WedgeFieldDoseCal

Multichannel dose calibration based on scans of EDW Wedge fields.

Environment

uv venv
source .venv/bin/activate
uv pip install -e .

Usage

The API stitches together TIFF film scans and RTDOSE DICOM planes, performs rigid mutual-information registration, crops to the overlap, and fits a multichannel calibration model with dose-dependent covariance estimates.

from wedge_field_dose_cal.pipeline import calibrate_from_files, dose_from_tiff

# Provide matching TIFF scans and RTDOSE files (e.g., 500, 1000, 2000, 4000 MU deliveries)
tiff_paths = ["edw_0500.tiff", "edw_1000.tiff", "edw_2000.tiff", "edw_4000.tiff"]
dicom_paths = ["edw_0500.dcm", "edw_1000.dcm", "edw_2000.dcm", "edw_4000.dcm"]

calibrator = calibrate_from_files(tiff_paths, dicom_paths)

# Convert a new film scan to dose
predicted_dose = dose_from_tiff("unknown_field.tiff", calibrator)

Key steps

1. Load TIFF and DICOM data with physical pixel spacing.
2. Resample the DICOM dose grid to the film DPI.
3. Register dose to film via 2D rigid mutual information and crop the valid overlap.
4. Calibrate using log optical density across all three channels with dose-binned 3×3 covariance maps.
5. Predict dose for future films by weighting dose bins with Mahalanobis distances (covariance handling is internal).