Pre-trained volumetric deep convolutional neural network for on MRIs of 43 stroke patients and 4 healthy individuals in the task of segmentation into 7 classes (gray matter, white matter, CSF, bone, air-cavities, skin and background), surpassing human performance. Code also available for training model for new tasks.
A fully automated Realistic vOlumetric-Approach to Simulate Transcranial electric stimulation. This is an open-source tool that runs on MATLAB and calls open-source C software packages such as iso2mesh and getDP. Starting from an MRI structural image, it segments the full head, places virtual electrodes, generates an FEM mesh and solves for voltage and electric field distribution -- at 1 mm resolution all this in about 15 minutes.
storEEG is system for storing and managing large amounts of EEG data. It allows for search and selection of EEG datasets, easy exporting and epoching of data, automated management of file structures, and easy-to-use interface to label data files. Base on the standard Brain Imagin Data Structure (BIDS).
We measures inter-subject correlationto (ISC) analyze stimulus responses in EEG in the absence of regressors or time markers. We use this to understand the responses to video and auditory narratives.
We developed Correlated Component Analysis (CorrCA) for the purpose of identifying components in the EEG with high inter-subject correlation. The technique is useful in any context where one would like to identify linear components with high reproducibility (across subjects, repeats, raters, etc).
This methods can identify what the brain encodes about the stimulus while simultaneously decoding the corresponding brain activity. It is a component extraction technique that works in the absence of precise time markers. We use this to understand the EEG responses to unique experiences, such as video games.
This is an anatomically detailed segmentation of the MNI standard head extended to include structures relevant to current-flow models (CSF, skull, scalp, neck). As we show in a recent paper, the resulting lead field is identical to the TES forward model and can thus be used for TES targeting as well as EEG source localization.
Individualized current-flow modeling starts with the MRI of an individual subject. The first task is to accurately segment the anatomy. These are a few tools based on SPM8, which we have developed for fully automated whole-head segmentation (not just brain).
Collaborating with Dr. Anli Liu at NYU Medical Center, we measured electric fields in vivo intracranially during TES on 10 subjects (~1300 electrodes). These data are used to validate and calibrate the computational models of TES. You can download these data to benchmark your own models.
This is a small software compiled from Matlab that can be used to quickly simulate tDCS on a sphere. Users can adjust the thickness of brain, CSF, skull and scalp, as well as their conductivities to see how these parameters affect the current-flow patterns inside the brain.