Lab Meeting

Published

April 17, 2026

HEMI-SLIM

  • First draft of manuscript complete, plan to submit in next few weeks
  • “Spin off” idea: quantitative PET extent

ring20

  • Collaborators secured at several centres (Italy, Vancouver, Dusseldorf, Munich, Innsbruck), total N~25
  • Waiting on legal for data sharing agreement

syn7T

All 3T and 7T scans have been downloaded, converted from DICOM to NIfTI using dcm2bids, and organized in valid BIDS format.

Next steps

  1. Preprocessing pipeline:
    • Resampling
    • Brain extraction
    • Coregistration of 3T to 7T space
  2. BME-X baseline model:
    • Run BME-X on native 3T as a “state of the art” baseline for image synthesis
    • Evaluate output quality using structural similarity index (SSIM) and visual information fidelity (VIF) as primary endpoints

The BME-X pipeline setup is underway in Docker. A session-level scans.tsv workaround is in place for BIDS age-parsing.

MRF-SEEG

  • PSD data from Balu on 8 patients
  • Within-channel total-power normalization
  • Contact-level EZ vs. non-EZ Spearman correlation framework
  • Mixed-effects modeling foundation
  • Heatmap approach refined (dropped at patient level due to small EZ contact counts)

- PyRadiomics

Next Steps

Comments

  • the AUC plot suggests there is some signal to the MRF data, and texture is important
  • spatial information lost when lumping things together - perhaps we should look by lobes for the heatmap

Things to think about

  • small # of EZ contacts, ROC curve may be inflating how good it looks consider precision-recall instead.

Action items