Machine learning algorithm for predicting seizure control after temporal lobe resection using peri-ictal electroencephalography.

Brain resection is curative for a subset of patients with drug resistant epilepsy but up to half will fail to achieve sustained seizure freedom in the long term. There is a critical need for accurate prediction tools to identify patients likely to have recurrent postoperative seizures. Results from preclinical models and intracranial EEG in humans suggest that the window of time immediately before and after a seizure ("peri-ictal") represents a unique brain state with implications for clinical outcome prediction. Using a dataset of 294 patients who underwent temporal lobe resection for seizures, we show that machine learning classifiers can make accurate predictions of postoperative seizure outcome using 5 min of peri-ictal scalp EEG data that is part of universal presurgical evaluation (AUC 0.98, out-of-group testing accuracy > 90%). This is the first approach to seizure outcome prediction that employs a routine non-invasive preoperative study (scalp EEG) with accuracy range likely to translate into a clinical tool. Decision curve analysis (DCA) shows that compared to the prevalent clinical-variable based nomogram, use of the EEG-augmented approach could decrease the rate of unsuccessful brain resections by 20%.

Toward a brighter constellation: multiorgan neuroimaging of neural and vascular dynamics in the spinal cord and brain.

Significance: Pain comprises a complex interaction between motor action and somatosensation that is dependent on dynamic interactions between the brain and spinal cord. This makes understanding pain particularly challenging as it involves rich interactions between many circuits (e.g., neural and vascular) and signaling cascades throughout the body. As such, experimentation on a single region may lead to an incomplete and potentially incorrect understanding of crucial underlying mechanisms. Aim: We aimed to develop and validate tools to enable detailed and extended observation of neural and vascular activity in the brain and spinal cord. The first key set of innovations was targeted to developing novel imaging hardware that addresses the many challenges of multisite imaging. The second key set of innovations was targeted to enabling bioluminescent (BL) imaging, as this approach can address limitations of fluorescent microscopy including photobleaching, phototoxicity, and decreased resolution due to scattering of excitation signals. Approach: We designed 3D-printed brain and spinal cord implants to enable effective surgical implantations and optical access with wearable miniscopes or an open window (e.g., for one- or two-photon microscopy or optogenetic stimulation). We also tested the viability for BL imaging and developed a novel modified miniscope optimized for these signals (BLmini). Results: We describe "universal" implants for acute and chronic simultaneous brain-spinal cord imaging and optical stimulation. We further describe successful imaging of BL signals in both foci and a new miniscope, the "BLmini," which has reduced weight, cost, and form-factor relative to standard wearable miniscopes. Conclusions: The combination of 3D-printed implants, advanced imaging tools, and bioluminescence imaging techniques offers a coalition of methods for understanding spinal cord-brain interactions. Our work has the potential for use in future research into neuropathic pain and other sensory disorders and motor behavior.

Machine learning polysomnographically-derived electroencephalography biomarkers predictive of epworth sleepiness scale.

Excessive daytime sleepiness (EDS) causes difficulty in concentrating and continuous fatigue during the day. In the clinical setting, the assessment and diagnosis of EDS rely mostly on subjective questionnaires and verbal reports, which compromises the reliability of clinical diagnosis and the ability to robustly discern candidacy for available therapies and track treatment response. In this study, we used a computational pipeline for the automated, rapid, high-throughput, and objective analysis of previously collected encephalography (EEG) data to identify surrogate biomarkers for EDS, thereby defining the quantitative EEG changes in individuals with high Epworth Sleepiness Scale (ESS) (n = 31), compared to a group of individuals with low ESS (n = 41) at the Cleveland Clinic. The epochs of EEG analyzed were extracted from a large overnight polysomnogram registry during the most proximate period of wakefulness. Signal processing of EEG showed significantly different EEG features in the low ESS group compared to high ESS, including enhanced power in the alpha and beta bands and attenuation in the delta and theta bands. Our machine learning (ML) algorithms trained on the binary classification of high vs. low ESS reached an accuracy of 80.2%, precision of 79.2%, recall of 73.8% and specificity of 85.3%. Moreover, we ruled out the effects of confounding clinical variables by evaluating the statistical contribution of these variables on our ML models. These results indicate that EEG data contain information in the form of rhythmic activity that could be leveraged for the quantitative assessment of EDS using ML.

Transient gamma events delineate somatosensory modality in S1.

Gamma band activity localized to the primary somatosensory cortex (S1) in humans and animals is implicated in the higher order neural processing of painful and tactile stimuli. However, it is unclear if gamma band activity differs between these distinct somatosensory modalities. Here, we coupled a novel behavioral approach with chronic extracellular electrophysiology to investigate differences in S1 gamma band activity elicited by noxious and innocuous hind paw stimulation in transgenic mice. Like prior studies, we found that trial-averaged gamma power in S1 increased following both noxious and innocuous stimuli. However, on individual trials, we noticed that evoked gamma band activity was not a continuous oscillatory signal but a series of transient spectral events. Upon further analysis we found that there was a significantly higher incidence of these gamma band events following noxious stimulation than innocuous stimulation. These findings suggest that somatosensory stimuli may be represented by specific features of gamma band activity at the single trial level, which may provide insight to mechanisms underlying acute pain.

SARS-CoV-2 alters neural synchronies in the brain with more severe effects in younger individuals.

Coronavirus disease secondary to infection by SARS-CoV-2 (COVID19 or C19) causes respiratory illness, as well as severe neurological symptoms that have not been fully characterized. In a previous study, we developed a computational pipeline for the automated, rapid, high-throughput and objective analysis of electroencephalography (EEG) rhythms. In this retrospective study, we used this pipeline to define the quantitative EEG changes in patients with a PCR-positive diagnosis of C19 (n = 31) in the intensive care unit (ICU) of Cleveland Clinic, compared to a group of age-matched PCR-negative (n = 38) control patients in the same ICU setting. Qualitative assessment of EEG by two independent teams of electroencephalographers confirmed prior reports with regards to the high prevalence of diffuse encephalopathy in C19 patients, although the diagnosis of encephalopathy was inconsistent between teams. Quantitative analysis of EEG showed distinct slowing of brain rhythms in C19 patients compared to control (enhanced delta power and attenuated alpha-beta power). Surprisingly, these C19-related changes in EEG power were more prominent in patients below age 70. Moreover, machine learning algorithms showed consistently higher accuracy in the binary classification of patients as C19 versus control using EEG power for subjects below age 70 compared to older ones, providing further evidence for the more severe impact of SARS-CoV-2 on brain rhythms in younger individuals irrespective of PCR diagnosis or symptomatology, and raising concerns over potential long-term effects of C19 on brain physiology in the adult population and the utility of EEG monitoring in C19 patients.

Time-dynamic pulse modulation of spinal cord stimulation reduces mechanical hypersensitivity and spontaneous pain in rats.

Enhancing the efficacy of spinal cord stimulation (SCS) is needed to alleviate the burden of chronic pain and dependence on opioids. Present SCS therapies are characterized by the delivery of constant stimulation in the form of trains of tonic pulses (TPs). We tested the hypothesis that modulated SCS using novel time-dynamic pulses (TDPs) leads to improved analgesia and compared the effects of SCS using conventional TPs and a collection of TDPs in a rat model of neuropathic pain according to a longitudinal, double-blind, and crossover design. We tested the effects of the following SCS patterns on paw withdrawal threshold and resting state EEG theta power as a biomarker of spontaneous pain: Tonic (conventional), amplitude modulation, pulse width modulation, sinusoidal rate modulation, and stochastic rate modulation. Results demonstrated that under the parameter settings tested in this study, all tested patterns except pulse width modulation, significantly reversed mechanical hypersensitivity, with stochastic rate modulation achieving the highest efficacy, followed by the sinusoidal rate modulation. The anti-nociceptive effects of sinusoidal rate modulation on EEG outlasted SCS duration on the behavioral and EEG levels. These results suggest that TDP modulation may improve clinical outcomes by reducing pain intensity and possibly improving the sensory experience.

Pain phenotypes classified by machine learning using electroencephalography features.

Pain is a multidimensional experience mediated by distributed neural networks in the brain. To study this phenomenon, EEGs were collected from 20 subjects with chronic lumbar radiculopathy, 20 age and gender matched healthy subjects, and 17 subjects with chronic lumbar pain scheduled to receive an implanted spinal cord stimulator. Analysis of power spectral density, coherence, and phase-amplitude coupling using conventional statistics showed that there were no significant differences between the radiculopathy and control groups after correcting for multiple comparisons. However, analysis of transient spectral events showed that there were differences between these two groups in terms of the number, power, and frequency-span of events in a low gamma band. Finally, we trained a binary support vector machine to classify radiculopathy versus healthy subjects, as well as a 3-way classifier for subjects in the 3 groups. Both classifiers performed significantly better than chance, indicating that EEG features contain relevant information pertaining to sensory states, and may be used to help distinguish between pain states when other clinical signs are inconclusive.

Automated and rapid self-report of nociception in transgenic mice.

There are currently no rapid, operant pain behaviors in rodents that use a self-report to directly engage higher-order brain circuitry. We have developed a pain detection assay consisting of a lick behavior in response to optogenetic activation of predominantly nociceptive peripheral afferent nerve fibers in head-restrained transgenic mice expressing ChR2 in TRPV1 containing neurons. TRPV1-ChR2-EYFP mice (n = 5) were trained to provide lick reports to the detection of light-evoked nociceptive stimulation to the hind paw. Using simultaneous video recording, we demonstrate that the learned lick behavior may prove more pertinent in investigating brain driven pain processes than the reflex behavior. Within sessions, the response bias of transgenic mice changed with respect to lick behavior but not reflex behavior. Furthermore, response similarity between the lick and reflex behaviors diverged near perceptual threshold. Our nociceptive lick-report detection assay will enable a host of investigations into the millisecond, single cell, neural dynamics underlying pain processing in the central nervous system of awake behaving animals.

Discovery and validation of biomarkers to aid the development of safe and effective pain therapeutics: challenges and opportunities.

Pain medication plays an important role in the treatment of acute and chronic pain conditions, but some drugs, opioids in particular, have been overprescribed or prescribed without adequate safeguards, leading to an alarming rise in medication-related overdose deaths. The NIH Helping to End Addiction Long-term (HEAL) Initiative is a trans-agency effort to provide scientific solutions to stem the opioid crisis. One component of the initiative is to support biomarker discovery and rigorous validation in collaboration with industry leaders to accelerate high-quality clinical research into neurotherapeutics and pain. The use of objective biomarkers and clinical trial end points throughout the drug discovery and development process is crucial to help define pathophysiological subsets of pain, evaluate target engagement of new drugs and predict the analgesic efficacy of new drugs. In 2018, the NIH-led Discovery and Validation of Biomarkers to Develop Non-Addictive Therapeutics for Pain workshop convened scientific leaders from academia, industry, government and patient advocacy groups to discuss progress, challenges, gaps and ideas to facilitate the development of biomarkers and end points for pain. The outcomes of this workshop are outlined in this Consensus Statement.

What does a pain 'biomarker' mean and can a machine be taught to measure pain?

Artificial intelligence allows machines to predict human faculties such as image and voice recognition. Can machines be taught to measure pain? We argue that the two fundamental requirements for a device with 'pain biomarker' capabilities are hardware and software. We discuss the merits and limitations of electroencephalography (EEG) as the hardware component of a putative embodiment of the device, and advances in the application of machine learning approaches to EEG for predicting pain.

An Electroencephalography Bioassay for Preclinical Testing of Analgesic Efficacy.

We present a multimodal method combining quantitative electroencephalography (EEG), behavior and pharmacology for pre-clinical screening of analgesic efficacy in vivo. The method consists of an objective and non-invasive approach for realtime assessment of spontaneous nociceptive states based on EEG recordings of theta power over primary somatosensory cortex in awake rats. Three drugs were chosen: (1) pregabalin, a CNS-acting calcium channel inhibitor; (2) EMA 401, a PNS-acting angiotensin II type 2 receptor inhibitor; and (3) minocycline, a CNS-acting glial inhibitor. Optimal doses were determined based on pharmacokinetic studies and/or published data. The effects of these drugs at single or multiple doses were tested on the attenuation of theta power and paw withdrawal latency (PWL) in a rat model of neuropathic pain. We report mostly parallel trends in the reversal of theta power and PWL in response to administration of pregabalin and EMA 401, but not minocycline. We also note divergent trends at non-optimal doses and following prolonged drug administration, suggesting that EEG theta power can be used to detect false positive and false negative outcomes of the withdrawal reflex behavior, and yielding novel insights into the analgesic effects of these drugs on spontaneous nociceptive states in rats.

Automated detection of electroencephalography artifacts in human, rodent and canine subjects using machine learning.

BACKGROUND: Electroencephalography (EEG) invariably contains extra-cranial artifacts that are commonly dealt with based on qualitative and subjective criteria. Failure to account for EEG artifacts compromises data interpretation. NEW METHOD: We have developed a quantitative and automated support vector machine (SVM)-based algorithm to accurately classify artifactual EEG epochs in awake rodent, canine and humans subjects. An embodiment of this method also enables the determination of 'eyes open/closed' states in human subjects. RESULTS: The levels of SVM accuracy for artifact classification in humans, Sprague Dawley rats and beagle dogs were 94.17%, 83.68%, and 85.37%, respectively, whereas 'eyes open/closed' states in humans were labeled with 88.60% accuracy. Each of these results was significantly higher than chance. COMPARISON WITH EXISTING METHODS: Other existing methods, like those dependent on Independent Component Analysis, have not been tested in non-human subjects, and require full EEG montages, instead of only single channels, as this method does. CONCLUSIONS: We conclude that our EEG artifact detection algorithm provides a valid and practical solution to a common problem in the quantitative analysis and assessment of EEG in pre-clinical research settings across evolutionary spectra.

Sub-paresthesia spinal cord stimulation reverses thermal hyperalgesia and modulates low frequency EEG in a rat model of neuropathic pain.

Paresthesia, a common feature of epidural spinal cord stimulation (SCS) for pain management, presents a challenge to the double-blind study design. Although sub-paresthesia SCS has been shown to be effective in alleviating pain, empirical criteria for sub-paresthesia SCS have not been established and its basic mechanisms of action at supraspinal levels are unknown. We tested our hypothesis that sub-paresthesia SCS attenuates behavioral signs of neuropathic pain in a rat model, and modulates pain-related theta (4-8 Hz) power of the electroencephalogram (EEG), a previously validated correlate of spontaneous pain in rodent models. Results show that sub-paresthesia SCS attenuates thermal hyperalgesia and power amplitude in the 3-4 Hz range, consistent with clinical data showing significant yet modest analgesic effects of sub-paresthesia SCS in humans. Therefore, we present evidence for anti-nociceptive effects of sub-paresthesia SCS in a rat model of neuropathic pain and further validate EEG theta power as a reliable 'biosignature' of spontaneous pain.

Thalamic Bursts Down-regulate Cortical Theta and Nociceptive Behavior.

We tested the relation between pain behavior, theta (4-8 Hz) oscillations in somatosensory cortex and burst firing in thalamic neurons in vivo. Optically-induced thalamic bursts attenuated cortical theta and mechanical allodynia. It is proposed that thalamic bursts are an adaptive response to pain that de-synchronizes cortical theta and decreases sensory salience.

Electroencephalographic frontal synchrony and caudal asynchrony during painful hand immersion in cold water.

Recent studies in our laboratory showed that cortical theta oscillations correlate with pain in rodent models. In this study, we sought to validate our pre-clinical data using EEG recordings in humans during immersion of the hand in ice cold water, a moderately noxious stimulus. Power spectral analysis shows that an increase in pain score is associated with an increase in power amplitude within a frequency range of 6-7Hz at the frontal (Fz) electrode. These results are consistent with our previous pre-clinical animal studies and the clinical literature. We also report a novel reduction in power at the caudal (O1) electrode within a broader 3-30Hz rand and decreased coherence between Fz and C3, C4 electrodes within the theta (4-8Hz) and low beta (13-21Hz) bands, while coherence (an indirect measure of functional connectivity) between Fz and O1 increased within the theta and alpha (8-12Hz) bands. We argue that pain is associated with EEG frontal synchrony and caudal asynchrony, leading to the disruption of cortico-cortical connectivity.

Electroencephalographic signatures of pain and analgesia in rats.

Pain modulates rhythmic neuronal activity recorded by Electroencephalography (EEG) in humans. Our laboratory previously showed that rat models of acute and neuropathic pain manifest increased power in primary somatosensory cortex (S1) recorded by electrocorticography (ECoG). In this study, we hypothesized that pain increases EEG power and corticocortical coherence in different rat models of pain, whereas treatments with clinically effective analgesics reverse these changes. Our results show increased cortical power over S1 and prefrontal cortex (PFC) in awake, freely behaving rat models of acute, inflammatory and neuropathic pain. Coherence between PFC and S1 is increased at a late, but not early, time point during the development of neuropathic pain. Electroencephalography power is not affected by ibuprofen in the acute pain model. However, pregabalin and mexiletine reverse the changes in power and S1-PFC coherence in the inflammatory and neuropathic pain models. These data suggest that quantitative EEG might be a valuable predictor of pain and analgesia in rodents.

T-type calcium channel blocker Z944 restores cortical synchrony and thalamocortical connectivity in a rat model of neuropathic pain.

Oscillations are fundamental to communication between neuronal ensembles. We previously reported that pain in awake rats enhances synchrony in primary somatosensory cortex (S1) and attenuates coherence between S1 and ventral posterolateral (VPL) thalamus. Here, we asked whether similar changes occur in anesthetized rats and whether pain modulates phase-amplitude coupling between VPL and S1. We also hypothesized that the suppression of burst firing in VPL using Z944, a novel T-type calcium channel blocker, restores S1 synchrony and thalamocortical connectivity. Local field potentials were recorded from S1 and VPL in anesthetized rats 7 days after sciatic chronic constriction injury (CCI). In rats with CCI, low-frequency (4-12 Hz) synchrony in S1 was enhanced, whereas VPL-S1 coherence and theta-gamma phase-amplitude coupling were attenuated. Moreover, Granger causality showed decreased informational flow from VPL to S1. Systemic or intrathalamic delivery of Z944 to rats with CCI normalized these changes. Systemic Z944 also reversed thermal hyperalgesia and conditioned place preference. These data suggest that pain-induced cortical synchrony and thalamocortical disconnectivity are directly related to burst firing in VPL.

Cortical theta is increased while thalamocortical coherence is decreased in rat models of acute and chronic pain.

Thalamocortical oscillations are critical for sensory perception. Although pain is known to disrupt synchrony in thalamocortical oscillations, evidence in the literature is controversial. Thalamocortical coherence has been reported to be increased in patients with neurogenic pain but decreased in a rat model of central pain. Moreover, theta (4 to 8 Hz) oscillations in primary somatosensory (S1) cortex are speculated to predict pain in humans. To date, the link between pain and network oscillations in animal models has been understudied. Thus, we tested the hypothesis that pain disrupts thalamocortical coherence and S1 theta power in two rat models of pain. We recorded electrocorticography (ECoG) waveforms over S1 and local field potentials (LFP) within ventral posterolateral thalamus in freely behaving rats under spontaneous (stimulus-independent) pain conditions. Rats received intradermal capsaicin injection (Cap) in the hindpaw, followed hours later by chronic constriction injury (CCI) of the sciatic nerve lasting several days. Our results show that pain decreases coherence between LFP and ECoG waveforms in the 2- to 30-Hz range, and increases ECoG power in the theta range. These changes are short-lasting after Cap and longer-lasting after CCI. These data might be particularly relevant to preclinical correlates of spontaneous pain-like behavior, with potential implications to clinical biomarkers of ongoing pain.