Accuracy of interpretation of full-length pediatric echocardiograms transmitted over an integrated services digital network telemedicine link.

ABSTRACT

Background. Transmission of echocardiograms via telemedicine links has allowed remote hospitals direct access to pediatric cardiology subspecialty care. This study assessed the accuracy of echocardiogram interpretation across an integrated services digital network (ISDN) telemedicine link.

Methods. Telemedicine systems were installed between Cape Fear Valley Medical Center neonatal intensive care unit and University of North Carolina Hospitals. One board-certified pediatric cardiologist interpreted 105 full-length echocardiograms that were videotaped and then transmitted over the system. Six months later, the same cardiologist reinterpreted the 105 original videotape studies from the off-site hospital and results were compared with the interpretations of the transmitted data.

Results. Interpretation of transmitted echocardiograms did not differ significantly from the original studies for diagnosis, evaluation of left ventricular function, valve function evaluation, and the presence of a ductus arteriosus. Minor differences in qualitative parameters were seen.

conclusion. Transmission of full-length echocardiograms over the ISDN telemedicine link is comparable to videotape review. There was no loss of significant clinical information, and the minor discrepancies noted did not impact management decisions.

**********

WITH THE ADVENT OF TELEMEDICINE, smaller hospitals have easier access to large academic centers where subspecialty expertise can assist in the care of patients with complex medical problems. This technology has been used to provide medical diagnosis, consultations, and case management. In pediatric cardiology, echocardiography has emerged as the standard for diagnosis and management of complex congenital heart disease in children. Since many infants are born at hospitals without the availability of on-site pediatric cardiology consultation, transmission of echocardiograms using telemedicine links is emerging as an important tool in the care of these infants. By transmitting an echocardiogram, evaluation of the infant can be made rapidly by a pediatric cardiologist, facilitating appropriate care.

The type of data connection between two telemedicine sites directly influences the amount of information that can be transmitted per second (bandwidth). Because digitalization of echocardiographic images requires a large amount of computer memory, a connection that can rapidly transfer large amounts of data can relay the best images without significant loss of information. Several previous studies have described transmission of echocardiograms either in cine-loop or full-length format over various types of telemedicine links. (1-5) In view of the complex anatomy, high heart rates, and amount of structural and functional information needed for diagnosis of congenital heart disease, we thought that review of full-length studies was essential to make an accurate interpretation. This study was designed to compare the accuracy of interpretations of full-length echocardiograms transmitted across a telemedicine link utilizing 3 bonded integrated services digital network (ISDN) lines at a rate of 384 kilobytes per se cond (kbps) with the original videotaped study.

METHODS

Two desktop telemedicine stations were linked between the University of North Carolina School of Medicine (UNC) in Chapel Hill, and Cape Fear Valley Medical Center (CFVMC) in Fayetteville, NC, approximately 90 miles away. Each station consisted of an IBM desktop computer, a videocamera, and a standard VHS videocassette recorder, allowing both transmission and recording of either live images or videotapes on either end. Each computer was equipped with a desktop videoconferencing system (Medpac, Bradford Group Ltd, Raleigh, NC), which digitized an analog videotape of an echocardiogram with a lossy video-compression algorithm (H.320). This digital data-stream was then sent to an internal videocard that could transmit at selected rates of n 64 kbps, where n is the number of B channels in an ISDN line (2 per line). The 2 computers were connected via 3 ISDN lines, with a total of 6 channels that were bonded using an internal BRI-6 multiplexing modem, resulting in a total transmission bandwidth of 384 kbps.

Pediatric echocardiograms were done with an HP 2000 echocardiography machine (Hewlett-Packard, Andover, Mass) and recorded on 1/2-inch VHS videotape. The technician then played the recorded study over the telemedicine system to a physician who received, recorded, and interpreted the study. A report was generated and faxed back to the referring physician. Additionally, live videoconferencing was used to speak to the referring physician as needed during the transmission.

A total of 114 consecutive pediatric echocardiograms were transmitted from CFVMC to UNC between January 1, 1997, and February 15, 1997. Of these, 105 transmitted studies were interpreted by a single, board-certified pediatric cardiologist (J.L.C.) at UNC. These studies ranged from 3 minutes to 15 minutes in duration and included 2-dimensional images, spectral and color Doppler images, and M-mode tracings, when appropriate. In order to eliminate intra-observer error, the same cardiologist reinterpreted the original videotape studies from CFVMC approximately 6 months later, and the results were compared with the 105 interpretations from the transmitted images read earlier. Approximately 450 additional studies had been transmitted during this 6-month period; it was thought that this interval, in addition to the number of echocardiograms interpreted in the interim, would minimize the cardiologist's memory of the initially transmitted study results.

In each study, the primary diagnosis, ventricular function, presence and amount of valvar regurgitation, presence or absence of a patent ductus arteriosus, and any other pertinent findings were recorded. The valvar regurgitation was graded as none, mild, moderate, or severe. These variables were compared utilizing Fisher's exact test or the chi-square test. Differences between the original and transmitted studies with a P value < .05 were determined to be significant.

RESULTS

Sixty patients underwent a total of 105 studies (mean, 1.75 studies per patient; range 1 to 9 studies per patient). The ages of the patients ranged from 1 day to 14 years (median 8 days, SD 2.8 years). There were 79 inpatient (75%) and 26 outpatient studies (25%). The indications for the study were to rule out congenital heart disease in 57 (54%), follow-up evaluation for a patent ductus arteriosus in 28 (27%), follow-up evaluation for left ventricular hypertrophy in 15 (14%), and other various indications in 5 studies (5%).

The original diagnosis from the recorded echocardiogram was normal in 73 studies (70%), left ventricular hypertrophy in 20 studies (19%), ventricular septal defect in 12 studies (11%), and 1 case of complex double-inlet left ventricle, ventricular inversion, L-transposition of the great arteries (Table 1). A patent ductus arteriosus and/or patent foramen ovale were considered normal in the newborn period. The transmitted studies were interpreted as normal in 71 (67%), left ventricular hypertrophy in 23 (22%), ventricular septal defect in 11 (11%), and double-inlet left ventricle, ventricular inversion, and L-transposition of the great arteries in 1.

A patent ductus arteriosus was found in 30 (29%) of the original studies. These were further graded by size as small in 7 cases (23%), moderate size in 18 cases (60%), and large size in 5 cases (17%). In the transmitted studies, 31 cases (30%) had a patent ductus arteriosus. These were classified as small in 10 cases (33%), moderate in 15 cases (50%), and large in 6 cases (17%). A patent foramen ovale was noted in 61 (59%) of the original studies and 58 (55%) of the transmitted studies.

Evaluation of ventricular function is displayed in Table 2. No significant differences were found between the transmitted and original study interpretations. Valve-function interpretation is seen in Table 3. Again, there were no significant differences noted between the 2 groups.

Overall, there was no significant statistical difference between any parameters reviewed in the interpretation of the transmitted studies and review of the original videotapes. Grading differences in function or regurgitation were always 1-category differences (eg, mild vs moderate).

DISCUSSION

Echocardiography has become an invaluable tool in the diagnosis and mangement of patients with congenital heart disease. Transmission of echocardiograms using various telemedicine system configurations has been previously reported. (1-5) These systems used differing imaging formats (cine-loop vs frill-length) and a variety of bandwidths, from a single telephone line (9.6 kbps) to T-1 communication lines (1.54 megabytes/second). This study evaluates the accuracy of transmission of large numbers of full-length echocardiograms using bonded ISDN lines. We show that transmission of echocardiograms from an outside hospital at 384 kbps resulted in reliable interpretation without loss of significant clinical information.

In our series of 105 echocardiograms, only 1 structural defect was misinterpreted; a small ventricular septal defect that was read as aortic insufficiency. Three studies interpreted as mild left ventricular hypertrophy when transmitted were found to be normal after review of the original tapes. These studies were done early in our experience when M-mode measurements of the left ventricular walls were not routinely performed. To address this error, we have since made M-mode measurements a standard part of any evaluation of ventricular hypertrophy. Ventricular function also tended to be graded as slightly worse on the transmitted studies than the original. This evaluation was qualitative, since ejection fractions or shortening fractions were not routinely obtained. Analysis of the discrepancies found only 1-category differences between interpretations (eg, moderate vs mild left ventricular depression). No significant differences in evaluation of valve function were found. Overall, the minor discrepancies found did not impact on any clinical management decisions.

The overall incidence of significant structural heart disease found in this series was low. This series was an unselected sample that represented a normal 6-week period of transmissions between the 2 institutions. The majority of findings involved patent ductus arteriosus, patent foramen ovale, and ventricular septal defects; this is consistent with the usual mix of pathology seen at a community-based hospital. The one case of cyanotic heart disease identified in this series was correctly diagnosed and appropriately triaged to the tertiary care center.

The quality of the recorded images was dependent upon the echocardiography technician who performed the scan. Initial and ongoing training of ultrasonographers in the proper techniques for scanning pediatric patients at the remote site is an integral reason for the high rate of accurate interpretation in this study. An example of the importance of training was seen with the initial discrepancies in the diagnosis of ventricular hypertrophy, which were eliminated when consistent M-mode scanning was performed. This training can be done with immediate interpreter feedback during a live, online transmission or, more commonly, by critiquing an examination while viewing the recorded images with the technician on the transmitting end. Both of these methods are useful in improving and maintaining a high standard of scanning, but we believe that the latter method is more efficient because of the large percentage of normal studies.

An ISDN-based system was chosen because of the widespread availability of the telephone equipment, the low cost of installation and maintenance, and the dialup capability that allows multiple sites to access our center. The bandwidth choice of 384 kbps was made because studies that were transmitted at slower speed tended to have significant loss of frames, which rendered the images choppy. Increasing the bandwidth above 384 kbps added only slight improvement to the transmitted images. Although we did not measure viewed-frame rates during the study, the H.320 decompression algorithm is 30 frames per second when transmitting at 384 kbps.

The field of telemedicine is rapidly changing. Changes in videocompression technologies, data-line bandwidth capabilities, and computer speed have resulted in a tremendous variation in the configuration of any given telemedicine link. Possibly the most image-intensive application to use this type of system is echocardiography, which generates images at up to 30 frames per second. Because each of these frames needs to be digitized, different methods have been used to minimize the necessary amount of data to be sent. The earliest studies involving transmission of echocardiograms used a microwave transmitter and receiver to send broadcast-quality images. (6) These were full-length studies that had excellent images, but the system was expensive and required access to the telephone company transmitters, which were hard-wired to each participating hospital. Advances in personal computers and modem technologies made it feasible to use a regular telephone line to transmit echocardiograms. Small, operatorselected segm ents of certain echocardiography views (cine-loops) were usually sent because of the large amount of data required to transmit full-length studies and the small bandwidth available. The pediatric studies were used to screen for congenital heart disease and facilitated transfer of appropriate patients for further treatment. (1) Cardiologists treating adults designed a network to transmit cine-loops of ventricular function after a myocardial infarction. (7) Trippi et al (3) transmitted cine-loop studies of dobutamine stress echocardiograms from an emergency department to help differentiate noncardiac chest pain from myocardial ischemia. This study showed good sensitivity and specificity of the transmitted loops versus clinical and catheterization findings. Good predictive accuracy was also shown in other indications for echocardiography in the emergency room, with good correlation with the original recorded images. Digital fetal echocardiograms have also been transmitted in the cine-loop format. (8) More recent studies have used a single ISDN line (2) or 3 ISDN lines (4) to transmit full-length pediatric echocardiograms on 3 patients and 22 patients, respectively. The largest study to date involved transmission of 200 fetal ultrasounds, (9) and revealed good correlation between the transmitted and recorded examinations. Interestingly, the authors noted that 2 fetal cardiac anomalies were not found on the transmitted images and believed this may have been due to either difference in the experience of the interpreting physician or reader bias. Finally, Murdison et al (5) reported their experience using a T-1 (1.54 megabytes per second) connection to remotely evaluate 13 pediatric patients using an electronic stethoscope, full-length echocardiography, and supplemental clinical information provided by the referring physician. Of these patients, 5 had some cardiac pathology, and 2 of those 5 patients required transportation to the tertiary care center for further assessment. They concluded that telemedicine has utility in evaluating new referrals and in postoperative follow-up.

CONCLUSION

From this study, we conclude that full-length pediatric echocardiograms can be transmitted over 3 bonded ISDN lines at 384 kbps with excellent sensitivity and specificity. Two-dimensional anatomy, spectral and color flow Doppler information, and M-mode measurements can be reliably interpreted without loss of significant clinical information. Sonographer training is an essential element in this accuracy. Use of this telemedicine system allows patients and physicians at a remote hospital access to tertiary care subspecialty expertise without need for physical transfer. Further research is needed to evaluate other aspects of this new technology, including cost-effectiveness, appropriate utilization, and effect on clinical outcomes.

TABLE 1.

Diagnosis Based Upon Original and Transmitted Echocardiograms

                                Original  Transmitted  P Value

Diagnosis *
  Normal                           73         71         NS
  Left ventricular hypertrophy     20         23
  Ventricular septal defect        12         11
  Other                             1          1
Patent ductus arteriosus
  None                             75         74         NS
  Small                             7         10
  Moderate                         18         15
  Large                             5          6
Patent foramen ovale
  Present                          61         58         NS
  Absent                           44         47

* Patients may have had more than one diagnosis.

NS = Not significant.
TABLE 2.

Ventricular Function Interpretation on Original and Transmitted
Echocardiograms

                     Original  Transmitted  P Value

Normal                 103         99         NS
Mild depression          2          5
Moderate depression      0          1

NS = Not significant.
TABLE 3.

Valve Function Interpretation on Original and Transmitted
Echocardiograms

                         Original  Transmitted  P Value

Mitral regurgitation
   None                     98         97         NS
   Mild                      7          8
Tricuspid regurgitation
   None                     63         61         NS
   Mild                     42         44
Aortic regurgitation
   None                    104        103         NS
   Mild                      1          2
Pulmonary regurgitation
   None                     99         97         NS
   Mild                      6          8

NS = Not significant.

References

(1.) Sobczyk WL, Solinger RE, Rees AH, et al: Transtelephonic echocardiography: successful use in a tertiary pediatric referral center. J Pediatr 1993; 122:S84-S88

(2.) Fisher JB, Alboliras ET, Berdusis K, et al: Rapid identification of congenital heart disease by transmission of echocardiograms. Am Heart J 1996; 131:1225-1227

(3.) Trippi JA, Lee KS, Kopp G, et al: Emergency echocardiography telemedicine: an efficient method to provide 24-hour consultative echocardiography. J Am Coil Cardiol 1996; 27:1748-1752

(4.) Alboliras E, Berdusis K, Fisher J, et al: Transmission of full-length echocardiographic images over ISDN for diagnosing congenital heart disease. Telemed J 1996; 2:251-258

(5.) Murdison KA: Telemedicine: a useful tool for the pediatric cardiologist. Telemed J 1997; 3:179-184

(6.) Finley JP, Human DG, Nanton MA, et al: Echocardiography by telephone-evaluation of pediatric heart disease at a distance. Am J Cardiol 1989; 63:1475-1477

(7.) Illiceto S, D'Ambrosio G, Scrutinio D, et al: A digital network for long-distance echocardiographic image and data transmission in clinical trials: the CEDIM study experience. J Am Soc Echocardiogr 1993; 6:583-592

(8.) Fyfe D, Doelling N, Graham D, et al: Digital fetal telemedicine: validation of echocardiography and image archiving (Abstract). Pediatrics 1997; 98:S533

(9.) Malone FD, Nores JA, Athanassiou A, et al: Validation of fetal telemedicine as a new obstetric imaging technique. Am J Obstet Gynecol 1997; 177:626-631

RELATED ARTICLE: KEY POINTS

* Transmission of full-length pediatric echocardiograms can be reliably performed over an ISDN-based telemedicine link.

* Interpretation of transmitted echocardiograms can result in reliable interpretations without significant loss of information.

* Minor discrepancies found co not impact on clinical management decisions.

From the Division of Pediatric Cardiology, Department of Pediatrics, University of North Carolina at Chapel Hill, and the Division of Neonatology, Cape Fear Valley Medical Center, Fayetteville, NC.

Reprint requests to John L. Cotton, MD, University of North Carolina at Chapel Hill, Division of Pediatric Cardiology, 311 Burnett-Womack, CB No. 7220, Chapel Hill, NC 27599-7220.