Project Ideas

For now these are only ideas. You can help make them happen.

Research Data Management with DataLad

Research data management is crucial for projects of any size. This tutorial introduces you to DataLad, an open-source solution for decentralized data management that is build atop Git. DataLad's features make it a perfect fit for tracking your research output, whether it's produced on your laptop or in "the cloud". DataLad helps you get ready for open-science from the very start.

BIDS-app-Based Searchlight Analysis Pipeline

Decoding analyses using a traveling "searchlight" are a widely used approach to map the availability of a given signal, and can also be used for a representational similarity analysis (RSA). This project aims to produce a BIDS-app for a PyMVPA-based searchlight analysis for decoding and RSA.

Basics of Image Reconstruction: From Raw Data to the Image

In this project, a basic reconstruction pipeline is to be developed for Cartesian sampled gradient echo data. This includes reading raw data, k space filtering, data synthesis in case of partial Fourier, coil combination (sum of squares and adaptive combination) and writing of image files. The aim is to reconstruct image with at least the quality of the vendors.

Containerized Nipype Pipelines on the Cluster

This project proposes taking custom or already available pipelines developed using Nipype, package them in a container using Docker or Singularity, and integrate them with cluster software (SLURM, SGE, Condor). The aim is to automate the process of taking those pipelines to a cluster as much as possible, either by developing scripts or extending Nipype's functionality.

Machine Learning with Dask and Neuroimaging Data

The aim of this project is to analyse large neuroimaging datasets — that do not fit in memory — even without access to powerful machines or clusters. To achieve this, we can use machine learning libraries (Nilearn, PyMVPA, scikit-learn) together with a library like Dask. During the event, the goal is to integrate these tools, pick a large dataset, and analyse it on "modest" hardware (e.g. a laptop).

Motor Imagery Classification with Deep Neural Networks

Motor imagery represents frequency phenomenon of Event-Related Desynchronization (ERD) and Event-Related Synchronization (ERS) in the motor cortex. In this tutorial, you will jointly train a logistic regression and a deep neural network combined model (Deep&Wide Learning) to evaluate the performance on four classes of motor imagery data.

Brain Network Analysis Pipelines in Neuroimaging Data

This project will focus on the pipelines of brain network analysis when using EEG data with different kinds of graphics measures, from mathematics to practical. The aim is to help visualize the brain dynamic connectivity changes and understand the network structures of brain activities in a different perspective.

Highest Resolution T1-weighted In Vivo Human Brain MRI Data

In this tutorial, the so far highest resolution T1-weighted in vivo human brain MRI data will be presented. Details will be given on what the entire data set consists of, how it was acquired, how it will be extended in the near future, and what potential use cases of it are. Besides simply showing some astonishing images of the brain.

How to Open Data (at OvGU)

Sharing your data openly is fantastic and may be beneficial for everyone. However, to improve visibility and enable citeability, data has to be provided with a digital object identifier (DOI). This tutorial will guide you through the steps to successfully apply for a DOI at the OvGU university library. Starting from the application itself, how and where to host data, and the regulations. Non-OvGU-members will learn how your institute is able to provide DOIs.

Hear What I Hear: Reconstructing the Music in Our Heads

Recently, research groups have shown that it is possible to read images from a person's brain as they are seen using both fMRI and EEG through machine learning strategies. These strategies have however not yet been applied to auditory stimuli. It should theoretically also be possible to record imagined visual or auditory perceptions. This project aims to utilize the same principles to extract simple musical details through EEG, and potentially even imagined music.

Nipype Tutorial

This tutorial introduces you to Nipype, an open-source neuroimaging software written in Python. Nipype provides a unified way of interfacing with heterogeneous neuroimaging software like SPM, FSL, FreeSurfer, AFNI, ANTS, Camino, and many more. It allows users to create flexible and complex workflows very quickly, that then can be executed in parallel either locally or on any computational cluster.

3D-Printing for Science

Additive manufacturing provides exciting opportunities in the lab: need a prototype? Just print it! More of these super expensive small parts that evil ACME Corp overcharges you for? 3D-printers for the rescue! In this project we will take a look at the workflow of FDM 3D printing and how you can implement it in your lab.

Seeing is Believing: Connecting VR Environments to EEG Output Interpretation

For this project, we would like to see what EEG signals will be generated when different VR environments or scenarios are presented to the user. It is up to the creativity of the programmer to either generate interesting VR environments or directly import existing 360 degree photos or videos from online databases (e.g. YouTube). We will record EEG signals and attempt to interpret them in accordance to the VR scenarios being watched.

Now You See It, Now You Don't: Virtual Reality and Vision

Virtual reality enables us not only to create artificial environments, we can also manipulate what — and how — participants are seeing. For this project, we will try to implement artificial limitations to vision (with and without VR-eye-tracking), create basic environments to explore and interact with, and implement basic experimental procedures to collect behavioral data in Unity3D.

Find The Bug: The Other Human in The Loop

Recently, we conducted a pilot study with humans to find selection strategies for a machine learning task (active learning). In this project, we develop the concept of an advanced version of the previously used paper card game extending the existing study. Furthermore, we take a look into existing data from the previous study to state interesting research hypotheses.