MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Automated High Contrast Resolution Test Analysis for MRI Daily QC

J Jimenez*, W Stefan , J Yung , D Reeve , R Stafford , J Hazle , UT MD Anderson Cancer Center, Houston, TX

Presentations

(Wednesday, 7/17/2019) 10:30 AM - 11:00 AM

Room: Exhibit Hall | Forum 8

Purpose: Develop an algorithm that analyzes images for the weekly high-contrast spatial resolution test using the large phantom as part of the American College of Radiology (ACR) accreditation program to monitor the performance and stability of MRI systems.

Methods: A python function was developed that attempts to mimic the ACR grading method. First, the algorithm assesses phantom circumference and angle of rotation. Next, all three resolution grids are registered to a template and the histogram of either the upper left or lower right zones of the grid is used to create a binary mask to count the number of resolvable holes at each column or row. Finally, a lower signal threshold value is used from the histogram and this process is repeated up to eleven times before declaring failure for a particular resolution area. A dataset of 5,500 images from 30 MRI systems (two vendors, various platforms and field strengths) that had been previously scored manually using the ACR criteria were used to test and validate the algorithm. Run time is 2 seconds on a laptop.

Results: We measured a 95% agreement between our algorithm and manually scored images. Two primary reasons for disagreement were identified. First, when angle of rotation is large the center of the resolution area is not accurately registered which leads to an erroneous histogram analysis. Secondly, low contrast between resolution holes and background is not scored as differentiable by readers but is measurable by the thresholding porting of the function, leading to a higher score given by the algorithm.

Conclusion: A simple algorithm for automated assessment of the ACR high-contrast spatial resolution test was developed and validated over a large cohort of images. The algorithm not only spares labor but can also provide a more consistent grading approach for routine MRI quality control (QC).

Keywords

Resolution, Quality Control, Image Analysis

Taxonomy

IM- MRI : Quality Control

Contact Email