Functional Connectivity fMRI

• Objective: To identify activity levels that are synchronized over time between spatially distant parts (voxels) of the brain.

Lots of Noise!

• Goal: Remove noise from fMRI signals to better analyze the data without losing information.

  • Types of Noise:

    • Head motion

    • Displacement of the body

    • Other factors

• Techniques:

  • Motion scrubbing

  • Stringent motion cutoffs

  • Framewise displacement of motion

Goals

1. Create a GUI application that compares the quality of fMRI data processed with different pipeline choices.

2. Improve access to learning materials for students not familiar with programming.

3. Create clear visualizations of the denoising pipeline and show how changing confounds/features can alter the output.

4. Improve personal data science skills and learn how to launch a web application.

GUI Application

• Function: Compare the quality of fMRI data processed with different pipeline choices (confounds).

• Features/Confounds Created by FMRIPrep:

  • 12 head-motion parameters and derivatives (translation, rotation derivatives, etc.)

  • Global signals (csf, white_matter, global_signal)

  • Motion outliers (motion_outlier00 … motion_outlierN)

Metrics

• Schaefer Atlas:

  • Splits brain into ‘parcels’ (anterior temporal, auditory, central, etc.)

  • Each parcel is a Region of Interest (ROI).

• Quality Control-Function Connectivity Metric (QC-FC):

  • Evaluates the efficacy of the denoising pipeline.

  • Distance dependence of QC-FC correlations:

    • Determines correlations of QC-FC and the distance between centroids of brain regions.

    • Higher values indicate poorly controlled motion if QC-FC relationships are highly distance-dependent.

Future Plans

• Launch Completed GUI: Allow options to compare timeseries plots for different confounds.

• Improve Runtime: Optimize code performance.

• Assess Efficacy of Denoising: Use QC-FC correlations.

• Introduce GUI: Teach students how to effectively use the application for their studies.

What I Learned

• Progress:

  • Followed documentation effectively.

  • Gained domain knowledge in neurology to create a useful application.

  • Enhanced understanding of programming through articles and videos.

  • Improved communication with professors/mentors about deadlines and progress.

• Challenges:

  • Accountability during remote work.

  • Focusing on audience needs rather than personal assumptions.

  • Ensuring accessibility for students.

  • Creating innovative yet impactful demonstrations of data science applications.