Diagnostic accuracy in remote expert consultation using standard video-conference technology

Abstract

An international multicenter study (Germany, Sweden, and Switzerland) was performed to investigate the feasibility and diagnostic reliability of standard video-conferencing (VC) technology for remote expert consultation in radiology. Three high-spatial-resolution films (hand-bone, mammography, chest) and two low-spatial-resolution image sets (liver CT and MRI) were studied (total 446 images taken from different examinations). The images were recorded by a video camera, transmitted via public broadband networks, and displayed on a video monitor. The resolution of the recorded images varied from 2.4 lp/mm to 4.8 lp/mm at maximum zoom. After 3-4 months, the images were reexamined using conventional light-box reading. Diagnostic reliability was evaluated by receiver operating characteristics (ROC) analysis. With video conferencing, there was a noticeable loss of diagnostic accuracy for the high-spatial-resolution films, whereas for liver CT and liver MRI images VC reading seemed to be satisfactory (average area value difference < 0.02).

Radiology Education 2.0—On the Cusp of Change

“…It can also be used for remote expert consultation and e-learning programs (24)(25)(26). Although e-mail or conference calls are usually relied for much of the long-distance communication, sometimes a virtual meeting is more efficient with higher impact, especially for an important initiative.…”

Visual Object Detection with DETR to Support Video-Diagnosis Using Conference Tools

“…The settings used by the best models are as follows: During the second phase experiments, we obtained almost the same level of mAP but with much better loss values (see Table 2). Suppose we apply the clustering approach to our data classification, in other words: instead of four different classes (text, static, no translate, container), we focus just on textual objects-then our models reach the best performance on a moderate number of training data with validation and in this way we can adapt real-time object detection models to real-time consultations [5,8] and multilingual collaborations tools and solutions. From the investigated models, DETECTRON and DETR can serve best as these approaches because they will provide the same level of accuracy, but the best results will provide DETR.…”

“…Nowadays, the importance of multilingual translation has grown exponentially thanks to the impact of the COVID-19 pandemic [1,2]. Visual Object Detection [3,4] (moreover, Textual Object Detection), as a central topic of video conference-based remote diagnosis and consultancy [5] has intensified the digital transformation in the medical field. The issue generates a new 'realtime multilingual data processing' market to catalyze the elimination of existing communication barriers of multilingual video conferences.…”

“…Furthermore, using AI models already established in other sectors, we can provide outcomes with a high degree of precision to detect the textual bounding boxes for videoconference tools. These models might be part of a further cooperative communication layer between video diagnosis [5] attendees.…”

The RETAIN project: dicom teleradiology over an ATM-based network

“…We did not intend to set up the whole capacity of the network. From a functional point of view, high-quality visiophony is far from being a ªnew toyº [9,10,11,12,13,14]. The users have been convinced of the added value of this service:…”