I'm a Stanford Data Science Postdoctoral Fellow and NINDS F32 Postdoctoral Fellow, and I work with Professor Todd Coleman in Bioengineering and Professor Sean Mackey in Pain Medicine. My research is on developing new technologies and methods to study the interactions between the brain, the autonomic nervous system, and the gut. Brain-gut interactions are poorly understood but involved in a number of disorders, such as functional gastrointestinal disorders, Parkinson’s disease, diabetes, migraine, and eating disorders. The goal of my research is to improve our ability to monitor and quantify these physiologic processes.

I completed my B.S. in Biomedical Engineering and Applied Mathematics & Statistics from Johns Hopkins University in 2015 and spent the next year as a Churchill Scholar at the University of Cambridge getting an M.Phil. in clinical neurosciences (all my research was computational). I then did my Ph.D. at MIT in the Harvard-MIT Health Sciences and Technology program, advised by Professor Emery Brown. During my PhD, I developed and tested models and methods to track unconscious pain under anesthesia in the operating room. I grew up in Grand Rapids, Michigan.

Honors & Awards

  • Schmidt Science Fellowship, Schmidt Science Fellows, Schmidt Futures, Rhodes Trust (2022-2024)
  • Stanford Data Science Postdoctoral Fellowship, Stanford University (2021-2024)
  • NIH NINDS F32 Postdoctoral Fellowship (Terminated early for Schmidt Science Fellowship), NIH NINDS (2021-2022)
  • Collamore-Rogers Fellowship, MIT Office of Graduate Education (2020-2021)
  • 3rd place, IEEE Engineering in Medicine and Biology Conference Student Paper Competition, IEEE Engineering in Medicine and Biology Society (2020)
  • Sloan School of Management Healthcare Certificate, MIT Sloan School of Management (2020)
  • Kauffman Teaching Certificate Program Graduate, MIT Teaching-Learning Lab (2019)
  • Society for Neuroscience Trainee Professional Development Award, Society for Neuroscience (2017)
  • MIT Presidential Fellow, MIT (2016)
  • NSF Graduate Fellow, National Science Foundation (2015-2020)
  • Churchill Scholar, The Winston Churchill Foundation (2015-2016)
  • Goldwater Scholarship, Barry Goldwater Scholarship and Excellence in Education Foundation (2013)

Boards, Advisory Committees, Professional Organizations

  • Member, Reviewer, IEEE Engineering in Medicine and Biology Society (2017 - Present)

Professional Education

  • Doctor of Philosophy, Massachusetts Institute of Technology (2021)
  • Master of Philosophy, University of Cambridge (2016)
  • Bachelor of Science, Johns Hopkins University, Biomedical Engineering (2015)
  • B.S., Johns Hopkins University, Biomedical Engineering, Applied Mathematics & Statistics (2015)
  • M.Phil., University of Cambridge, Clinical Neurosciences (2016)
  • Ph.D., Harvard-MIT Division of Health Sciences and Technology, Medical Engineering and Medical Physics (2021)

Stanford Advisors


  • Emery N Brown, Riccardo Barbieri, Sandya Subramanian. "United States Patent PCT/US2020/042031 Tracking Nociception Under Anesthesia Using a Multimodal Metric", Massachusetts Institute of Technology
  • Sandya Subramanian, Riccardo Barbieri, Emery N Brown. "United StatesSurgical Cautery Artifact Removal from Electrodermal Activity Data", Massachusetts Institute of Technology
  • Sandya Subramanian, Todd Coleman. "United StatesAutomated classification of sleep and wake from single day triaxial accelerometer data", Stanford University
  • Aaron Chang, Melinda Chen, Piyush Poddar, Rohil Malpani, Peter Malamas, Sandya Subramanian, Joon Eoh, Kevin George, Todd J. Cohen. "United States Patent 9,474,892 Method and System for Decreasing Transthoracic Impedance for Cardioversion and Defibrillation", Cardiac Inventions, Oct 25, 2016
  • Piyush Poddar, Aaron Chang, Melinda Chen, Peter Malamas, Sandya Subramanian, Todd J. Cohen. "United States Patent 9,320,884 Method and System for Switching Shock Vectors and Decreasing Transthoracic Impedance for Cardioversion and Defibrillation", Cardiac Inventions, Apr 26, 2016
  • Sridevi Vedula Sarma, Sandya Subramanian, Stephanie Hao. "United States Patent 9,277,873 Computational tool for pre-surgical evaluation of patients with medically refractory epilepsy", Johns Hopkins University, Mar 8, 2016

Research Interests

  • Brain and Learning Sciences
  • Data Sciences

Current Research and Scholarly Interests

I would like to focus on platform technology development for at-home monitoring of chronic disease, by studying gut-autonomic nervous system interactions. I am trained as an engineer and computational researcher, and I have experience developing computational algorithms from physiology, collecting data from patients in complex clinical scenarios, and collaborating with diverse clinical and regulatory teams. I am developing expertise in hardware-software interfacing and bioelectronics.

Lab Affiliations

All Publications

  • An unsupervised automated paradigm for artifact removal from electrodermal activity in an uncontrolled clinical setting. Physiological measurement Subramanian, S., Tseng, B., Barbieri, R., Brown, E. N. 2022


    OBJECTIVE: Electrodermal activity (EDA) reflects sympathetic nervous system activity through sweating-related changes in skin conductance and could be used in clinical settings in which patients cannot self-report pain, such as during surgery or when in a coma. To enable EDA data to be used robustly in clinical settings, we need to develop artifact detection and removal frameworks that can handle the types of interference experienced in clinical settings while salvaging as much useful information as possible.APPROACH: In this study, we collected EDA data from 70 subjects while they were undergoing surgery in the operating room. We then built a fully automated artifact removal framework to remove the heavy artifacts that resulted from the use of surgical electrocautery during the surgery and compared it to two existing state-of-the-art methods for artifact removal from EDA data. This automated framework consisted of first utilizing three unsupervised machine learning methods for anomaly detection, and then customizing the threshold to separate artifact for each data instance by taking advantage of the statistical properties of the artifact in that data instance. We also created simulated surgical data by introducing artifacts into cleaned surgical data and measured the performance of all three methods in removing it.MAIN RESULTS: Our method achieved the highest overall accuracy and precision and lowest overall error on simulated data. One of the other methods prioritized high sensitivity while sacrificing specificity and precision, while the other had low sensitivity, high error, and left behind several artifacts. These results were qualitatively similar between the simulated data instances and operating room data instances.SIGNIFICANCE: Our framework allows for robust removal of heavy artifact from EDA data in clinical settings such as surgery, which is the first step to enable clinical integration of EDA as part of standard monitoring.

    View details for DOI 10.1088/1361-6579/ac92bd

    View details for PubMedID 36113446

  • Automated classification of sleep and wake from single day triaxial accelerometer data. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference Subramanian, S., Coleman, T. P. 2022; 2022: 3665-3668


    Actigraphy allows for the remote monitoring of subjects' activity for clinical and research purposes. However, most standard methods are built for proprietary measures from specific devices that are not widely used. In this study, we develop an algorithm for classifying sleep and awake using a single day of triaxial accelerometer data, which can be acquired from all smart devices. This algorithm consists of two stages, clustering and hidden Markov modeling, and outperforms standard algorithms in sensitivity (94%), specificity (93 %), and overall accuracy (93%) across seven subjects. This method can help automate actigraphy analyses at scale using widely available technology using even a single day's worth of data. Clinical Relevance- Automated monitoring of patients' activity at home can help track recovery trajectories after surgery and injury, disease progression, treatment response.

    View details for DOI 10.1109/EMBC48229.2022.9871823

    View details for PubMedID 36086032

  • Tonic Electrodermal Activity is a Robust Marker of Psychological and Physiological Changes during Induction of Anesthesia. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference Tseng, B., Subramanian, S., Barbieri, R., Brown, E. N. 2022; 2022: 418-421


    Electrodermal activity (EDA), which tracks sweat gland activity as a proxy for sympathetic activation, has the potential to be a biomarker of physiological and psychological changes in the clinic. To show this, in this study, we demonstrate that the tonic component of EDA responds consistently and robustly during induction of anesthesia in the operating room in 8 subjects during surgery. This response is seen bilaterally. The response shows a significant increase in EDA in anticipation of induction and then a gradual decrease in response to the administration of medication, which agrees with both the expected psychological effects of stress and anxiety and the physiological effects of anesthetic medication on sweat glands. The results also show a slightly faster response to drug in the arm directly receiving the medication intravenously compared to the opposite, though the magnitude of the effect evens out over time. Clinical Relevance- EDA can serve as a robust non-invasive biomarker in the clinic to track both psychologically and physiologically induced autonomic changes.

    View details for DOI 10.1109/EMBC48229.2022.9871080

    View details for PubMedID 36086567

  • How Machine Learning is Powering Neuroimaging to Improve Brain Health. Neuroinformatics Singh, N. M., Harrod, J. B., Subramanian, S., Robinson, M., Chang, K., Cetin-Karayumak, S., Dalca, A. V., Eickhoff, S., Fox, M., Franke, L., Golland, P., Haehn, D., Iglesias, J. E., O'Donnell, L. J., Ou, Y., Rathi, Y., Siddiqi, S. H., Sun, H., Westover, M. B., Whitfield-Gabrieli, S., Gollub, R. L. 2022


    This report presents an overview of how machine learning is rapidly advancing clinical translational imaging in ways thatwill aid in the early detection, prediction, and treatment of diseases that threaten brain health. Towards this goal, we aresharing the information presented at a symposium, "Neuroimaging Indicators of Brain Structure and Function - Closingthe Gap Between Research and Clinical Application", co-hosted by the McCance Center for Brain Health at Mass GeneralHospital and the MIT HST Neuroimaging Training Program on February 12, 2021. The symposium focused on the potentialfor machine learning approaches, applied to increasingly large-scale neuroimaging datasets, to transform healthcare deliveryand change the trajectory of brain health by addressing brain care earlier in the lifespan. While not exhaustive, this overviewuniquely addresses many of the technical challenges from image formation, to analysis and visualization, to synthesisand incorporation into the clinical workflow. Some of the ethical challenges inherent to this work are also explored, as aresome of the regulatory requirements for implementation. We seek to educate, motivate, and inspire graduate students, postdoctoralfellows, and early career investigators to contribute to a future where neuroimaging meaningfully contributes to themaintenance of brain health.

    View details for DOI 10.1007/s12021-022-09572-9

    View details for PubMedID 35347570

  • A Model-Based Framework for Assessing the Physiologic Structure of Electrodermal Activity IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING Subramanian, S., Purdon, P. L., Barbieri, R., Brown, E. N. 2021; 68 (9): 2833-2845


    We present a statistical model for extracting physiologic characteristics from electrodermal activity (EDA) data in observational settings.We based our model on the integrate-and-fire physiology of sweat gland bursts, which predicts inverse Gaussian (IG) inter-pulse interval structure. At the core of our model-based paradigm is a subject-specific amplitude threshold selection process for EDA pulses based on the statistical properties of four right-skewed models including the IG. By performing a sensitivity analysis across thresholds and fitting all four models, we selected for IG-like structure and verified the pulse selection with a goodness-of-fit analysis, maximizing capture of physiology at the time scale of EDA responses.We tested the model-based paradigm on simulated EDA time series and data from two different experimental cohorts recorded during different experimental conditions, using different equipment. In both the simulated and experimental data, our model-based method robustly recovered pulses that captured the IG-like structure predicted by physiology, despite large differences in noise level. In contrast, established EDA analysis tools, which attempted to estimate neural activity from slower EDA responses, did not provide physiological validation and were susceptible to noise.We present a computationally efficient, statistically rigorous, and physiology-informed paradigm for pulse selection from EDA data that is robust across individuals and experimental conditions, yet adaptable to varying noise level.The robustness of the model-based paradigm and its physiological basis provide empirical support for the use of EDA as a clinical marker for sympathetic activity in conditions such as pain, anxiety, depression, and sleep states.

    View details for DOI 10.1109/TBME.2021.3071366

    View details for Web of Science ID 000688216600004

    View details for PubMedID 33822719

    View details for PubMedCentralID PMC8425954

  • Elementary integrate-and-fire process underlies pulse amplitudes in Electrodermal activity PLOS COMPUTATIONAL BIOLOGY Subramanian, S., Purdon, P. L., Barbieri, R., Brown, E. N. 2021; 17 (7): e1009099


    Electrodermal activity (EDA) is a direct read-out of sweat-induced changes in the skin's electrical conductance. Sympathetically-mediated pulsatile changes in skin sweat measured as EDA resemble an integrate-and-fire process, which yields an inverse Gaussian model as the inter-pulse interval distribution. We have previously showed that the inter-pulse intervals in EDA follow an inverse Gaussian distribution. However, the statistical structure of EDA pulse amplitudes has not yet been characterized based on the physiology. Expanding upon the integrate-and-fire nature of sweat glands, we hypothesized that the amplitude of an EDA pulse is proportional to the excess volume of sweat produced compared to what is required to just reach the surface of the skin. We modeled this as the difference of two inverse Gaussian models for each pulse, one which represents the time required to produce just enough sweat to rise to the surface of the skin and one which represents the time requires to produce the actual volume of sweat. We proposed and tested a series of four simplifications of our hypothesis, ranging from a single difference of inverse Gaussians to a single simple inverse Gaussian. We also tested four additional models for comparison, including the lognormal and gamma distributions. All models were tested on EDA data from two subject cohorts, 11 healthy volunteers during 1 hour of quiet wakefulness and a different set of 11 healthy volunteers during approximately 3 hours of controlled propofol sedation. All four models which represent simplifications of our hypothesis outperformed other models across all 22 subjects, as measured by Akaike's Information Criterion (AIC), as well as mean and maximum distance from the diagonal on a quantile-quantile plot. Our broader model set of four simplifications offered a useful framework to enhance further statistical descriptions of EDA pulse amplitudes. Some of the simplifications prioritize fit near the mode of the distribution, while others prioritize fit near the tail. With this new insight, we can summarize the physiologically-relevant amplitude information in EDA with at most four parameters. Our findings establish that physiologically based probability models provide parsimonious and accurate description of temporal and amplitude characteristics in EDA.

    View details for DOI 10.1371/journal.pcbi.1009099

    View details for Web of Science ID 000674290300001

    View details for PubMedID 34232965

    View details for PubMedCentralID PMC8289084

  • Unsupervised Machine Learning Methods for Artifact Removal in Electrodermal Activity. Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference Subramanian, S., Tseng, B., Barbieri, R., Brown, E. N. 2021; 2021: 399-402


    Artifact detection and removal is a crucial step in all data preprocessing pipelines for physiological time series data, especially when collected outside of controlled experimental settings. The fact that such artifact is often readily identifiable by eye suggests that unsupervised machine learning algorithms may be a promising option that do not require manually labeled training datasets. Existing methods are often heuristic-based, not generalizable, or developed for controlled experimental settings with less artifact. In this study, we test the ability of three such unsupervised learning algorithms, isolation forests, 1-class support vector machine, and K-nearest neighbor distance, to remove heavy cautery-related artifact from electrodermal activity (EDA) data collected while six subjects underwent surgery. We first defined 12 features for each halfsecond window as inputs to the unsupervised learning methods. For each subject, we compared the best performing unsupervised learning method to four other existing methods for EDA artifact removal. For all six subjects, the unsupervised learning method was the only one successful at fully removing the artifact. This approach can easily be expanded to other modalities of physiological data in complex settings.Clinical Relevance- Robust artifact detection methods allow for the use of diverse physiological data even in complex clinical settings to inform diagnostic and therapeutic decisions.

    View details for DOI 10.1109/EMBC46164.2021.9630535

    View details for PubMedID 34891318

  • Quantitative assessment of the relationship between behavioral and autonomic dynamics during propofol-induced unconsciousness PLOS ONE Subramanian, S., Purdon, P. L., Barbieri, R., Brown, E. N. 2021; 16 (8): e0254053


    During general anesthesia, both behavioral and autonomic changes are caused by the administration of anesthetics such as propofol. Propofol produces unconsciousness by creating highly structured oscillations in brain circuits. The anesthetic also has autonomic effects due to its actions as a vasodilator and myocardial depressant. Understanding how autonomic dynamics change in relation to propofol-induced unconsciousness is an important scientific and clinical question since anesthesiologists often infer changes in level of unconsciousness from changes in autonomic dynamics. Therefore, we present a framework combining physiology-based statistical models that have been developed specifically for heart rate variability and electrodermal activity with a robust statistical tool to compare behavioral and multimodal autonomic changes before, during, and after propofol-induced unconsciousness. We tested this framework on physiological data recorded from nine healthy volunteers during computer-controlled administration of propofol. We studied how autonomic dynamics related to behavioral markers of unconsciousness: 1) overall, 2) during the transitions of loss and recovery of consciousness, and 3) before and after anesthesia as a whole. Our results show a strong relationship between behavioral state of consciousness and autonomic dynamics. All of our prediction models showed areas under the curve greater than 0.75 despite the presence of non-monotonic relationships among the variables during the transition periods. Our analysis highlighted the specific roles played by fast versus slow changes, parasympathetic vs sympathetic activity, heart rate variability vs electrodermal activity, and even pulse rate vs pulse amplitude information within electrodermal activity. Further advancement upon this work can quantify the complex and subject-specific relationship between behavioral changes and autonomic dynamics before, during, and after anesthesia. However, this work demonstrates the potential of a multimodal, physiologically-informed, statistical approach to characterize autonomic dynamics.

    View details for DOI 10.1371/journal.pone.0254053

    View details for Web of Science ID 000684029800070

    View details for PubMedID 34379623

    View details for PubMedCentralID PMC8357089

  • Point process temporal structure characterizes electrodermal activity PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA Subramanian, S., Barbieri, R., Brown, E. N. 2020; 117 (42): 26422-26428


    Electrodermal activity (EDA) is a direct readout of the body's sympathetic nervous system measured as sweat-induced changes in the skin's electrical conductance. There is growing interest in using EDA to track physiological conditions such as stress levels, sleep quality, and emotional states. Standardized EDA data analysis methods are readily available. However, none considers an established physiological feature of EDA. The sympathetically mediated pulsatile changes in skin sweat measured as EDA resemble an integrate-and-fire process. An integrate-and-fire process modeled as a Gaussian random walk with drift diffusion yields an inverse Gaussian model as the interpulse interval distribution. Therefore, we chose an inverse Gaussian model as our principal probability model to characterize EDA interpulse interval distributions. To analyze deviations from the inverse Gaussian model, we considered a broader model set: the generalized inverse Gaussian distribution, which includes the inverse Gaussian and other diffusion and nondiffusion models; the lognormal distribution which has heavier tails (lower settling rates) than the inverse Gaussian; and the gamma and exponential probability distributions which have lighter tails (higher settling rates) than the inverse Gaussian. To assess the validity of these probability models we recorded and analyzed EDA measurements in 11 healthy volunteers during 1 h of quiet wakefulness. Each of the 11 time series was accurately described by an inverse Gaussian model measured by Kolmogorov-Smirnov measures. Our broader model set offered a useful framework to enhance further statistical descriptions of EDA. Our findings establish that a physiologically based inverse Gaussian probability model provides a parsimonious and accurate description of EDA.

    View details for DOI 10.1073/pnas.2004403117

    View details for Web of Science ID 000580597300063

    View details for PubMedID 33008878

    View details for PubMedCentralID PMC7584910

  • Multimodal vs Unimodal Estimation of Sympathetic-Driven Arousal States Subramanian, S., Brown, E., Barbieri, R., IEEE IEEE. 2020
  • Analyzing Transitions in Anesthesia by Multimodal Characterization of Autonomic State Subramanian, S., Barbieri, R., Purdon, P. L., Brown, E. N., IEEE IEEE. 2020
  • Detecting Loss and Regain of Consciousness during Propofol Anesthesia using Multimodal Indices of Autonomic State Subramanian, S., Barbieri, R., Purdon, P. L., Brown, E. N., IEEE IEEE. 2020: 824-827


    We have traditionally defined `loss of consciousness' (LOC) and `regain of consciousness' (ROC) during general anesthesia in terms of behavioral correlates. We are starting to understand the dynamics in brain activity that may help define those events; however, we have not yet explored the possible autonomic correlates of LOC and ROC. In this study, we investigated the autonomic dynamics immediately surrounding loss and regain of consciousness in nine healthy volunteers under controlled propofol sedation. We used multimodal autonomic indices generated from physiologically accurate models and found that just before and after LOC and ROC could be differentiated with an AUC of 0.80. In addition, we saw that some of the autonomic changes accompanying LOC and ROC verify known information about the mechanism of action of propofol, while others indicate new avenues for exploration of propofol's effect on the autonomic nervous system. Overall, our work suggests that the autonomic dynamics surrounding the events of loss and regain of consciousness are worthy of further investigation.Clinical Relevance-This introduces the possibility of autonomic biomarkers for loss and regain of consciousness during general anesthesia that are more precise than behavioral tracking alone.

    View details for Web of Science ID 000621592201039

    View details for PubMedID 33018112

  • Risk-taking bias in human decision-making is encoded via a right-left brain push-pull system PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA Sacre, P., Kerr, M. D., Subramanian, S., Fitzgerald, Z., Kahn, K., Johnson, M. A., Niebur, E., Eden, U. T., Gonzalez-Martinez, J. A., Gale, J. T., Sarma, S. V. 2019; 116 (4): 1404-1413


    A person's decisions vary even when options stay the same, like when a gambler changes bets despite constant odds of winning. Internal bias (e.g., emotion) contributes to this variability and is shaped by past outcomes, yet its neurobiology during decision-making is not well understood. To map neural circuits encoding bias, we administered a gambling task to 10 participants implanted with intracerebral depth electrodes in cortical and subcortical structures. We predicted the variability in betting behavior within and across patients by individual bias, which is estimated through a dynamical model of choice. Our analysis further revealed that high-frequency activity increased in the right hemisphere when participants were biased toward risky bets, while it increased in the left hemisphere when participants were biased away from risky bets. Our findings provide electrophysiological evidence that risk-taking bias is a lateralized push-pull neural system governing counterintuitive and highly variable decision-making in humans.

    View details for DOI 10.1073/pnas.1811259115

    View details for Web of Science ID 000456336100048

    View details for PubMedID 30617071

    View details for PubMedCentralID PMC6347682

  • A Systematic Method for Preprocessing and Analyzing Electrodermal Activity Subramanian, S., Barbieri, R., Brown, E. N., IEEE IEEE. 2019: 6902-6905


    Electrodermal activity (EDA) is a measure of sympathetic tone using sweat gland activity that has applications in research and clinical medicine. We previously identified never-before-seen statistical structure in EDA. However, there is no systematic method to preprocess and analyze EDA data to capture such statistical structure. Therefore, in this study, we analyzed the data of two healthy volunteers while awake and at rest. We used a systematic process that takes advantage of the tail behavior of various statistical distributions to ensure capturing the point process structure in EDA. We verified the presence of this temporal structure in a new dataset of subjects. Our results demonstrate for the first time that point process structure of EDA pulses can be identified across multiple datasets using a systematic method that is still rooted in the underlying physiology.

    View details for Web of Science ID 000557295307078

    View details for PubMedID 31947426

  • Multitaper Infinite Hidden Markov Model for EEG Song, A. H., Chlon, L., Soulat, H., Tauber, J., Subramanian, S., Ba, D., Prerau, M. J., IEEE IEEE. 2019: 5803-5807


    Electroencephalographam (EEG) monitoring of neural activity is widely used for identifying underlying brain states. For inference of brain states, researchers have often used Hidden Markov Models (HMM) with a fixed number of hidden states and an observation model linking the temporal dynamics embedded in EEG to the hidden states. The use of fixed states may be limiting, in that 1) pre-defined states might not capture the heterogeneous neural dynamics across individuals and 2) the oscillatory dynamics of the neural activity are not directly modeled. To this end, we use a Hierarchical Dirichlet Process Hidden Markov Model (HDP-HMM), which discovers the set of hidden states that best describes the EEG data, without a-priori specification of state number. In addition, we introduce an observation model based on classical asymptotic results of frequency domain properties of stationary time series, along with the description of the conditional distributions for Gibbs sampler inference. We then combine this with multitaper spectral estimation to reduce the variance of the spectral estimates. By applying our method to simulated data inspired by sleep EEG, we arrive at two main results: 1) the algorithm faithfully recovers the spectral characteristics of the true states, as well as the right number of states and 2) the incorporation of the multitaper framework produces a more stable estimate than traditional periodogram spectral estimates.

    View details for Web of Science ID 000557295306054

    View details for PubMedID 31947171

    View details for PubMedCentralID PMC7029542

  • Arousal Detection in Obstructive Sleep Apnea Using Physiology-Driven Features Subramanian, S., Chamadia, S., Chakravarty, S., IEEE IEEE. 2018
  • A Point Process Characterization of Electrodermal Activity Subramanian, S., Barbieri, R., Brown, E. N., IEEE IEEE. 2018: 37-40


    Electrodermal activity (EDA) is a measure of sympathetic activity using skin conductance that has applications in research and in clinical medicine. However, current EDA analysis does not have physiologically-based statistical models that use stochastic structure to provide nuanced insight into autonomic dynamics. Therefore, in this study, we analyzed the data of two healthy volunteers under controlled propofol sedation. We identified a novel statistical model for EDA and used a point process framework to track instantaneous dynamics. Our results demonstrate for the first time that point process models rooted in physiology and built upon inherent statistical structure of EDA pulses have the potential to accurately track instantaneous dynamics in sympathetic tone.

    View details for Web of Science ID 000596231900009

    View details for PubMedID 30440335

  • Using network analysis to localize the epileptogenic zone from invasive EEG recordings in intractable focal epilepsy NETWORK NEUROSCIENCE Li, A., Chennuri, B., Subramanian, S., Yaffe, R., Gliske, S., Stacey, W., Norton, R., Jordan, A., Zaghloul, K. A., Inati, S. K., Agrawal, S., Haagensen, J. J., Hopp, J., Atallah, C., Johnson, E., Crone, N., Anderson, W. S., Fitzgerald, Z., Bulacio, J., Gale, J. T., Sarma, S. V., Gonzalez-Martinez, J. 2018; 2 (2): 218-240


    Treatment of medically intractable focal epilepsy (MIFE) by surgical resection of the epileptogenic zone (EZ) is often effective provided the EZ can be reliably identified. Even with the use of invasive recordings, the clinical differentiation between the EZ and normal brain areas can be quite challenging, mainly in patients without MRI detectable lesions. Consequently, despite relatively large brain regions being removed, surgical success rates barely reach 60-65%. Such variable and unfavorable outcomes associated with high morbidity rates are often caused by imprecise and/or inaccurate EZ localization. We developed a localization algorithm that uses network-based data analytics to process invasive EEG recordings. This network algorithm analyzes the centrality signatures of every contact electrode within the recording network and characterizes contacts into susceptible EZ based on the centrality trends over time. The algorithm was tested in a retrospective study that included 42 patients from four epilepsy centers. Our algorithm had higher agreement with EZ regions identified by clinicians for patients with successful surgical outcomes and less agreement for patients with failed outcomes. These findings suggest that network analytics and a network systems perspective of epilepsy may be useful in assisting clinicians in more accurately localizing the EZ.

    View details for DOI 10.1162/netn_a_00043

    View details for Web of Science ID 000449590300006

    View details for PubMedID 30215034

    View details for PubMedCentralID PMC6130438

  • The influences and neural correlates of past and present during gambling in humans SCIENTIFIC REPORTS Sacre, P., Subramanian, S., Kerr, M. D., Kahn, K., Johnson, M. A., Bulacio, J., Gonzalez-Martinez, J. A., Sarma, S. V., Gale, J. T. 2017; 7: 17111


    During financial decision-making tasks, humans often make "rational" decisions, where they maximize expected reward. However, this rationality may compete with a bias that reflects past outcomes. That is, if one just lost money or won money, this may impact future decisions. It is unclear how past outcomes influence future decisions in humans, and how neural circuits encode present and past information. In this study, six human subjects performed a financial decision-making task while we recorded local field potentials from multiple brain structures. We constructed a model for each subject characterizing bets on each trial as a function of present and past information. The models suggest that some patients are more influenced by previous trial outcomes (i.e., previous return and risk) than others who stick to more fixed decision strategies. In addition, past return and present risk modulated with the activity in the cuneus; while present return and past risk modulated with the activity in the superior temporal gyrus and the angular gyrus, respectively. Our findings suggest that these structures play a role in decision-making beyond their classical functions by incorporating predictions and risks in humans' decision strategy, and provide new insight into how humans link their internal biases to decisions.

    View details for DOI 10.1038/s41598-017-16862-9

    View details for Web of Science ID 000417353600005

    View details for PubMedID 29214997

    View details for PubMedCentralID PMC5719351

  • Computing Network-based Features from Intracranial EEG Time Series Data: Application to Seizure Focus Localization Hao, S., Subramanian, S., Jordan, A., Santaniello, S., Yaffe, R., Jouny, C. C., Bergey, G. K., Anderson, W. S., Sarma, S. V., IEEE IEEE. 2014: 5812-5815


    The surgical resection of the epileptogenic zone (EZ) is the only effective treatment for many drug-resistant epilepsy (DRE) patients, but the pre-surgical identification of the EZ is challenging. This study investigates whether the EZ exhibits a computationally identifiable signature during seizures. In particular, we compute statistics of the brain network from intracranial EEG (iEEG) recordings and track the evolution of network connectivity before, during, and after seizures. We define each node in the network as an electrode and weight each edge connecting a pair of nodes by the gamma band cross power of the corresponding iEEG signals. The eigenvector centrality (EVC) of each node is tracked over two seizures per patient and the electrodes are ranked according to the corresponding EVC value. We hypothesize that electrodes covering the EZ have a signature EVC rank evolution during seizure that differs from electrodes outside the EZ. We tested this hypothesis on multi-channel iEEG recordings from 2 DRE patients who had successful surgery (i.e., seizures were under control with or without medications) and 1 patient who had unsuccessful surgery. In the successful cases, we assumed that the resected region contained the EZ and found that the EVC rank evolution of the electrodes within the resected region had a distinct "arc" signature, i.e., the EZ ranks first rose together shortly after seizure onset and then fell later during seizure.

    View details for Web of Science ID 000350044705200

    View details for PubMedID 25571317