Background
Pocket-size ultrasound has increased echocardiographic portability, but expert point-of-care interpretation may not be readily available. The aim of this study was to test the hypothesis that remote interpretation on a smartphone with dedicated medical imaging software can be as accurate as on a workstation.
Methods
Eighty-nine patients in a remote Honduran village underwent echocardiography by a nonexpert using a pocket-size ultrasound device. Images were sent for verification of point-of-care diagnosis to two expert echocardiographers in the United States reading on a workstation. Studies were then anonymized, randomly ordered, and reinterpreted on a smartphone with a dedicated, Health Insurance Portability and Accountability Act–compliant application. Point-of-care diagnosis was considered accurate if any abnormal finding was matched and categorized at the same level of severity (mild, moderate, or severe) by either expert interpretation.
Results
The mean age was 54 ± 23 years, and 57% of patients were women. The most common indications for echocardiography were arrhythmia (33%), cardiomyopathy (28%), and syncope (15%). Using the workstation, point-of-care diagnoses were changed in 38% of cases by expert overread (41% left ventricular function correction, 38% valvulopathy correction, 18% poor image quality). Expert interobserver agreement was excellent at 82%, with a Cohen’s κ value of 0.82 (95% confidence interval, 0.70–0.94). Intraobserver agreement comparing interpretations on workstations and smartphones was 90%, with a Cohen’s κ value of 0.86 (95% confidence interval, 0.76–0.97), signifying excellent intertechnology agreement.
Conclusions
Remote expert echocardiographic interpretation can provide backup support to point-of-care diagnosis by nonexperts when read on a dedicated smartphone-based application. Mobile-to-mobile consultation may improve access in previously inaccessible locations to accurate echocardiographic interpretation by experienced cardiologists.
Pocket-size cardiac ultrasound (PCU) devices improve the portability of echocardiographic image acquisition. In the hands of experienced operators, PCU has shown promising accuracy compared with traditional echocardiography. However, without trained personnel capable of accurately interpreting the acquired information, the potential for misapplication or incorrect interpretation may hinder its adoption by clinicians. Remote real-time expert echocardiographic interpretation has been demonstrated to be feasible, but such validation has required dedicated satellite connections and custom wireless transmitters, which may limit its applicability outside the military setting. Remote interpretation with electronic transfer of data also brings concerns about breaches of the security of confidential patient information, of particular concern in the United States as the rules enforcing the Health Insurance Portability and Accountability Act (HIPAA) include civil penalties against physicians who transmit such data without reasonable safeguards against impermissible disclosure. We sought to develop a HIPAA-compliant, low-cost, mobile-to-mobile echocardiographic transmission system that would enable images to be acquired at remote locations previously inaccessible to ultrasound technology and allow expert echocardiographers to interpret studies remotely on smartphones. Such a system could provide expert interpretation to patients in isolated locations where traditional echocardiography would otherwise be unavailable. We hypothesized that interpretation on a smartphone with dedicated medical imaging software would be as accurate as conventional workstation-based interpretation.
Methods
During a humanitarian mission in a remote, mountainous region of Honduras where ultrasound was not previously available, relief workers, including a cardiology fellow (who by the time of the mission had rotated through 3 dedicated months in an academic echocardiography lab under the supervision of a level 3 full-time director, personally performed 50 transthoracic echocardiographic studies, and participated in the interpretation of >750 studies), brought a PCU device to augment physical examinations. Patients were referred for cardiovascular evaluation by local primary care providers (i.e., noncardiologists). Relief workers alternatively transmitted the images directly from the field via dialup modem or during periods of limited connectivity couriered the electronic files to an urban center for upload via a broadband connection to expert echocardiographers for remote interpretation on an established desktop-based system. Analysis of acquired images was approved by the George Washington University institutional review board.
Echocardiographic Image Acquisition
The cardiology fellow performed limited, indication-focused echocardiographic studies on 89 patients with standard parasternal and apical views. In each patient, a recording of single parasternal long-axis, short-axis, and apical four-chamber views was obtained. Additional loops were recorded if deemed diagnostic of an abnormality at the point of care or if the fellow was unsure of a finding, but if a view revealed clearly normal findings, a loop was not recorded, to minimize data size. PCU images were acquired using Vscan (GE Healthcare, Wauwatosa, WI), a 1.4 × 7.3 × 2.8 cm device weighing 390 g with an 8.9-cm diagonal, 240 × 320 pixel display. Vscan uses a phased-array transducer (1.7–3.8 MHz) with a fixed sector angle of 75° and maximum depth of 25 cm. Color flow Doppler is available within a 30° maneuverable sector. Spectral Doppler or M-mode imaging is not available with the device. Images are stored on the device as JPEG still captures or MPEG-4 cine loops.
Workstation-Based Analysis
A preliminary point-of-care diagnosis was provided by the nonexpert cardiology fellow. Images were then sent for expert interpretation by two independent expert echocardiographers reading on Vscan Gateway workstation version 1.0.0.12 (GE Healthcare). A change in the point-of-care diagnosis was recommended if any identified major abnormality (i.e., ventricular function, valvulopathy, chamber size, congenital abnormality) was not matched and categorized at the same level of severity (mild, moderate, or severe) by either expert interpretation.
Smartphone-Based Analysis
A minimum of 4 weeks after the initial interpretation to minimize recall bias, the studies were randomly ordered, and in blinded fashion, one of the experts interpreted the same studies using mVisum onDemand version 1.0.0 (mVisum, Inc., Camden, NJ) on an iPhone 3GS with iOS version 4.1 (Apple Corporation, Cupertino, CA), which has an 8.9-cm diagonal, 480 × 320 pixel display. Figure 1 shows the layout of the system, which is adaptable to other smartphone operating systems (BlackBerry OS version 5.0 or later, Research in Motion, Waterloo, Ontario, Canada; Android version 1.6 or later, Google, Mountain View, CA; Apple iOS version 2.2.1 or later, Apple Corporation, Cupertino, CA; and Windows Mobile version 5.0 or later, Microsoft Corporation, Redmond, WA), but for consistency, only iOS version 4.1was tested. The mVisum handheld client software allows remote viewing of studies through the Internet (Wi-Fi) or cell phone data networks. To maintain greater consistency of connectivity, this analysis was performed only via Wi-Fi. The mVisum Platform server delivers images through a lossless review technique that scales the downloaded pixel densities to the pixel densities of the mobile device for any given level of magnification.
The iPhone client software allows standard window width and window level functions. The client software supports both still and cine image review within a single study. It allows pinch zoom capabilities (by pinching two fingers apart or together on the screen) for both still images as well as cine loops. It also supports panning around the zoomed image using the standard Apple touch interface. For cine loops, it allows continuous looping along with play and pause as well as scroll bar–driven fast-forward and rewind functions. Images can be viewed in landscape or portrait mode.
Statistical Analysis
All means are stated with their standard deviations. Interobserver agreement between expert readers on the workstation and intertechnology agreement with the same expert reader were assessed using Cohen’s κ test, and 95% confidence intervals were calculated. Statistical significance was assessed using χ 2 tests. SAS version 9.2 (SAS Institute Inc., Cary, NC) was used for statistical testing. P values < .05 were considered statistically significant.
Results
Patient Demographics
The characteristics of the 89 evaluated patients are detailed in Table 1 . Indications for echocardiography are listed in Table 2 . The most common echocardiographic abnormalities were left ventricular systolic dysfunction (24%), left ventricular dilatation (8%), mitral regurgitation (7%), right ventricular dilatation (6%), tricuspid regurgitation (4%), abnormal right ventricular systolic function (3%), and mitral stenosis (3%). All other significant findings were seen in <3% of patients; 46% of studies had no identified abnormalities.
| Variable | Value |
|---|---|
| Age (y), mean ± SD | 54 ± 23 |
| Women | 58% |
| Hypertension | 37% |
| Cardiomyopathy | 13% |
| Complete heart block | 13% |
| Hyperlipidemia | 10% |
| Sick sinus syndrome | 8% |
| Atrial fibrillation | 7% |
| Diabetes | 7% |
| Coronary artery disease | 6% |
| Ventricular tachycardia | 3% |
| Tobacco use | 3% |
| Indication | n (%) |
|---|---|
| Arrhythmia | 29 (33) |
| Cardiomyopathy | 25 (28) |
| Syncope/presyncope | 13 (15) |
| Palpitations | 11 (12) |
| Chest pain | 10 (11) |
| Dyspnea | 10 (11) |
| Coronary artery disease | 5 (6) |
| Valvular heart disease | 3 (3) |
| Congenital heart disease | 3 (3) |
Image Characteristics
A mean of 8.5 ± 3.0 cines were acquired per study. The average file size per cine was 249 ± 144 KB. By Wi-Fi, smartphone download time approximated 24 ± 10 sec/study.
Point-of-Care Diagnostic Accuracy
Compared with remote expert interpretations, the point-of-care diagnoses provided by the cardiology fellow were changed in 35 of the 89 patients (38% of cases): 41% of these with corrections in left ventricular function (e.g., change from mildly reduced to normal function), 38% with valvulopathy corrections (e.g., change from moderate to mild mitral regurgitation), 18% with image quality insufficient for verification of point-of-care diagnosis, 12% with chamber size corrections (e.g., change from a mildly dilated left ventricle to normal left ventricular size), and 12% with congenital heart disease diagnoses (e.g., suspected atrial septal defect to no identified atrial septal defect). The details of the diagnostic changes are listed in Table 3 . Representative images from a study for which the point-of-care diagnosis was changed are shown in Figures 2 and 3 and Videos 1 to 5 (view video clips online).
| Point-of-care diagnosis | Corrected diagnosis |
|---|---|
| Correction in LV systolic function | |
| Moderate LV dysfunction | Mild LV dysfunction |
| Mild LV dysfunction | Normal LV function |
| Mild LV dysfunction | Normal LV function |
| Inferior wall hypokinesis with preserved LV function | Severely reduced LV function |
| Normal LV function | Moderate LV dysfunction |
| Moderate LV dysfunction | Normal LV function |
| Mild LV dysfunction | Severe LV dysfunction |
| Normal LV function | Mild LV dysfunction |
| Mild LV dysfunction | Normal LV function |
| Correction in valvulopathy | |
| Mild mitral regurgitation, trace aortic regurgitation | Trace mitral regurgitation, mild aortic regurgitation |
| Moderate tricuspid regurgitation | Mild tricuspid regurgitation |
| Mild mitral regurgitation | Trace mitral regurgitation |
| No aortic regurgitation | Mild aortic regurgitation |
| Moderate mitral regurgitation | Mild aortic regurgitation |
| Severe aortic stenosis | No aortic stenosis |
| Moderate aortic stenosis | No aortic stenosis |
| Trace mitral regurgitation | Mild mitral regurgitation |
| Inadequate study quality to verify diagnosis | |
| Normal study | Inadequate study quality |
| Mild LV systolic dysfunction | Inadequate study quality |
| Mild LV systolic dysfunction | Inadequate study quality |
| Mild biventricular dilation | Inadequate study quality |
| Moderate LV systolic dysfunction | Inadequate study quality |
| Normal study | Inadequate study quality |
| Multiple corrections | |
| Mild RV dilation, mild mitral regurgitation | No RV dilation, trace mitral regurgitation |
| Severe LV systolic dysfunction, severe RV dilation | Moderate LV systolic dysfunction, no RV dilation |
| Mild biventricular dilation, membranous ventricular septal defect | No biventricular dilation, no ventricular septal defect |
| Mild LV systolic dysfunction, mild biventricular dilation | Normal LV systolic function, no biventricular dilation |
| Moderate LV systolic dysfunction, no mitral stenosis | Normal LV systolic function, mild mitral stenosis |
| Mild LV systolic dysfunction, severe mitral stenosis | Normal LV systolic function, no mitral stenosis |
| Atrial septal defect, moderate tricuspid regurgitation | No atrial septal defect, mild tricuspid regurgitation |
| Inferior wall hypokinesis with preserved LV systolic function, moderate aortic regurgitation | Normal LV systolic function without regional wall motion abnormality, trace aortic regurgitation |
| Other corrections | |
| Mild LV hypertrophy | No LV hypertrophy |
| Membranous ventricular septal defect | No ventricular septal defect |
| Ventricular septal defect, no atrial septal defect | Atrial septal defect, no ventricular septal defect |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
