Algorithm development for automatic laser alignment assessment on an ACR CT phantom and its evaluation on sixteen CT scanners

Abstract

Purpose. The aim of this study is to develop software to automatically assess the laser alignment on the ACR CT phantom and evaluate its accuracy on sixteen CT scanners. Methods. Software for an automated method of laser alignment assessment on the ACR CT phantom was developed. Laser alignment assessment was based on the positions of the ball-bearing markers at the edge of the ACR CT phantom. The automatic assessment was performed using several steps, including segmentation to acquire the coordinates of the ball-bearing markers and determination of the distances between lines connecting them with lines through the center of the image. A comparison of the results from the automatic method with those from the manual method was performed. The manual measurements were carried out using MicroDicom Viewer. A Mann–Whitney U test was performed to determine the statistical difference between both methods. The evaluation was performed on images of the ACR CT phantom scanned with 16 CT scanners from 5 different CT manufacturers. Results. The results confirmed that our software successfully segments the ball-bearing markers and determines the laser alignment assessment on the ACR CT phantom. Evaluation of the algorithm with images from the 16 CT scanners revealed that the difference between the results from automatic and manual methods were about 0.2 mm with a p value of around 0.7 (no statistical difference). Misalignment in the y-axis was larger than the misalignment in the x-axisfor the majority of the scanners tested. It was found that the phantom tended to be placed 2 mm higher than the iso-center. Conclusions. Software to automatically assess CT laser alignment with the ACR CT phantom was successfully developed and evaluated. The automatic assessment was comparable to manual assessment. In addition, the automatic method was user independent and fast.