Dr. Razavi's clinical interests are in medically refractory epilepsies and using high density EEG (electroencephalogram) for better localization of seizure foci. His research areas include using engineering techniques for analyzing EEGs, medical devices for evaluation and treatment of epilepsy, and using seizures as a model for understanding consciousness.
Clinical Assistant Professor, Neurology & Neurological Sciences
Boards, Advisory Committees, Professional Organizations
Member, American Academy of Neurology (2010 - Present)
Member, American Epilepsy Society (2012 - Present)
PhD Training:University of Rochester (2009) NY
Fellowship:Stanford University Medical Center (2015)
Board Certification: Neurology, American Board of Psychiatry and Neurology (2013)
Residency:UC Davis Medical Center (2013) CA
Internship:University of Rochester, Strong Memorial Hospital (2010) NY
Medical Education:University of Rochester (2009)
PhD, University of Rochester, Biomedical Engineering (2009)
Quantitative EEG Metrics Differ Between Outcome Groups and Change Over the First 72 h in Comatose Cardiac Arrest Patients
2018; 28 (1): 51–59
Forty to sixty-six percent of patients resuscitated from cardiac arrest remain comatose, and historic outcome predictors are unreliable. Quantitative spectral analysis of continuous electroencephalography (cEEG) may differ between patients with good and poor outcomes.Consecutive patients with post-cardiac arrest hypoxic-ischemic coma undergoing cEEG were enrolled. Spectral analysis was conducted on artifact-free contiguous 5-min cEEG epochs from each hour. Whole band (1-30 Hz), delta (δ, 1-4 Hz), theta (θ, 4-8 Hz), alpha (α, 8-13 Hz), beta (β, 13-30 Hz), α/δ power ratio, percent suppression, and variability were calculated and correlated with outcome. Graphical patterns of quantitative EEG (qEEG) were described and categorized as correlating with outcome. Clinical outcome was dichotomized, with good neurologic outcome being consciousness recovery.Ten subjects with a mean age = 50 yrs (range = 18-65) were analyzed. There were significant differences in total power (3.50 [3.30-4.06] vs. 0.68 [0.52-1.02], p = 0.01), alpha power (1.39 [0.66-1.79] vs 0.27 [0.17-0.48], p < 0.05), delta power (2.78 [2.21-3.01] vs 0.55 [0.38-0.83], p = 0.01), percent suppression (0.66 [0.02-2.42] vs 73.4 [48.0-97.5], p = 0.01), and multiple measures of variability between good and poor outcome patients (all values median [IQR], good vs. poor). qEEG patterns with high or increasing power or large power variability were associated with good outcome (n = 6). Patterns with consistently low or decreasing power or minimal power variability were associated with poor outcome (n = 4).These preliminary results suggest qEEG metrics correlate with outcome. In some patients, qEEG patterns change over the first three days post-arrest.
View details for DOI 10.1007/s12028-017-0419-2
View details for Web of Science ID 000425631100008
View details for PubMedID 28646267
Detecting silent seizures by their sound
2018; 59 (4): 877-884
The traditional approach to interpreting electroencephalograms (EEGs) requires physicians with formal training to visually assess the waveforms. This approach can be less practical in critical settings where a trained EEG specialist is not readily available to review the EEG and diagnose ongoing subclinical seizures, such as nonconvulsive status epilepticus.We have developed a novel method by which EEG data are converted to sound in real time by letting the underlying electrophysiological signal modulate a voice tone that is in the audible range. Here, we explored whether individuals without any prior EEG training could listen to 15-second sonified EEG and determine whether the EEG represents seizures or nonseizure conditions. We selected 84 EEG samples to represent seizures (n = 7), seizure-like activity (n = 25), or nonperiodic, nonrhythmic activity (normal or focal/generalized slowing, n = 52). EEGs from single channels in the left and right hemispheres were then converted to sound files. After a 4-minute training video, medical students (n = 34) and nurses (n = 30) were asked to designate each audio sample as "seizure" or "nonseizure." We then compared their performance with that of EEG-trained neurologists (n = 12) and medical students (n = 29) who also diagnosed the same EEGs on visual display.Nonexperts listening to single-channel sonified EEGs detected seizures with remarkable sensitivity (students, 98% ± 5%; nurses, 95% ± 14%) compared to experts or nonexperts reviewing the same EEGs on visual display (neurologists, 88% ± 11%; students, 76% ± 19%). If the EEGs contained seizures or seizure-like activity, nonexperts listening to sonified EEGs rated them as seizures with high specificity (students, 85% ± 9%; nurses, 82% ± 12%) compared to experts or nonexperts viewing the EEGs visually (neurologists, 90% ± 7%; students, 65% ± 20%).Our study confirms that individuals without EEG training can detect ongoing seizures or seizure-like rhythmic periodic patterns by listening to sonified EEG. Although sonification of EEG cannot replace the traditional approaches to EEG interpretation, it provides a meaningful triage tool for fast assessment of patients with suspected subclinical seizures.
View details for DOI 10.1111/epi.14043
EEG with Fewer Electrodes Is More Specific for Detecting Seizures and Seizure-Like Activity
WILEY. 2017: S158
View details for Web of Science ID 000413198700326
Utility of electroencephalography: Experience from a U.S. tertiary care medical center.
2016; 127 (10): 3335-3340
To investigate the utility of electroencephalography (EEG) for evaluation of patients with altered mental status (AMS).We retrospectively reviewed 200 continuous EEGs (cEEGs) obtained in ICU and non-ICU wards and 100 spot EEGs (sEEGs) obtained from the emergency department (ED) of a large tertiary medical center. Main outcomes were access time (from study request to hookup), and diagnostic yield (percentage of studies revealing significant abnormality).Access time, mean±SD (maximum), was 3.5±3.2 (20.8) hours in ICU, 4.8±5.0 (25.6) hours in non-ICU, and 2.7±3.6 (23.9) hours in ED. Access time was not significantly different for stat requests or EEGs with seizure activity. While the primary indication for EEG monitoring was to evaluate for seizures as the cause of AMS, only 8% of cEEGs and 1% of sEEGs revealed seizures. Epileptiform discharges were detected in 45% of ICU, 24% of non-ICU, and 9% of ED cases, while 2% of ICU, 15% of non-ICU, and 45% of ED cases were normal.Access to EEG is hampered by significant delays, and in emergency settings, the conventional EEG system detects seizures only in a minority of cases.Our findings underscore the inefficiencies of current EEG infrastructure for accessing diagnostically important information, as well as the need for more prospective data describing the relationship between EEG access time and EEG findings, clinical outcomes, and cost considerations.
View details for DOI 10.1016/j.clinph.2016.08.013
View details for PubMedID 27611442