Task based optimization of system parameters in image guided interventions

Fluoroscopy imaging is an essential tool in the field of interventional radiology and cardiology, giving physicians the possibility to look at the internals of a patient in real-time. Having dynamic images available of the patient gives physicians a fast and easy guidance tool during image-guided interventions, like for example during stent placement in a stenotic artery. Fluoroscopy also gives physicians the possibility to look at the functionality of a certain organ, e.g. the pumping of the heart or the motion of swallowing. The downside of using fluoroscopy imaging however, are the x-rays needed to create the images leading to radiation dose absorbed by the patient and the medical personnel. A fundamental subunit of a fluoroscopy system that searches for the crucial balance between image quality and dose is called the Automatic Dose Rate Control (ADRC). The ADRC automates the selection of the x-ray factors to cope with changes in patient thickness during the exam. In so doing, the ideal ADRC provides consistent and adequate image quality for the medical tasks on a broad range of patients.

The goal of this thesis was to investigate a proper method of describing the efficiency of an ADRC and to investigate new methods to optimize the acquisition parameters during fluoroscopy guided interventions using a dedicated image quality metric: the spatial frequency dependent signal difference to noise ratio, SDNR(u). This image quality metric adjusts the standard SDNR by adding geometric and motion blurring in its calculation. By using this dedicated image quality metric in combination with a radiation dose metrics of interest, the work in this study compares the efficiency, defined as the image quality squared divided by the dose, of several ADRC operation methods. In chapter 2, measurements of stents and guidewires showed that SDNR(u) is much more closely linked to the visual perception of image quality than the standard image quality metric SDNR, providing good evidence that using this new method is useful and relevant in the field of interventional radiology and cardiology.

Chapter 1 shows that the efficiency of a 5-parameter ADRC is superior to that of a 3-parameter system, indicating the benefit of a more flexible ADRC. Similarly, chapter 3 shows that improvements in the x-ray generator have a beneficial effect on imaging performance. The study compared the newer Siemens Artis Q with the older Siemens Artis Zee, concentrating on the ADRC parameter selection, and the results show a clear gain in efficiency for the newer system.

Chapter 4 is the main chapter of the thesis and introduces a new method for the ADRC to optimize the exposure parameter selection. Currently implemented ADRC systems measure the image brightness in specified regions of the image and keep the pixel value in this area constant. The new approach tries to reach a target SDNR(u) value instead of a target pixel value and it does so while simultaneously selecting the exposure parameters with the lowest dose to the patient, therefore continuously reaching optimal efficiency. The measurements, performed over a large range of patient thicknesses and using many different materials, show a significant increase in the imaging efficiency for the new method compared to the conventional ADRC approach. Expectations are that this new ADRC method can lead to important dose reductions. 

Finally, chapter 5 implements the techniques learned in the previous chapters to determine the exposure parameters of a concept hybrid Angio-MRI system proposed by Siemens Healthineers. The x-ray part of this concept system has a very different geometry and spectral filtration than standard angio systems and is operating near a magnetic field. This study therefore tried to investigate whether a conventional x-ray tube had enough power output to select the exposure parameters necessary to deliver the same SDNR(u) as a conventional system. To answer this question, a simulation platform to calculate image quality and dose was created. Results showed that under certain conditions, the hybrid system should be able to match the image quality of a conventional system.