P022] Assessment of the detectability of a moving guidewire in fluoroscopically-guided procedures using a mathematical model observer

Abstract

Purpose Fluoroscopically-guided procedures are among the most irradiating medical interventions. Thus, a fair amount of pressure is put on reducing patient exposure. This can be achieved by adapting the unit’s protocols, and by using technical upgrades. However, it is important to check if any image information content is lost. To our knowledge, there is no existing metric able to objectively assess the detectability of moving structures in fluoroscopic images. This work proposes a new method to assess image quality for a clinically relevant detection task in dynamic conditions using a mathematical observer. Methods A thin PMMA plate, supporting a 0.014” guidewire, was attached to a motion simulation engine via a cantilever, allowing motion with velocities and amplitudes representative of coronary arteries. The guidewire was placed at the unit’s isocenter, while 20 cm of PMMA around it simulated a patient. Images were acquired using pre-defined fluoroscopy quality levels (low, medium, and high) on two different units: One dedicated for interventional cardiology (IC), (Philips AlluraClarity FD10) the other for a broader purpose range (Philips Veradius Neo). Regions of interest were extracted from the images, either in the homogenous zones (signal-absent), or along the guidewire (signal-present). A mathematical model observer computed a decision variable for each signal-absent and signal-present image. The signal-to-noise ratio (SNR) between both decision variable distributions was taken as the figure of merit of the guidewire detectability. Dose rates were measured at the entrance of the PMMA using a pencil ionisation chamber. Results Preliminary results show that, for the IC unit, the SNR is equal to 3.5, 4.1 and 4.9 for respectively low (reference dose rate), medium (×1.6 rate) and high (×3.7 rate) quality fluoroscopy settings. In comparison, the same measurement on the non IC-dedicated unit yields a SNR of only 2.0 (high quality setting), at a dose rate comparable (11.5 vs. 12.2 mGy min−1) to medium quality fluoroscopy on the IC unit. Conclusions We have shown the feasibility of using a model observer to assess the detectability of a moving coronary guidewire. Further developments will include the fine tuning of the model, and matching between human observer results.