Siemens Healthineers AG, Siemensstrasse 3, 91301 Forchheim, Germany.
Diagnostikum Linz GmbH, Saporoshjestrasse 3, 4030 Linz, Austria.
Radiography (Lond). 2024 Jul;30(4):1073-1079. doi: 10.1016/j.radi.2024.05.002. Epub 2024 May 18.
Intelligent virtual and AI-based collimation functionalities have the potential to enable an efficient workflow for radiographers, but the specific impact on clinical routines is still unknown. This study analyzes primarily the influence of intelligent collimation functionalities on the examination time and the number of needed interactions with the radiography system.
An observational study was conducted on the use of three camera-based intelligent features at five clinical sites in Europe and the USA: AI-based auto thorax collimation (ATC), smart virtual ortho (SVO) collimation for stitched long-leg and full-spine examinations, and virtual collimation (VC) at the radiography system workstation. Two people conducted semi-structured observations during routine examinations to collect data with the functionalities either activated or deactivated.
Median exam duration was 31 vs. 45 s (p < 0.0001) for 95 thorax examinations with ATC and 94 without ATC. For stitched orthopedic examinations, 34 were performed with SVO and 40 without SVO, and the median exam duration was 62 vs. 82 s (p < 0.0001). The median time for setting the ortho range - i.e., the time between setting the upper and the lower limits of the collimation field - was 7 vs. 16 s for 39 examinations with SVO and 43 without SVO (p < 0.0001). In 109 thorax examinations with ATC and 112 without ATC, the median number of system interactions was 1 vs. 2 (p < 0.0001). VC was used to collimate in 2.4% and to check the collimation field in 68.5% of 292 observed chest and other examinations.
ATC and SVO enable the radiographer to save time during chest or stitched examinations. Additionally, ATC reduces machine interactions during chest examinations.
System and artificial intelligence can support the radiographer during the image acquisition by providing a more efficient workflow.
智能虚拟和基于人工智能的准直功能有可能为放射技师实现高效的工作流程，但具体对临床常规的影响仍不清楚。本研究主要分析智能准直功能对检查时间和与放射系统交互次数的影响。
在欧洲和美国的五个临床地点，对三种基于相机的智能功能（基于人工智能的自动胸部准直（ATC）、用于拼接长肢和全脊柱检查的智能虚拟正交（SVO）准直以及在放射系统工作站上的虚拟准直（VC））的使用情况进行了观察性研究。在常规检查期间，有两个人进行半结构化观察，以收集功能激活或未激活时的数据。
95 例使用 ATC 的胸部检查的平均检查时间为 31 秒，而 94 例未使用 ATC 的检查时间为 45 秒（p<0.0001）。对于拼接骨科检查，34 例使用 SVO，40 例未使用 SVO，平均检查时间为 62 秒和 82 秒（p<0.0001）。设置正交范围的中位时间，即设置准直区域上下限之间的时间，对于 39 例使用 SVO 和 43 例未使用 SVO 的检查，分别为 7 秒和 16 秒（p<0.0001）。在 109 例使用 ATC 的胸部检查和 112 例未使用 ATC 的检查中，中位系统交互次数分别为 1 次和 2 次（p<0.0001）。在 292 例观察到的胸部和其他检查中，2.4%使用 VC 进行准直，68.5%使用 VC 检查准直区域。
ATC 和 SVO 使放射技师能够在胸部或拼接检查期间节省时间。此外，ATC 减少了胸部检查期间的机器交互次数。
系统和人工智能可以通过提供更高效的工作流程来支持放射技师在图像采集过程中的工作。
