Automated quality control for radiation therapy

(c) Santeri Rytky, Oulu University Hospital, 2024

Background

Multiple quality assurance tests are done periodically at radiation therapy clinics. The analysis of the measurements is often time-consuming and subject to errors from manual processing. This repository acts as a framework that automatically analyses images from a given data folder. The main dependencies are the Pylinac package for calculating the analyses and the Watchdog package for monitoring the data folder.

Requirements

There are small modifications to the Pylinac 3.22 package that should be done. These are likely fixed in future updates.

Fix for reading multiple CT series (Github issue)

Installation

• Anaconda installation

git clone https://github.com/MIPT-Oulu/automated-rt-qa.git
cd automated-rt-qa

Either create a conda environment for running the code

conda create --name rt-env --file requirements.txt

or install the packages using pip

pip install -r requirements.txt

Populate the input arguments from main.py and main_offline.py. If you need to map a network drive for the data folder, include a share.txt file. The file should include the drive letter, shared path, username and password. Example:

Z:
\\network\drive
user
pwd

Usage

The automated analysis is started using main.py. This could be automated for example with task scheduler in Windows systems. Other option is to run the analysis using main_offline.py, which runs the analysis pipeline once.

Features

Automated QA pipeline

The code lists the files of listed data types (default: .dcm, .tif, .tiff) in the folder, and runs the analysis for each measurement date (Series date) and Patient ID (test patient). The code detects the available test types and runs all the different measurements found. The results are saved either as a pdf report and/or a row in an Excel file.

Available tests

• VMAT (T2/T3)
• Catphan analysis
• ACR phantom analysis (for MRI)
• Winston-Lutz

Logging

Different events during the analysis pipeline are logged in the repository root.

Assumptions

The pipeline is tested for Varian Truebeam and Halcyon accelerators.

VMAT (T2/T3) analysis

The pipeline assumes that T2 images have RT label MV_243. In T3 tests the label is MV_32 or MV_190. Open beam images have the 5000, 2500 (Curve label) DICOM tag.

For Halcyon, the T2 test is split into two separate tests. The Halcyon dose-rate test is detected from a lower exposure (less than 100 MU). Halcyon tests are conducted with custom regions of interest (constants.py).

Catphan analysis

The DICOM modality attribute is CT. CBCT images have either RT Plan Storage or RT Plan or RT Ion Plan or Radiation Set to be verified identifier. Diagnostic CT images are detected based on CT Image Storage identifier. Phantom is oriented head first supine (HFS). If feet first supine is planned, the code inverts the stack Z-axis.

ACR analysis

Currently only passed for MR images. The modality should be MR.

Winston-Lutz analysis

Modality should be RTIMAGE. The test is tried for images that are not labeled as other test types.

License

This software is distributed under the MIT License.