(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: American College of Radiology (ACR) requires a Qualified Physicist to review a weekly Quality Control (QC) program, which can be challenging at multi-site institutions. This work set out to automate the requirement to increase compliance and consistency. This project focused on MRI weekly QC but could be modified for any DICOM modality.
Methods: server hosts a webpage for data entry by the Tech and acts as a PACS node for Query and Retrieve of ACR DICOMs. The tech completes a webform and pushes the DICOMs to the main PACS node. An XML file stores the Visual Checklist data, and a MATLAB script extracts the Large Phantom measurements from the Sagittal and Axial T1 scans, while pulling the Center Frequency and Transmit Gain from the header. Ghosting Ratio and Uniformity are done weekly instead of annually. The Large Phantom measurements are stored in an XML file, and the weekly QC numbers are added to an Excel spreadsheet by scanner name and emailed to the Physics and Operations teams weekly.
Results: automated system acts as a backup for all ACR weekly testing, and since the data entry becomes Pass/Fail errors are eliminated. Physicist Annual Exams were used as a gold standard. The average distance measurement error (excluding one air bubble “outlier”) was 0.7 mm, 34/36 High Resolution sections were correctly identified (Y/N), all of the Low Contrast Detectability sections were identified, the average Uniformity percentage error was 1.5% and the average Ghosting Ratio error was 5e-4. A Ghosting Ratio failure alert enabled a faster call to the manufacturer, and the format makes tracking Center Frequency and Transmit Gain simple.
Conclusion: automated analysis is robust enough to use on a weekly basis for easier and more consistent QC measurements. The automated QC system increases compliance and provides a backup of ACR data.
Keywords
PACS, Quality Control, Image Processing
Taxonomy
IM- MRI : Quality Control and Image Quality Assessment
Contact Email
