Development of methods and software for rapid quality control in fluoroscopy

Sammanfattning: Background: Fluoroscopy is a common imaging technique which uses X-ray to derive a real time imaging of patient anatomy to determine diagnosis and positioning of patients for interventional procedures. It is therefore important that the fluoroscopy systems maintain their performance. Assessment of image quality parameters (such as: low contrast resolution, uniformity, homogeneity and detection of defective pixels and artifacts) is one way to assess if they maintain their performance. This study aims to develop and implement a Matlab script to do a remote Quality Control (QC) and set up tolerance levels on different types of fluoroscopy systems.                                        Method: Three fluoroscopy systems were used in this project, Siemens Axiom Artis Zee MP, Siemens Cios Alpha and Ziehm Vision RFD. There were two setups used in the study for image acquisition by adding a 2 mm Cu filter as the attenuating material placed directly on the X-ray tube. A Cylindrical aluminum contrast detail of dimension 4 mm thick and 6 mm diameter was placed in the middle of X-ray field (Setup 1 on patient couch and setup 2 directly on the flat panel detector). The fluoroscopic images were acquired with and without contrast detail over a period of six month. The image quality parameter SNR2rate was determined from signal and background images while other quality parameters such as kerma-area product rate PKA, rate, uniformity, homogeneity, low contrast resolution, SNR, defective pixels and artefact detection were examined and determined from the background images. Two additional supporting experiments were performed, one with a chest phantom and 13 cm PMMA and the other one a human visual detection 4-AFC experiment.                 Result: The image quality index SNR2rate and the dose rate index PKA, rate, the low contrast resolution parameter (LCRP), uniformity, homogeneity and SNR values were within ±2 standard deviation for repeated measurements in each system. Nevertheless, the result indicates that Siemens Axiom Artis Zee MP has the best performance while Ziehm Vision RFD has the worst performance between these three systems. The result from the defective pixel method indicate that for 20 % tolerance there were no defective pixels for Siemens Axiom and Cios Alpha. Ziehm Vision had also no defective pixels for 30 % tolerance. The artefact detection shows that artefact level is high for fluoroscopy systems and Ziehm Vision RFD has artefact level more than 50 % tolerance.  The chest phantom experiment indicate that SNR2rate varies considerably over the lung anatomy as expected.           The 4-AFC experiment indicates that the effective time was 0.14 s for human observers to integrate the information in the live image.           Conclusion: The methods developed and implemented in this project were successfully able to determine and assess the image quality parameters such as SNR2rate,PKA, rate, low contrast resolution, uniformity, homogeneity, SNR and detection of defective pixels. Further effort is needed for installation of Matlab script on our local server, connection with Excel program and internal website (SharePoint) and adding more clinical fluoroscopy systems to do remote QC in Region Östergötland.