CRCI Thesis Code – FC and SD-BOLD Analysis

This repository contains all code for the analysis of Functional Connectivity (FC) and BOLD Signal Variability (SD-BOLD) in resting-state fMRI data. It evaluates how these neural features relate to cognitive performance (memory and executive function) in healthy female participants, forming a baseline for identifying biomarkers of cancer-related cognitive impairment (CRCI).

Used for the bachelor thesis:
"Functional Connectivity and BOLD Variability as Neural Biomarkers for Cancer-Related Cognitive Impairment"
Author: Emre Pelzer | Maastricht University | 2025

---

What This Code Does

1. Preprocesses raw rs-fMRI in native functional space (motion correction, 6mm smoothing).
2. Extracts Functional Connectivity (FC) using seed-based correlation from 13 brain regions (DMN, FPN, SN, hippocampus).
3. Extracts SD-BOLD by computing voxelwise signal variability.
4. Performs cognitive analyses:
   • Correlations with CVLT1 (memory) and SES (executive function)
   • Group comparisons between high vs. low memory performers
5. Outputs:
   • Per-subject CSV files with extracted features
   • Statistical test results
   • Figures and summary plots

---

How to Run

1. Install dependencies

pip install -r requirements.txt

2. Prepare your data

• Download from: OpenNeuro ds004796 – PEARL-Neuro Dataset
• Select female participants with task-rest_dir-PA_bold.nii.gz
• Organize data in BIDS format under /data/

3. Preprocess

python generate_sdbold_per_subject_resampled.py

4. Extract features

python generate_fc_features.py

5. Analyze correlations

python analyze_fc_cognition.py
python correlate_sdbold_scores.py

6. Group comparisons

python analyze_fc_cvlt_groups.py
python analyze_sdbold_cvlt_groups.py

---

Folder Structure

• /code/: all scripts
• /data/: raw data in BIDS format (publicly available via OpenNeuro)

---

Notes

• Preprocessing is done without T1w normalization
• SD-BOLD is computed voxelwise and averaged per ROI
• FC values are Fisher z-transformed
• Only female participants included to match CRCI population
• Cognitive scores (CVLT1, SES) are z-normalized

---

License

Creative Commons BY-NC-ND 4.0 – see LICENSE file for details. If reusing thesis text or figures, please credit the author.

---

Contact

Emre Pelzer
B.Sc. Data Science & AI – Maastricht University Mail: emrepel03@gmail.com