Author: James R Duncan, Nikhat Fatima Mansoori, Allan Thomas 👨🔬
Affiliation: Mallinckrodt Institute of Radiology, Washington University School of Medicine 🌍
Purpose: Collimation during fluoroscopically-guided interventions (FGIs) can not only reduce radiation exposure to patients and personnel but also improve image quality. Nonetheless, there are concerns that collimation is neither routinely employed nor optimized. Improvement strategies, including using artificial intelligence, have been proposed. To assess the potential impact of improved collimation, we objectively measured the use of collimation across a wide range of different FGIs. In addition, collimation during uterine artery embolization (UAE) procedures was studied in detail given efforts to optimize radiation use for these patients.
Methods: Video was routinely captured and archived from the live fluoroscopy monitors of 7 different interventional suites. Videos from 84 randomly selected FGIs performed over a three-month period were processed to create a series of > 700,000 still images. Images corresponding to the start and end of every irradiation event were identified and further processed to determine the extent of collimation during each irradiation event. In addition, videos from 15 uterine artery embolization (UAE) procedures were studied to assess collimation including use during different segments of these procedures.
Results: For the randomly selected procedures, >10% collimation was seen in 51% of all irradiation events and average degree of collimation was 20+/-17%. The frequency and extent of collimation varied according to the event type and procedure category.
For the directed sample of UAE procedures, the frequency of >10% collimation rose to 70% and the average degree of collimation was 38+/-15%. During the embolization segments of the UAE procedures, frequency of collimation increased further to 92% and the average collimation index rose to 62%.
Conclusion: The combination of a large video archive of FGIs and computer vision techniques allowed a detailed and objective analysis of collimation across a wide range of procedures. The results demonstrate substantial differences in routine use of collimation and opportunities for improvement.