| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Computers |
1 Dartmouth Medical School, Hanover, NH.
2 Department of Radiology, University of Belgrade Medical School, Belgrade,
Serbia.
3 Department of Radiology, St. John Surgical Hospital, Zrenjanin, Serbia and
Montenegro.
4 Department of Radiology, Dartmouth-Hitchcock Medical Center, 1 Medical Center
Dr., Lebanon, NH.
Received December 15, 2005; accepted after revision August 1, 2006.
Address correspondence to R. D. Harris
(robert.harris{at}hitchcock.org).
OBJECTIVE. We sought to show that sonography can be performed in teleconference settings, "telesonography," in which a remotely located sonography interpreter can monitor the examination in real-time and guide the examiner with voice commands while the patient simultaneously undergoes imaging, albeit at low resolution, thus helping to overcome the lack of trained operators in certain areas.
CONCLUSION. This system of image transfer offers the potential for sonography to be performed at a remote underdeveloped region and interpreted in real-time at a distant site by trained radiologists, thereby extending the presence of physicians in virtual space.
Keywords: computers in radiology • digital images • Internet • radiology practice • sonography • teleradiology
The advent of inexpensive portable sonographic instruments not only has increased the use of sonography in primary care practices and in nonhospital settings, but also has promoted the use of sonography as the primary imaging tool in many civilian, humanitarian, and military missions. Largely because of its low cost, portability, and benign safety profile (i.e., no ionizing radiation exposure, no contrast agents necessary), diagnostic sonography has experienced widespread use in clinical settings; sonography is one of the most widely used diagnostic imaging techniques worldwide. However, its utility in remote locations has been hampered by the need for extensive operator training or experienced sonologists to perform the examination and interpret the scans. Although interpretation of radiologic data from a distance is readily available through PACS and standard teleradiology systems in developed countries, such as the United States and European countries, the currently used systems for image transfer were developed for high bandwidth and cannot support the transmission of sonograms or radiographs from remote or underdeveloped countries because of the limited bandwidth available from those locations.
We sought to show that sonography can be performed in teleconference settings, "telesonography," in which a remotely located sonography interpreter can monitor an examination in real-time and guide the examiner with voice commands while simultaneously imaging the patient, albeit at low resolution, thus helping to overcome the lack of trained operators.
In this study, we hoped to establish a model for low-cost telesonography with real-time transmission between a remote clinic in an underdeveloped area and a large U.S. medical center to assess the frame rate of such a system and to compare the quality of both the original (real-time) and transmitted (store-forward) sonographic images. The ultimate goal was for trained paramedical personnel or nurses at remote (Third World) locations to perform diagnostic-quality sonography and for an expert sonologist located anywhere in the world to interpret the images essentially simultaneously for diagnostic or screening sonography.
Materials and Methods
All the sonography studies were performed on a portable sonography machine (Sonosite 180, Sonosite Inc.) donated by the department of radiology at the Dartmouth-Hitchcock Medical Center to a surgical clinic in Zrenjanin, Serbia. During the method development period of 18 months, 50 patients were examined and the images from those examinations were transmitted to the U.S. medical center. Then 50 images of eight subjects, obtained by a single examiner, were used for the evaluation study. Standard Sonosite curvilinear abdominal (3 MHz) and linear (7.5 MHz) high-resolution transducers were used to image the abdomen, aorta, and thyroid gland of the subjects.
In the examination room, the sonography unit was connected to a television tuner video-capture card installed on a PC (Compaq Presario 2100, Hewlett-Packard). The PC at the surgical clinic was on the 100BaseT local area network (LAN) and connected to an Internet service provider (ISP) via a 68 kB/s ISDN (integrated services digital network) modem (Fig. 1A, 1B). This connection enabled real-time transfer and analog-to-digital conversion of both video stream and still images from the portable sonography machine to the PC. We used Microsoft AmCap and VidCap software packages, both available free of charge, to view these images on the PC [1]. With both applications, it was possible to record video clips from the examination in an audio video interleave (AVI) file format at approximately 1 MB/s. To maximize the efficiency of image transfer, we matched the video card's resolution to the inherent resolution of the compact sonography machine (i.e., 360 x 240 pixels).
|
|
Capitalizing on the ability of the H.323 protocol, an International Telecommunications Union (ITU) standard, to constantly monitor bandwidth and adjust the compression ratio, we organized the telesonography examination in two steps. During the first step, a sonologist could, via a low-resolution real-time video at a low frame rate (2-4 frames per second [fps]), follow the sonography examination and guide the sonographer to obtain a clear image of the organ or organs of interest. As the probe was slowed to optimize the image, the resolution of the transmitted image improved, reaching its maximum when the image on the sonography unit was frozen for recording. During this time, marked as the second step of the examination, one or more still images could be recorded by both the sonologist and sonographer nearly simultaneously (usually a lag of between 5 and 10 seconds in our experience). As soon as an adequate image was obtained, the sonographer could be instructed to continue scanning.
After exploring several H.323 protocol-based teleconferencing applications for real-time transmission, we used NetMeeting (Microsoft) because of its low cost, superior performance (minimal technical difficulties, good performance on computers with low CPU speed and limited memory), and global availability. NetMeeting is sold in a package with Microsoft Windows 98, 2000, Me, and XP operating systems. The videoconferencing session was initiated from a local PC in Zrenjanin by calling the remote computer's Internet protocol (IP) number. Transmission of sounds and video signals from the remote computer occurred immediately on acceptance of the standard telephone call. During an examination, video stream from the sonography unit in Zrenjanin was teleconferenced in real-time at estimated frame rates of 3-6 fps, which enabled a sonologist in the United States to follow and guide the examiner in Zrenjanin. On the receiving end of the transmission (Dartmouth Medical School, Hanover, NH), images were viewed on a commercially available PC (Compaq Presario 2100 Notebook, Hewlett-Packard) linked through the standard 802.11b wireless network to a 128 or 256 kB/s DSL (digital subscriber line) router (MR814, Netgear).
When the examiner in Zrenjanin recorded a still image during the examination (e.g., the left kidney) the physician in the United States captured the same image on the computer screen virtually simultaneously. Transmitted images underwent software-based compression and decompression in which the compression rate depended on the bandwidth and varied between 1:280 (for CIF [common interchange/intermediate format]) and 1:75 (for QCIF [quarter common intermediate format]).
The same 10-15 images (recorded by the sonologist) were recorded on the portable sonography machine during each examination and were later transferred to the PC via AmCap software. Still images (in bitmap format) were incorporated in Microsoft Word documents that were annotated and, in addition, labeled and sent to the reviewer's PC via e-mail. We sent images embedded in Word files, rather than as attachments to an e-mail, mainly for convenience; we stored the series of images in one large Word file, which made it easier to annotate the images with comments. Thus, at the end of the examination, a data set of 10-15 corresponding images was created on each side.
The original noncompressed images were later compared with the transmitted compressed and decompressed images as obtained by the sonologist in Hanover. Three radiologists (two fellowship-trained body imagers and one senior radiology resident) and a medical student, all blinded to the origin of the images, interpreted both the real-time compressed images and their noncompressed store-forward partner images to assess the loss of image contrast and resolution. Reviewers were asked four questions about the image pairs. For the first question, they were asked to grade the quality of the images from 1 to 5, with 1 being the lowest and 5 being the highest and 5 indicating the image was equal to an image obtained on a cart-based high-resolution sonography machine used in most radiology departments in the United States. For the second question, they were asked to state their preference for one image over the other. For the third question, they were asked to identify an organ, and for the fourth question, they were asked to attempt a diagnosis. The Student's paired t test was used to analyze for statistical significance (p < 0.05).
To quantitatively determine the difference between the original and transmitted images, we used digital subtraction and "Mean Gray Value" (MGV) available in ImageJ software (version 1.32j, National Institutes of Health).
|
|
MGV =0.299red + 0.587green + 0.114blue.
Used in this setting, MGVs provide a semiquantitative method by which to compare two images and estimate the amount of lost information.
Results
The existing setup was sufficient to transmit sonographic examination data in realtime from the remote site to the academic center, albeit at a slower frame rate (estimated to be 3-6 fps) than is customary (20-30 fps). The amount of data transmitted was directly related to the speed of transducer movement, with a higher speed of movement requiring higher transmission speeds. Nonetheless, the image resolution improved as the probe was slowed and reached maximum levels when a static image was transferred. At that point, the loss of image quality was negligible.
The panel of three radiologists and a medical student analyzed the transmitted images with respect to image quality and overall accuracy on a scale of 1-5 (1, acceptable; 2, marginal; 3, adequate; 4, good quality; and 5, excellent quality). The average scores (± SD) for compressed real-time images and noncompressed store-forward images were, respectively, 2.72 ± 0.82 and 2.75 ± 0.84 (p = 0.30), neither a statistically nor clinically significant difference. When asked which image they thought was superior, our panelists chose the noncompressed image in 37.5% of the cases and the compressed image in 21.5% of the cases, and they had no preference in 40% of the cases.
The MGVs for the original and transmitted images remained comparable (MGV original = 48.88, MGV transmitted = 47.92). The MGV of the image obtained by digital subtraction (MGV = 1.89) of the pair indicates that only about 4% of data are lost during the compression and decompression step. The percentage of data lost for all analyzed images was between 2% and 5%.
The second part of our study suggested that transmitted images were as good as their original counterparts and that there was no degradation in perceived image quality during the transmission.
Discussion
Several reports of guided sonography examinations from remote locations [1, 2] describe the use of dedicated high-bandwidth Internet links and the simultaneous use of proprietary software packages to transmit video and audio data. Recent reports of sonography transmissions from missions in outer space [3] and from other remote locations indicate the potential that exists for using diagnostic sonography in remote settings. Despite these numerous reports, there is still no simple, comprehensive software package designed for telesonography applications. Most telemedicine systems rely on proprietary hardware and software packages that require costly high-bandwidth connections between centers (e.g., T1-T10 lines, intranet, LANs). Both the proprietary hardware and software and the costly underlying networks add to the system's expense and, in many cases, make telemedicine practical for only large medical centers in developed locations whose geographic locations permit highspeed Internet access. Some studies sponsored by the National Aeronautics and Space Administration or the U.S. Department of Defense have used high-bandwidth systems with expensive hardware and software at dedicated research facilities [4, 5], but these systems are not readily available for humanitarian missions in Third World countries or tertiary medical centers in the United States.
Our research study in telemedicine was made possible by several advancements: the availability of inexpensive compact sonography; the increasing accessibility of low- to medium-bandwidth Internet connections, via terrestrial- and satellite-based links; and the availability of inexpensive hardware and software solutions for transmission of sonography data in teleconferencing settings. The conversion of any analog video signal, including sonography (National Television System(s) Committee standard), into a digital signal viewable on a PC can be achieved with an analog-to-digital conversion card, which is commonly used in households to transfer videos from videotapes to CDs or DVDs. Similarly, any still or moving video image can be readily transmitted between two computers by either streaming or teleconferencing.
H.323 protocol-based teleconferencing software applications, widely used in private and commercial teleconferencing, are capable of transmitting video and audio streams in realtime, transferring files, and allowing remote access and control of the computers. The H.323 protocol, an ITU standard, includes protocols that compress video (H.261, H.263) and audio (G.711, G.723) streams for transmission over low-bandwidth Internet lines.
By combining real-time video streams and still-image capture, the bandwidth requirements decreased substantially and image transmission over telephone lines or possibly by satellite became possible. Images can be viewed by the sonologist and sonographer virtually simultaneously, and the novice examiner in the remote area can be directed to the appropriate anatomic areas by the trained sonologist.
We anticipate that the emergence of new telecommunication technologies will make this approach even more feasible. For example, new video transfer protocols, such as H.264, with higher compression ratios will enable better quality video to be transmitted over lower bandwidths. Alternatively, some other media might be used for the remotely guided telesonography. In a recent report, Blaivas et al. [6] used camera-equipped cell phones to transmit sonography images. This approach seems promising for the transmission of static sonography images in situations in which the examiners have some experience performing sonography.
However, disaster areas and Third World locations may lack cell telephone coverage because cell phones rely on terrestrial antennas, which may be nonexistent or damaged in mass disaster areas, and they also may lack the trained personnel to perform sonography examinations. Such situations underscore the importance of remote guidance in which a more experienced sonologist has access to real-time video and simultaneous audio, currently available only via broadband Internet links (cell telephone bandwidth is 14 kilobits per second [kbs] vs 68, 128, or 256 kbs for ISDN vs DSL lines, respectively). In disaster areas or the Third World, satellite-based Internet links would ensure global applicability of telesonography.
Limitations of the Study
Lower-than-expected grades for both original (2.72 of 5) and transmitted (2.75 of 5) sonography images were, in our opinion, partially due to the fact that all four panelists are accustomed to viewing sonograms on high-resolution machines and were largely unfamiliar with the image quality attained with a portable sonography machine. Second, to ease the grading in our study, all images were printed on analog hard-copy black-and-white printers, which resulted in further image resolution loss. When using digital subtraction of the original and transmitted images, a more quantitative method, the results suggested a minimal loss of data during telephone-line transmission, and although measurable, this loss of data was thought to be clinically insignificant.
Because sonography examinations are operator-dependent and often require realtime input from the interpreter (sonologist), the time delay between the examination and interpretation needs to be minimal in order not to compromise sonography's ability to display "real-time" (25-30 fps) images of moving structures, such as a fetus or blood flow. Remote imaging interpretation in most medical centers is done in a stop-forward fashion, in which data are first stored as a series of still images in a database, often through a PACS system, and then are reviewed and interpreted by a radiologist at a later time or at a distance.
To streamline image collection and exchange and reduce sonographers' time and engagement in image manipulation, our research team is currently working on a new application containing interfaces for teleconferencing and sonography, the latter of which will contain patient header data and record options and Image Forward and Back buttons, to further facilitate the process of telesonography.
Conclusion
Our pilot study revealed that transmitting sonography images using real-time and store-forward methods via a low-bandwidth Internet network is possible with little degradation in quality from a compact sonography machine placed abroad. We believe the transmission of sonography data from remote locations, which often lack optimally trained sonography operators, is desirable and probably feasible. We think that this system can be further refined to allow trained paramedical personnel to image patients for both diagnostic and screening sonographic examinations at a remote underdeveloped region with real-time interpretation at a distant site by trained radiologists, thereby extending the presence of physicians in virtual space.
Acknowledgments
We thank Steven Sargent, Kevin Buffington, and Sabrina Selim for evaluating and ranking the images. We would also like to acknowledge Jared Adams for his help in conducting statistical analysis and Robyn Mosher for editorial assistance.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
