About

Todd Parrish is a Professor of Radiology and Biomedical Engineering at Northwestern University. His research interests include magnetic resonance imaging, functional neuroimaging, MR cardiac imaging, and clinical sequence and protocol development. Parrish has contributed extensively to understanding brain activity related to chronic pain, social hierarchy, speech perception, insight facilitation, sleep deprivation, odor discrimination, familiarity and priming, verb argument processing, linguistic pitch learning, sexual arousal, stroke-induced aphasia, language network alterations, and chronic pain, among other topics. His work involves exploring neural mechanisms through advanced imaging techniques, and he has authored numerous publications that investigate the neural correlates of various cognitive and emotional processes.

Research topics

• Computer Science
• Medicine
• Neuroscience
• Artificial Intelligence
• Psychology
• Audiology
• Data Mining
• Pathology
• Radiology
• Anatomy
• Cognitive psychology
• Internal medicine
• Statistics
• Nuclear medicine
• Mathematics
• Medical physics
• Speech recognition

Selected publications

• MRI mapping of hemodynamics in the human spinal cord

  OpenNeuro · 2026-01-01

  datasetOpen access
  PublisherDOI

• MRI mapping of hemodynamics in the human spinal cord

  OpenNeuro · 2026-01-01

  datasetOpen access
  PublisherDOI

• 2051 Age and Sex-Specific Morphometric Predictors of Disease Progression and Surgical Outcomes in Degenerative Cervical Myelopathy

  Neurosurgery · 2026-03-26

  article
  PublisherDOI

• A Customized Neural Transcranial Magnetic Stimulation Target for Functional Disability Among Veterans With Co-Occurring Alcohol Use Disorder and Mild Traumatic Brain Injury: Protocol for a Pilot Randomized Controlled Trial

  JMIR Research Protocols · 2025-03-24

  articleOpen access

  BACKGROUND: Alcohol use disorder (AUD) and mild traumatic brain injury (mTBI) commonly co-occur, exacerbating symptoms and negatively impacting function. Co-occurring AUD and mTBI (AUD+mTBI) represents a unique and heterogeneous brain state impacting symptoms and function. Repetitive transcranial magnetic stimulation (rTMS) is a noninvasive neuromodulatory treatment with burgeoning evidence for improving brain function and symptoms for AUD and mTBI each alone. However, there is no consensus on the optimal neural target or treatment site of stimulation for either condition alone or when they co-occur. OBJECTIVE: This study aims to (1) test the preliminary efficacy of high-frequency rTMS provided to a customized neural target to treat functional disability among veterans with AUD+mTBI and (2) assess the sustainability of rTMS effects on functional disability among veterans with AUD+mTBI. METHODS: This single-blind randomized controlled trial (RCT) involved treatment-seeking veterans with AUD+mTBI recruited from a Department of Veterans Affairs hospital. Veterans will be randomly assigned to (1) a novel TMS target site using neuronavigation or (2) standard clinical left dorsolateral prefrontal cortex using the Beam F3 method. All participants first receive 10 daily sessions of sham rTMS, followed by 10 daily sessions of active rTMS (10 Hz) provided by a trained TMS administrator. Veterans will complete self-report study questionnaires before and after sham and active rTMS session blocks as well as at 2-week, 1-month, 3-month, and 6-month posttreatment follow-up time points. The primary outcome is the WHO Disability Assessment Schedule 2.0. The secondary outcomes include alcohol use on the Timeline Follow-Back calendar, the Penn Alcohol Craving Scale, the Obsessive-Compulsive Drinking Scale, the Alcohol Craving Questionnaire, the Neurobehavioral Symptom Inventory, the PTSD Checklist for DSM-5, the Beck Depression Inventory-II, the Beck Anxiety Inventory, and the Mood and Anxiety Symptom Questionnaire. RESULTS: This study protocol was approved by the institutional review board of the Edward Hines Jr Department of Veterans Affairs Hospital (19-021). This study includes a Food and Drug Administration investigational device exemption (G180292). A 6-year research plan timeline was developed, including cost and no-cost extensions due to the COVID-19 pandemic. As of June 2025, overall, 27 veterans with AUD+mTBI who had been enrolled in the study had completed the neural target identification phase and were awaiting enrollment in the RCT phase. Data collection for the RCT phase will be initiated soon and is expected to be completed by April 2026. We expect the results of this study to be available by May 2027. CONCLUSIONS: We will be able to provide preliminary evidence of the efficacy, safety, and feasibility of a novel TMS target for veterans with AUD+mTBI. TRIAL REGISTRATION: ClinicalTrials.gov NCT04043442; https://www.clinicaltrials.gov/study/NCT04043442.

  PublisherDOI

• Automated Imaging Differentiation for Parkinsonism

  JAMA Neurology · 2025-03-17 · 36 citations

  articleOpen access

  Importance: Magnetic resonance imaging (MRI) paired with appropriate disease-specific machine learning holds promise for the clinical differentiation of Parkinson disease (PD), multiple system atrophy (MSA) parkinsonian variant, and progressive supranuclear palsy (PSP). A prospective study is needed to test whether the approach meets primary end points to be considered in a diagnostic workup. Objective: To assess the discriminative performance of Automated Imaging Differentiation for Parkinsonism (AIDP) using 3-T diffusion MRI and support vector machine (SVM) learning. Design, Setting, and Participants: This was a prospective, multicenter cohort study conducted from July 2021 to January 2024 across 21 Parkinson Study Group sites (US/Canada). Included were patients with PD, MSA, and PSP with established criteria and unanimous agreement in the clinical diagnosis among 3 independent, blinded neurologists who specialize in movement disorders. Patients were assigned to a training set or an independent testing set. Exposure: MRI. Main Outcomes and Measures: Area under the receiver operating characteristic curve (AUROC) in the testing set for primary model end points of PD vs atypical parkinsonism, MSA vs PSP, PD vs MSA, and PD vs PSP. AIDP was also paired with antemortem MRI to test against postmortem neuropathology in a subset of autopsy cases. Results: A total of 316 patients were screened and 249 patients (mean [SD] age, 67.8 [7.7] years; 155 male [62.2%]) met inclusion criteria. Of these patients, 99 had PD, 53 had MSA, and 97 had PSP. A retrospective cohort of 396 patients (mean [SD] age, 65.8 [8.9] years; 234 male [59.1%]) was also included. Of these patients, 211 had PD, 98 had MSA, and 87 had PSP. Patients were assigned to the training set (78%; 104 prospective, 396 retrospective) or independent testing set, which included 145 (22%; 60 PD, 27 MSA, 58 PSP) prospective patients (mean age, 67.4 [SD 7.7] years; 95 male [65.5%]). The model was robust in differentiating PD vs atypical parkinsonism (AUROC, 0.96; 95% CI, 0.93-0.99; positive predictive value [PPV], 0.91; negative predictive value [NPV], 0.83), MSA vs PSP (AUROC, 0.98; 95% CI, 0.96-1.00; PPV, 0.98; NPV, 0.81), PD vs MSA (AUROC, 0.98; 95% CI, 0.96-1.00; PPV, 0.97; NPV, 0.97), and PD vs PSP (AUROC, 0.98; 95% CI, 0.96-1.00; PPV, 0.92; NPV, 0.98). AIDP predictions were confirmed neuropathologically in 46 of 49 brains (93.9%). Conclusions and Relevance: This prospective multicenter cohort study of AIDP met its primary end points. Results suggest using AIDP in the diagnostic workup for common parkinsonian syndromes.

  PublisherOA PDFDOI

• The White Matter of Aha! Moments

  Research Square · 2025-05-22

  preprintOpen access
  PublisherOA PDFDOI

• Amygdala parcellation reveals unique subregions of atrophy in FTLD‐tauopathies

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Primary progressive aphasia (PPA) is a dementia syndrome characterized by language decline that can be caused by frontotemporal lobar degeneration with tau pathology (FTLD-tau), involving three or four microtubule binding repeat domains (i.e., 3R and 4R tau). Evidence suggests that differential atrophy in limbic regions such as the amygdala may play a role in the clinical heterogeneity of 3R and 4R FTLD-tau. This study examined the distribution of atrophy across amygdala subregions in these tauopathies. METHOD: Eighteen right-handed PPA cases and 35 age-matched controls from the Northwestern PPA Program underwent structural MRI. PPA cases were chosen based on postmortem pathology of 3R (N = 9) or 4R (N = 9) FTLD-tauopathy. T1-weighted images were acquired on a Siemens 3T scanner and underwent FreeSurfer (v7) processing. Amygdala nuclei were parcellated using FreeSurfer's amygdala subregion module and normalized to each individual's estimated total intracranial volume. Mixed linear regressions, adjusting for disease duration and incorporating random intercepts, were used to compare amygdala subregion volumes between tauopathies and controls. RESULT: PPA/3R had a mean age of 63.5 years (SD = 3.9), PPA/4R were 72.5 years (SD = 5.8), and controls were 63.4 years (SD = 6.8). In the left hemisphere, PPA/3R cases had smaller mean volumes than controls in the whole amygdala (p = 0.001) and all nuclei except the medial and paralaminar nuclei (i.e., lateral, basal, accessory basal, anterior amygdaloid area, cortical, central, and corticoamygdaloid transition area, all p <0.05). PPA/4R cases demonstrated smaller volumes than controls in only the paralaminar (p = 0.03) and corticoamygdaloid transition area (p = 0.009). Compared to PPA/4R, PPA/3R cases had reduced volumes in the whole amygdala (p = 0.04), lateral (p = 0.04), and cortical (p = 0.03) nuclei. There were no significant differences between disease groups in the right hemisphere. CONCLUSION: Results highlight the leftward asymmetry of disease in PPA and demonstrate increased atrophy of the amygdala in PPA/3R. Further, distinct portions of the amygdala were implicated in 3R versus 4R FTLD- tauopathy. Findings support prior research showing divergent neuropsychiatric phenotypes in FTLD-tauopathies, which may result from disruptions in amygdalar circuits governing emotional and behavioral processes. Future functional connectivity studies could offer greater insight into the pathways of limbic disruption in tauopathies.

  PublisherOA PDFDOI

• A novel focus operator from statistical dispersion of deconvolution error for infrared focus ranging of low-contrast regions at meter-order distances

  2025-06-23

  preprintOpen accessSenior author

  In monocular variable-focus imaging systems, focus operators are key to autofocus and passive distance estimation. However, canonical operators including the gradient 1-norm and Laplacian 1-norm lose robustness as distance degrades contrast, exacerbated by thermal diffusion in long-wave infrared thermography. To address this limitation, we developed a novel focus operator, K, which generates a scalar cost function. The cost function is the standard deviation in Kullback-Leibler divergence between each image slice in a focus sweep and a set of deconvolved slices simulating the entire focus sweep as a function of assumed object distance. Deconvolution is performed with Gaussian beam point spread functions (PSFs) derived from the paraxial Helmholtz equation. The K operator locates focus by identifying the assumed depth minimizing KL divergence dispersion, reflecting the depth where the optical model introduces the least sensitivity to model mismatch. When benchmarked against canonical focus operators using focus sweeps centered about the distance of a weakly textured target, K demonstrated highly competent accuracy and markedly lower standard deviation in depth estimation error, with remarkable performance at farther, meter-order ranges. A second, computational experiment truncated focus sweeps from the benchmark experiment so the thermal target was not centered in the sweep range. Our results suggest that K enables robust, model-based focus estimation in conditions where classical focus operators may lose reliability, with accuracy and precision improved in cases where the target appears early in the sweep range. Because the method relies solely on an analytical PSF model, it is in principle wavelength-agnostic and may generalize to other monocular imaging systems with known optical geometry and a tractable PSF model.

  PublisherOA PDFDOI

• Intramedullary Lesions and Perilesional Tissue Bridges Predict Motor Recovery in Degenerative Cervical Myelopathy: A Multicenter Study

  medRxiv · 2025-08-07

  preprintOpen access

  ABSTRACT Background and Objectives Degenerative cervical myelopathy (DCM) is the leading cause of spinal cord-related disability worldwide. T2-weighted (T2w) intramedullary hyperintensities are common MRI findings in DCM and may reflect irreversible pathology, namely intramedullary lesions. The prognostic value of both quantitative lesion characteristics and the integrity of surrounding spared spinal cord tissue—tissue bridges—remains unclear. Using automated quantification of intramedullary lesions and tissue bridges, this study aimed to determine whether these structures can serve as imaging biomarkers of clinical severity and predict functional recovery in DCM. Methods This retrospective, multicenter study included 94 patients with DCM from four tertiary centers. All underwent baseline MRI and clinical assessment, with 6-month follow-up in 48.9% (n=46) of patients at six months. We used SCIsegV2, an automated tool integrated in the Spinal Cord Toolbox, to segment T2w hyperintense lesions and midsagittal tissue bridges. Lesion characteristics (volume, length, width, maximal axial damage ratio) and tissue bridge widths were extracted. Clinical assessments included mJOA, Neck Disability Index (NDI), hand dexterity, and balance. We evaluated associations between imaging metrics and clinical outcomes by using correlation and multivariable linear regression models adjusted for baseline clinical score, age, sex, and surgical status. Results Intramedullary lesions were present in 45.7% (n=43) of patients and were associated with greater clinical severity. Lesion volume, length, and maximal axial damage ratio (MADR) correlated with dexterity and balance impairments, but not with mJOA or NDI. Wider tissue bridges were associated with better dexterity at both time points. Multivariable models showed lesion volume and MADR independently predicted poorer balance at follow-up, while tissue bridge width was positively linked to with dexterity improvements. In surgical patients, lesion magnitude and tissue bridge integrity explained up to 71% of the variance in follow-up outcomes. Discussion Automated quantification of intramedullary lesion extent and spared tissue bridges provide robust biomarkers for structural damage and preserved function in DCM. These features better predict recovery, particularly in dexterity and balance, than conventional metrics. Integrating these metrics into clinical workflows could enhance surgical decision-making and support personalized prognosis. Future studies should incorporate 3D segmentation and multimodal imaging to refine predictions of long-term outcomes prediction.

  PublisherOA PDFDOI

• EPISeg: Automatic Segmentation of Spinal Cord fMRI Data

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  articleOpen access

  Motivation: Spinal cord fMRI holds promise for somatosensory and motor research, but challenges in data acquisition and preprocessing limit its potential. Goal(s): Develop an automated segmentation tool for spinal cord fMRI data that minimizes manual intervention and improves segmentation accuracy which is important for data analysis. Approach: We introduce EPISeg, a deep learning-based segmentation method trained on an open-source multi-site dataset of 406 subjects using active learning with human-in-the-loop feedback. Results: EPISeg outperformed established methods like PropSeg, DeepSeg, and a Contrast-agnostic model, achieving a Dice score of 0.88. Impact: EPISeg significantly enhances spinal fMRI research by enabling automated, accurate segmentations of EPI data, overcoming limitations of manual segmentation. Its integration into SCT broadens accessibility and reproducibility, facilitating robust group-level analyses essential for advancing studies of spinal processes and disorders.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R21AT000401

  NIH · $310k · 2003

• NIH Grant R01EB002449

  NIH · $1.5M · 2008

• BIGDATA: IA: Collaborative Research: Asynchronous Distributed Machine Learning Framework for Multi-Site Collaborative Brain Big Data Mining

  NSF · $120k · 2019–2022

Frequent coauthors

• James Higgins

  Northwestern University

  547 shared

• Theresa Pape

  Northwestern University

  542 shared

• Amy A. Herrold

  541 shared

• Ann Guernon

  Lewis University

  540 shared

• Joshua M. Rosenow

  Northwestern University

  538 shared

• Sherri Livengood

  535 shared

• Dulal K. Bhaumik

  535 shared

• Trudy Mallinson

  George Washington University

  534 shared