Research

A fundamental problem in psychiatry is that there are no biological markers for diagnosing mental illness or for indicating how best to treat it. Treatment decisions are based entirely on symptoms, and doctors and their patients will typically try one treatment, then if it does not work, try another, and perhaps another. Our group hopes to change this picture, and our research suggests that individual brain scans and speaking patterns can hold valuable information for guiding psychiatrists and patients. Current areas include depression, suicide, anxiety disorders, autism, Parkinson disease, and brain tumors.

To support this broader goal, our group develops novel analytic platforms that use such information to create robust, predictive models around human health. We believe that solving this problem will require complex integration of different types of sensors into an adaptive learning system together with patient, caregiver, and community feedback.

Our research interests span computer science and neuroscience, specifically in the areas of applied machine learning, signal processing, and translational medicine. Our current research portfolio comprises projects on spoken communication, brain imaging, and informatics to address gaps in scientific knowledge in three areas: the neural basis and translational applications of speaking, precision psychiatry and medicine, and preserving information for reproducible research.

Many of the tools we develop can be used across domains. If you have a need we can address, we would like to hear from you.

If you have solved problems associated with any of the projects below, we would love to hear from you. For us, a solution typically implies available data, code, and/or replicated results.

Highlights

For a full list see below

Full List

DANDI: Distributed Archives for Neurophysiology Data Integration
DANDI team (MIT, Dartmouth, Kitware)
A platform for data ingestion, search, and computing targeted towards cellular neurophysiology. Supported by NIMH R24MH117295 (PI: Ghosh and Halchenko).
Project website.

Nobrainer: A Robust And Validated Neural Netwrok Tool Suite For Imagers
Nobrainer team (MIT, MGH, GSU)
Develop an open Python library to simplify integrating deep learning into neuroimaging research. We will build and distribute user-friendly and cloud enabled end-user applications for the neuroimaging community.
Project summary from NIH.

Nipype: Dataflows for Reproducible Biomedical Research
Nipype team
A Python-based platform for scalable scientific workflow design and execution. Supported by NIBIB R01 EB020740.
Read more about Nipype

ReproNim: A Center for Reproducible Neuroimaging Computation
ReproNim team
A Center dedicated to improving reproducibility in neuroimaging through the development of products designed to enhance efficiency and provenance tracking within laboratories and in collaboration with external stakeholders such as data archives and publishers. Supported by NIBIB P41 EB019936 (PI: David Kennedy, UMass Medical School).
Read more about ReproNim

Neuroscout: A cloud-based platform for rapid re-analysis of naturalistic fMRI datasets
Neuroscout team
A platform for fast and flexible re-analysis of (naturalistic) fMRI studies. Supported by NIMH R01 MH109682 (PI: Tal Yarkoni, UTexas, Austin).
Neuroscout Website

Naturalistic Parcellation and audiovisual integration
Jeff Mentch
A Neuroscout project to generate cortical parcellations using naturalistic (movie viewing) neuroimaging datasets.
Project details

Voice and language biomarkers of mental health
Daniel Low
The goal of this project is to observe and analyze speech variations in psychiatric disorders by analyzing different types of data including ecological momentary assessments, recordings of clinical interviews, and social media posts. Key questions we hope to answer are: 1) Is speech a good longitudinal biomarker? 2) How much speech information is needed for stable tracking? 3) Can linguistic information augment voice models? 4) Can we create a neurocomputational model of such systems?
Project details

Detecting unilateral vocal fold paralysis
D Low, P Song, G Randolph, V Rao
The aim of this project was to develop a model to detect unilateral vocal fold paralysis from audio recording of voice.
Preprint

Subcortical anatomy of the auditory system
K Sitek, OF Gulban, F de Martino, GA Johnson, E Calabrese
The aim of this project is to identify brainstem structures and their connectivity. This is Kevin Sitek’s NIH F31 project involving high resolution imaging and analysis of post mortem and in vivo data.
Poster from OHBM 2019

Imaging genetics analysis of schizophrenia
D Burdinski, Y Sanchez-Araujo, K Batmanghelich, G Blokland, T Petryshen
This is a collaboration with the Genus Consortium to apply multivariate data mining and visualizatin tools to the GENUS dataset. An additional target is to support data harmonization in multi-site studies. Key questions are: 1) Do multivariate deterministic and probabilistic tools provide a more sensitive approach to detecting genetic markers of brain structure and cognitive variation? 2) Can a common vocabulary help organize and analyze multi-site data?
OHBM 2019 presentation

Using neurofeedback to reduce auditory hallucination in schizophrenia
S Whitfield-Gabrieli, C Bauer, K Okano, M Niznikiewicz, A van der Kouwe, P Wighton
This is a collaboration with the PEN Lab to use realtime fMRI biofeedback to to increase the positive diametric activity in patients with schizophrenia.
Description at PEN Lab

Mental well-being platform
D Low, D Burdinski, A Dulski, K Wang
The aim of this project is to create a platform outlining tasks that have been shown to improve well-being with the option of input by the user community. The platform includes an optional study protocol to monitor the effects of performing the tasks over time. The website will serve as the basis for future well-being studies.
Mentalwellbeing.us

Disease risk prediction from imaging genetics
D Burdinski
The aim of this project is to harness both genetic and brain imaging data from the Adolescent Brain Cognitive Development Study and the UK Biobank in order to design models that predict disease risk from genes and brain endophenotypes. The project will explore to what extent incorporating intermediate phenotypes that are more closely tied to biology and may capture environmental effects increases predictive ability compared to polygenic risk scores alone, and what additional insights can be gained with this method (eg. new loci discovery, better mechanistic understanding, delineation of new disease subtypes).

Vocallect: Detecting Speech Biomarkers of Communication Disorders
K Nguyen
The purpose of this study is to understand and classify the impacts that communication disorders may have on speech production and acoustic quality. Subjects will respond to questions regarding their current health, and provide sample voice recordings and free and read speech through a smartphone application and/or web browser. By collecting and correlating self-reported symptoms with audio features, we intend to develop and validate a voice-based technology using signal processing and machine learning that can be used for in-hospital and in-home monitoring of communication disorders, treatment, and recovery.
Project details

Mumble Melody: Musically Modulated Auditory Feedback to Increase Fluency for People Who Stutter
R Kleinberger, A Kodibagkar, M Vemuri
The aim of this project is to explore assistive technologies to help increase fluency for people who stutter. Our studies investigate the use of Altered Auditory Feedback (AAF), a stuttering intervention in which an individual’s speech is played back to them in an altered manner. Below are further details about this project as well as a page about our study.
Project details