalp and intracranial electroencephalogram (iEEG) recordings in 16 patients being evaluated for anterior CC alone or CC in combination with focal resection, to determine the role of the iEEG in predicting postoperative seizure outcomes. In our cohort, CC improved generalized atonic seizures and focal seizures with impaired awareness but did not alter outcomes for generalized tonic–clonic or tonic seizures. Invasive EEG prior to CC did not refine the prediction of postsurgical seizure outcomes in patients with inconclusive scalp EEG.