Anirudh Wodeyar

Anirudh Wodeyar

Assistant Professor of Statistical Learning

Department of Advanced Computing Sciences · Maastricht University

Statistics  ·  Signal Processing  ·  Neuroscience

Email Google Scholar ResearchGate BlueSky CV

About

My research goal is to give people back agency. Epilepsy and Parkinson's disease take it away: seizures and epileptic spikes interrupt a life without warning, and altered rhythms in deep brain structures erode motor control. Getting that agency back means tracking the dynamics the brain is governed by accurately enough to move them, and to move them somewhere better. I am an Assistant Professor of Statistical Learning in the Department of Advanced Computing Sciences at Maastricht University, where I build the statistical and signal-processing tools that make this possible.

Tracking rhythms in real time. I introduced a state-space approach to real-time phase estimation (eLife, 2021). It treats each brain rhythm as a latent state with per-sample uncertainty, instead of something to filter out of the signal. Independent groups have since adopted it, benchmarked against it or extended it. I then showed that phase is not uniquely defined for real signals, so closed-loop use needs uncertainty estimates (eNeuro, 2023). I am now extending this into a general framework, the State-Space Model of Rhythms. Its first extension detects beta bursts in Parkinson's disease sample by sample (IEEE EMBC 2026, oral presentation).

Resolving how rhythms drive one another. I develop methods that infer brain networks: a connectome-constrained graphical lasso for MEG, and a generative structure–function model of stroke damage. With my students I extended this to sparse event data. Group-lasso point-process models infer directed epileptic spike networks to help localise where seizures start (EUSIPCO 2026).

Methods that change what others can measure. Analytic tools I contributed helped establish low-frequency oscillations as a biomarker of injury and recovery after stroke (Stroke, 2020), my most-cited work. I showed that thalamic epileptic spikes disrupt sleep spindles (Brain, 2024). I also showed that a cascade of sleep rhythms supporting motor memory is hijacked by epileptic spikes (PNAS, 2026). This work was recognised with the American Epilepsy Society Young Investigator Award (2024), an AES Fellowship (2023) and an Epilepsy Foundation New England Blue Skies award as co-PI (2023).

Teaching and mentoring. I coordinate quantitative courses across three bachelor's programmes (Data Science & AI, Computer Science, Brain Science). I bring students in as collaborators rather than assistants. They have first-authored work at EUSIPCO, BNAIC and FENS, several came back to do their master's theses with me.

Open science. I treat released code as part of the result, because a real-time method nobody can inspect cannot be trusted. I publish open access wherever I can, and I write for a general audience about technology and society, including a piece in The Hindu on what we lose with "giant AI".

Before Maastricht, I was a postdoctoral researcher at Boston University and at Massachusetts General Hospital / Harvard Medical School, after a PhD at UC Irvine with Ramesh Srinivasan.

Research

My research sits at the intersection of systems neuroscience and real-time signal processing, with recurring themes of oscillations, networks, epilepsy and sleep. A few directions I am currently excited about:

The difficulty of estimating networks from time-series

Estimating networks is critical across a range of fields but especially in neuroscience. Our current work looks at how we can estimate networks from point-processes such as those that represent epileptic spike events.

State-space modeling for non-sinusoidal and bursty rhythms

By modeling oscillations as damped harmonic oscillators with time-varying frequencies driven by noise, a state-space framework lets us track — in real time — rhythms that are poorly represented as simple band-limited oscillations.

Constraints from sleep oscillations

Sleep rhythms have been implicated in overnight memory consolidation. How can we quantify this in a way that gives explanatory power to the rhythms themselves, rather than only to the neuronal spikes that support them?

Publications

My work grouped by theme, newest first within each group. For citation counts, see my Google Scholar profile. (My name is in bold; titles link to the freely available version where there is one.)

Tracking rhythms in real time

  1. Conference Real-time sub-cycle oscillatory beta burst detection. Wodeyar A, Karel J, Peeters R. IEEE EMBC 2026 (oral presentation). Adds a switching layer to the state-space rhythm model so that beta bursts in Parkinson's disease can be detected within a single cycle, as the data arrive.
  2. Conference Real-time navigational intent detection from hippocampal EEG: a proof of concept. Savvides N, Bonizzi P, Herff C, Wodeyar A. 37th Benelux Conference on Artificial Intelligence (BNAIC), 2025. A student-led proof of concept that decodes where a person intends to navigate from hippocampal EEG in real time.
  3. Different methods to estimate the phase of neural rhythms agree but only during times of low uncertainty. Wodeyar A, Marshall FA, Chu CJ, Eden UT, Kramer MA. eNeuro, 10(11), 2023. Shows that phase is not uniquely defined for real signals: standard estimators disagree whenever uncertainty is high, so closed-loop use needs uncertainty estimates.
  4. A state-space modeling approach to real-time phase estimation. Wodeyar A, Schatza M, Widge AS, Eden UT, Kramer MA. eLife, 10:e68803, 2021. Introduces the state-space phase estimator: each rhythm is tracked as a latent oscillator, giving its phase sample by sample, with uncertainty and without windowing.

Inferring brain networks

  1. Conference Directed epileptic spike network inference with group-lasso regularised point-process GLMs. Delfo Furno LL, Kelly L, Tsikhanovich G, Savvides N, Pradas A, Sankar S, Gommer E, Karel J, Wodeyar A. 34th European Signal Processing Conference (EUSIPCO), Bruges, 2026. Student-led work that infers directed networks from sparse epileptic spike trains more accurately than Granger-causality baselines, to help localise where seizures start.
  2. Structural connectome constrained graphical lasso for MEG partial coherence. Wodeyar A, Srinivasan R. Network Neuroscience, 6(4):1219–1242, 2022. Uses the brain's anatomical wiring as a constraint in a graphical lasso to estimate direct (partial) coherence between MEG sources.
  3. Damage to the structural connectome reflected in resting-state fMRI functional connectivity. Wodeyar A, Cassidy JM, Cramer SC, Srinivasan R. Network Neuroscience, 4(4):1197–1218, 2020. A generative structure–function model showing how stroke damage to the structural connectome shows up in resting-state fMRI connectivity.
  4. Preprint Network structure during encoding predicts working memory performance. Wodeyar A, Srinivasan R. bioRxiv, 2018. Shows that the structure of EEG functional networks while items are being encoded predicts how well they are later held in working memory.
  5. Manuscript Functional connectivity using complex-Gaussian graphical models of EEG. Wodeyar A, Srinivasan R. 2017. Graphical models for complex-valued EEG Fourier coefficients that separate direct from indirect functional connections.

Sleep rhythms and epilepsy

  1. A hierarchical cascade of sleep rhythms supports motor memory and is hijacked by epileptic spikes in human epilepsy. Wodeyar A, Chinappen D, Kwon H, Shi W, Richardson RM, Kramer MA, Chu CJ. PNAS, 123(27):e2517454123, 2026. Free preprint Shows that slow oscillations, spindles and ripples form a cascade across brain regions that supports motor memory, and that epileptic spikes hijack it. Recognised with the AES Young Investigator Award.
  2. Thalamic engagement by epileptic spikes as a mechanism for widespread slow oscillation–spindle decoupling. Wodeyar A, Kramer MA, Chu CJ. Epilepsia, 66(7):2600, 2025. Shows that when a thalamic epileptic spike coincides with a cortical slow oscillation, thalamic spindles are suppressed, across epilepsy types and ages.
  3. Auditory-evoked changes in slow oscillations and spindles correlate with memory consolidation in children with epilepsy and controls. Kwon H, Chinappen DM, Wodeyar A, Kinard EA, Goodman SK, Shi W, Baxter BS, Manoach DS, Kramer MA, Chu CJ. Clinical Neurophysiology, 2025. Sound played during sleep changes slow oscillations and spindles, and the size of those changes tracks overnight memory consolidation in children with and without epilepsy.
  4. Thalamic epileptic spikes disrupt sleep spindles in patients with epileptic encephalopathy. Wodeyar A, Chinappen D, Mylonas D, Baxter B, Manoach DS, Eden UT, Kramer MA, Chu CJ. Brain, 147(8):2803–2816, 2024. Uses point-process models of thalamic and cortical recordings to show that epileptic spikes in the thalamus disrupt the sleep spindles thought to support memory.

Stroke injury and recovery

  1. Preprint Structural and EEG motor networks distinguish level of motor impairment after stroke. Zhou Z, Wodeyar A, Cramer SC, Srinivasan R. medRxiv, 2025. Combines structural and EEG motor-network measures to tell apart patients with different levels of motor impairment after stroke.
  2. Sensory stimulation-based protection from impending stroke following MCA occlusion is correlated with desynchronization of widespread spontaneous local field potentials. Rasheed W, Wodeyar A, Srinivasan R, Frostig RD. Scientific Reports, 12(1):1744, 2022. In an animal model, sensory stimulation that protects against an impending stroke goes together with desynchronised cortical activity.
  3. Coherent neural oscillations inform early stroke motor recovery. Cassidy JM, Wodeyar A, Srinivasan R, Cramer SC. Human Brain Mapping, 42(17):5636–5647, 2021. EEG coherence in the motor network in the first days after stroke helps predict how much motor function patients recover.
  4. Low-frequency oscillations are a biomarker of injury and recovery after stroke. Cassidy JM, Wodeyar A, Wu J, Kaur K, Masuda AK, Srinivasan R, Cramer SC. Stroke, 51(5):1442–1450, 2020. Establishes low-frequency EEG oscillations as a marker of both the injury caused by stroke and the recovery that follows.
  5. Rapid development of strong, persistent, spatiotemporally extensive cortical synchrony and underlying oscillations following acute MCA focal ischemia. Wann EG, Wodeyar A, Srinivasan R, Frostig RD. Scientific Reports, 10(1):21441, 2020. In an animal model, an acute ischemic stroke quickly produces strong, long-lasting synchrony and oscillations across wide areas of cortex.

Thesis and public writing

  1. Opinion What we lose when we work with a 'giant AI' like ChatGPT. Wodeyar A. The Hindu, 28 May 2023. Argues that building large language models top-down flattens the local knowledge that knowledge work depends on, and that we need many differently incentivised models rather than one giant one.
  2. PhD thesis Linking structure to function in resting state macroscale neural activity. Wodeyar A. University of California, Irvine, 2019. How the brain's anatomical wiring shapes the large-scale functional networks measured at rest.

Contact

I am always happy to talk about oscillations, real-time methods, and potential collaborations or student projects.