This study evaluated whether a dedicated image processing workstation installed at a remote location could be operated without performance degradation. The same examiner processed the same data in both the OP and RW environments, and the processing times for each task were measured. The overall impact on workflow was assessed over time. A comprehensive evaluation by all examiners showed that the RW environment reduced processing time by about 1 min. Initially, we anticipated that performance drawbacks would arise in tasks, such as re-slicing and rendering, which involve complex procedures, in the RW environment. However, despite data transfers and operations over a wide-area network (WAN), rendering performance was found to be equivalent to that of the OP environment. Furthermore, during the re-slicing of large image data sets, an average time reduction of 70 s was completed with the dedicated image processing device in the RW environment. This is likely because re-slicing requires processing the entire data set, which is a heavy task, demonstrating the advantage of the dedicated device. In actual workflow, the benefits are not limited to simple reductions in waiting time but also include a significant decrease in the time the device is occupied for processing. While local image processing devices also manage image transfer, image transfers may stop during intensive image processing, causing workflow interruptions. Using RW to distribute workloads could bring about additional workflow improvements that this study could not measure. In addition, while a dedicated image processing device is expensive and challenging to prepare specifically for the reconstruction of dental CBCT images, cost-effectiveness could be achieved using medical imaging devices during available time slots.

As a medical service utilizing networks, there are various applications such as inter-facility information sharing, cloud services for personal health records, or web services to assist in diagnosing specific diseases [4,5,6,7]. Systems for remote monitoring or image recognition using artificial intelligence have also been developed [8]. Cloud computing has become common for sharing devices with high processing power, such as computer assisted design or large-scale language models requiring complex data processing [9,10,11]. In medical imaging, high-performance processing devices are needed to handle large volumes of images with various modifications or analyses [12]; however, cloud-based services are still limited in the medical field [13], and while there is research on web services for 3D image processing in medicine [14, 15], they are not yet widely adopted. Services using the Remote Desktop Protocol (RDP) provide a solution to reduce network load when using devices installed remotely [16, 17], and it has been reported that performance can be adequately maintained in a local area network environment. In this study, we used a commercial network with a 2Gbps bandwidth, sharing it with all traffic in the medical information systems of two hospital facilities. This traffic includes electronic medical records and image viewing in diagnostics and the registration of image information from CT, MRI, endoscopy, and ultrasound examinations. Since remote operation uses a remote display protocol, the proportion of network bandwidth used is kept low. It could be used without delay, even in a shared environment with other systems, which is essential. Moreover, from the viewpoint of image processing devices, the local terminal is a thin client system. As long as it can run a remote display protocol, it can be used without being restricted by installation location or model, allowing work locations to be changed without physical or spatial constraints.

Within the scope of our search, we were unable to find any papers discussing the impact of environmental factors, such as internet speed and PC performance on medical imaging, comparisons across different applications, or the effects of physical distance on communication quality.

Theoretically, differences in PC performance among workstations affect processing time. In this study, the image processing environment used in RW typically exhibits higher performance than OP setups, with applications specifically designed for image processing, leading to an expected reduction in processing time. In addition, while the image sizes include data with varying FOV, the same data are processed and compared in both groups, effectively eliminating the influence of differences in data size.

Although there are studies on web-based architectures, we could not find any research evaluating the impact of inter-site distance in a configuration that directly utilizes a WAN as in this study. However, using RDP and a bandwidth-guaranteed line, we believe that sufficient communication quality was ensured, enabling real-time processing.

In summary, even when the image processing equipment is located at a remote site, it can be operated without compromising usability. The fact that workflow improvements are observed depending on the performance of the image processing equipment demonstrates that high-performance resources can be utilized via a WAN environment with performance equivalent to that of a LAN connection.

Limitations include the variation in processing times among examiners, the inability to fully evaluate the impact of other systems on network bandwidth usage, the use of a dedicated remote display protocol, and the fact that the image processing device used for remote access does not support multi-client simultaneous access. Regarding the variation in examiners' processing times, one specific examiner showed no significant difference in processing times between the OP and RW environments. This examiner generally had longer operation times than others, and the variation in processing times was considerable in the RW environment, suggesting a possible lack of proficiency in image processing operations.

While it is difficult to eliminate the impact of other systems' network traffic, the network utilization bandwidth is constantly monitored centrally, and based on the average values of network traffic every 2 h, the peak utilization during the day is about 20%, with an average of around 6%, suggesting that the impact of network bandwidth limitations is limited. In addition, to address network failures, we use a dualized 1Gbps line, which is expected to provide an environment with no logical restrictions due to bandwidth, even if one of the lines is completely down.

The dedicated remote display protocol could be replaced by the Windows OS standard RDP (winRDP), as suggested [16]. This would enable operation without restrictions on location or terminal type, although verification of winRDP-based performance in practical settings has not yet been conducted. The image processing device used for remote access does not support multi-client simultaneous access, which needs to be addressed in the future. Still, this is managed by checking machine availability through operational management.

In conclusion, dental CBCT image processing with a remote display protocol can be performed at a remote location with processing times equal to or less than those using on-premises devices, suggesting potential benefits for workflow improvement.