T1w Fractal Dimension Repository

This repository contains de-identified example data and resources for computing fractal dimensions from T1-weighted MRI scans. It includes code and patient-derived, de-identified data for analysis and processing. The repository is part of ongoing research in the Radiology-Morrison-lab-UCSF.

Contents

• ExamplePatient_FractalDimension_ROIS.csv:
  • A CSV file containing fractal dimension values for 90 regions of interest (ROIs) based on the AAL atlas.
• ExamplePatient_T1w_brain.nii.gz:
  • A de-identified, skull-stripped T1-weighted MRI scan in NIfTI format with identifying features (ears, nose, mouth) removed.
• LEDD_Score_ExamplePatient.xlsx:
  • An Excel file with LEDD scores for the example patient.
• MDSUPDRS_Scores_ExamplePatient.xlsx:
  • An Excel file with MDS-UPDRS scores for the example patient.
• FeatureSelection_and_ClassificatioinLearning.ipynb:
  • A jupyter notebook (python 3.7) that performs feature selection and also performs a classification learning task using HGNNs from the DHG package in PyTorch
• T1_metrics.xlsx:
  • An Excel file with all features and charecteristics used in the predictive modeling for our cohort.

Data De-Identification

The included example patient data is derived from a real patient but has been de-identified to comply with ethical guidelines:

• All dates (e.g., surgeries and follow-up visits) have been altered, but the time intervals between events remain consistent to ensure accurate calculations.
• The T1-weighted MRI scan has been skull-stripped to remove identifiable features, such as ears, nose, and mouth.
• Metadata an

Future Additions

This repository will soon include:

• Code:
  • Scripts for processing weighted MRI images and calculating fractal dimensions.
  • Statistical analysis scripts for downstream interpretation and visualization.
• Data:
  • De-identified MRI scans and associated metadata.
  • All data will be made publicly available through OpenNeuro upon completion of ongoing analyses.

License

• The code in this repository is licensed under the MIT License.
• The de-identified data in this repository is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) License.

Acknowledgments

This project is part of ongoing research in the Radiology-Morrison-lab-UCSF. Special thanks to all contributors and collaborators involved in this work.