Diagnostic Accuracy of a Hand-Held Ultrasound Scanner in Routine Patients Referred for Echocardiography


The aim of this study was to investigate the imaging capabilities of recent hand-held ultrasound scanners.


Three hundred forty-nine patients were scanned with hand-held ultrasound (HAND) and high-end echocardiography (HIGH). Segmental endocardial border delineation was scored (2 = good, 1 = poor, 0 = invisible) to describe image quality. Assessments of left ventricular (LV) dimensions, regional and global LV function, and grades of valve disease were compared.


The mean endocardial visibility grades were 1.6 ± 0.5 with HAND and 1.7 ± 0.4 with HIGH ( P < .01). Regional wall motion was scored very similarly (κ = 0.73, P < .01). Ejection fraction assessment (bias = 1.8%, 1.96 × SD = 8.3%) and LV measurements ( r = 0.99, P < .01; interventricular septum: bias = 0.91 mm, 1.96 × SD = 2.1 mm; LV end-diastolic diameter: bias = 0.5 mm, 1.96 × SD = 4.1 mm; LV posterior wall: bias = 0.61 mm, 1.96 × SD = 2.4 mm) showed negligible deviations. No pericardial effusion or valve stenosis was missed. Regurgitations missed by HAND were all graded “minimal” on HIGH. Regurgitations were mildly overestimated by HAND. Overall concordance for detection of regurgitations was very good (κ = 0.9, P < .01).


Handheld echocardiography was feasible and missed no relevant findings. Given the future implementation of spectral Doppler capabilities, this handheld scanner can safely be used in clinical routine.

Since its introduction by Edler and Hertz in 1954, echocardiography has become an indispensable diagnostic tool in cardiac care. However, ultrasound machines were once large and heavy, limiting their easy use at the bedside. Early portable machines were heavy, had limited functionality and image quality, and were shown to miss important clinical findings.

The recent introduction of handheld ultrasound (HAND) scanners the size of mobile phones allows these devices for the first time to be carried in one’s white coat, facilitating true bedside routine use. (Note that previous publications have used inconsistent terms for smaller echocardiographic machines. In this paper, we use the term hand-held for pocket scanners the size of mobile phones and the term portable for all other hand-carried devices.) However, concerns about image quality remain. Therefore, we investigated the feasibility and imaging capabilities of the most recent and most advanced representative of this class of devices in comparison with contemporary high-end echocardiography (HIGH).


Study Population and Echocardiographic Protocol

During a 5-week period, 349 unselected consecutive routine patients from the echocardiography lab at University Hospital Gasthuisberg (Leuven, Belgium) were scanned with a HAND scanner (Vscan; GE Vingmed Ultrasound AS, Horten, Norway; Figures 1 A and 1 B) and modern HIGH scanners, such as Vivid 7 or E9 (GE Vingmed Ultrasound AS) as the gold standard (using the M3S probe [GE Vingmed Ultrasound AS] at 1.5–4.0 MHz). Harmonic imaging was used in HIGH scanners by default. The examinations with HAND were performed by an experienced cardiologist after informed consent was obtained, directly before or after the regular echocardiographic examination with the HIGH scanner, which was performed independently by an experienced echocardiographer. Each examiner was blinded to the results of the other examination. Grayscale and color Doppler recordings from both scanners (apical four-chamber, three-chamber, and two-chamber and parasternal long-axis and short-axis views) were digitally stored for further offline assessment. Feasibility and handling issues were documented after each examination.

Figure 1

The Vscan HAND device.

During offline analysis, images were read from a regular computer display using dedicated software. Image size was slightly bigger than on the device, but image resolution was identical. A single echocardiographer reviewed offline HAND and HIGH examinations of all subjects separately and in changing order, completely blinded to results of previous readings.

Specification and Capabilities of the HAND Device

The display unit of the Vscan measures 135 × 73 × 28 mm. The probe measures 120 × 33 × 26 mm. The device and probe together weigh 390 g. The portrait-oriented (3:4) display has a diagonal dimension of 3.5 in (8.9 cm), with a resolution of 240 × 320 pixels. The displayed image sector for black and white imaging is 75°, with a maximum depth of 25 cm. The device offers regular gryascale imaging and color blood flow mode. The color flow box has a fixed size but can be moved with cursor keys. The bandwidth of the phased-array probe is 1.7 to 3.8 MHz. The gain is adjusted automatically for all depths. The device offers an image based “Auto-Cycle” function for the automatic detection of a full heart cycle beginning with end-diastole. If this detection fails, a 2-second loop is stored. Distance measurements can be performed during an examination. The minimalistic keyboard contains only eight buttons. Voice recording is used for patient identification. Stills, image loops, and voice recordings are stored on a 4-GB micro-SD memory card in MP4 format and can be copied to a PC via the included docking station or directly from the card.

Data Analysis

For analysis, HIGH data were transferred to a PC running EchoPAC software (GE Vingmed Ultrasound AS). Data from the HAND scanner were also transferred to a PC and reviewed using the dedicated Vscan Gateway software. Data analysis was performed in random order by an experienced cardiologist blinded to any patient data. A score for segmental endocardial border delineation was used to describe image quality in all recordings, as follows: 2 = good, 1 = poor, and 0 = not possible. Standard measurements of left ventricular (LV) dimensions (interventricular septal diameter, LV end-diastolic diameter, and LV posterior wall diameter) as well as visual assessment of regional (segmental wall motion score: 1 = normokinesia, 2 = hypokinesia, 3 = akinesia) and global (visually estimated ejection fraction [EF]) LV function were compared between HAND and HIGH. In both HAND and HIGH readings, EF was visually assessed using prespecified levels in intervals of 5%.The severity of valve regurgitation (0 = none, 1 = minimal, 2 = mild, 3 = moderate, 4 = severe) was graded according to cardiac morphology and the visual interpretation of the color Doppler jet. On the HIGH readings, the severity of a valve stenosis (0 = none, 1 = mild, 2 = moderate, 3 = severe) was evaluated using standard methods of routine echocardiography, such as pressure gradients and opening area calculation by continuity equitation or pressure half-time. On HAND readings, grading of valve stenosis was based on the interpretation of the available grayscale (chamber hypertrophy, valve calcification, and mobility) and color Doppler information (turbulence). The assessments were compared between machines.

Statistical Analysis

Continuous data are expressed as mean ± SD. Statistical analyses were performed using PASW (SPSS, Inc., Chicago, IL). For continuous and normally distributed data, paired t tests were used, and for non-normally distributed data, Wilcoxon’s signed-rank tests were used. A two-tailed P value < .05 was considered significant. Continuous measurements were compared using Spearman’s correlation and Bland-Altman analysis. To determine the level of agreement between grades, we used weighted κ statistics. Kappa values < 0.2 were interpreted as slight, 0.21 to 0.4 as fair, 0.41 to 0.6 as moderate, 0.61 to 0.8 as substantial, and 0.81 to 1.00 as very good agreement.


Patient Population

We examined 349 consecutive routine patients (196 men). The mean age was 61.5 ± 15.0 years (range, 20.0–89.0 years). The patients had a mean body mass index of 25.8 ± 4.7 kg/m 2 (range, 15.0–48.0 kg/m 2 ). The LV EF ranged from 20% to 70% (mean, 55.6 ± 10.0%).

Use of the HAND Device

Operating the device was easy and intuitive. Basic functions were found quickly by trial and error. After reading an A4-size instruction card, all functions could be controlled without problems. Lifting the screen switches the device on and initiates a new examination. We measured a booting time of 25 seconds. Voice recordings are used to identify the patient. The rechargeable battery allowed approximately 70 minutes of work. The time to recharge the battery was about 2 hours.

For proper data storage without electrocardiographic triggering, the device relies on detecting cardiac cycles on the basis of the cyclic changes in the image. This algorithm works rather robustly and failed in 8.2% of all cases, leading to a sometimes disturbed image sequence.

Image Quality

The segmental visibility of the endocardium was graded with mean scores of 1.6 ± 0.5 per patient with HAND and 1.7 ± 0.4 with HIGH as the gold standard ( P < .01). In both echogenic and more difficult to scan patients, image quality was comparable. Only in patients with very bad echogenicity did HIGH tend toward advantageously better image quality ( Figures 2 A– 2 C). No frame rate information is provided on the display of the device. Analysis of the stored image loops revealed a constant frame rate of 20 frames/sec independent of depth settings for grayscale image loops and 13.6 frames/sec in color Doppler mode.

Figure 2

Comparison of image quality of the HAND device (top) and a HIGH device (bottom) . (A) Echogenic patient. (B) Patient with suboptimal echogenicity. (C) Patient with bad echogenicity. (D) Patient with pericardial effusion. The pathology is somewhat more easily detected with the wider image sector of the HIGH device.

Functional Assessment

Visual EF estimates from HAND and HIGH image data correlated well ( r = 0.91, P < .01). Bland-Altman analysis revealed no relevant bias (1.8%) and a normal variation (1.96 × SD = 8.3%) ( Figure 3 ). Regional wall motion was scored similarly on HAND and HIGH readings (κ = 0.73, P < .01; Table 1 ).

Figure 3

(Left) A good correlation ( r = 0.91, P < .01) was found between readings with both types of devices. (Right) Bland-Altman plot shows no relevant bias and a normal variation of assessments. Data points of several patients are clustered because of the EF assessment in prespecified levels of 5%.

Table 1

Grading of segmental wall motion ( n = 6,175 segments) with the HAND device and the HIGH scanner

HIGH scanner
NG 1 2 3
HAND device (Vscan)
NG 6 119 6 2
1 3 5614 4 0
2 0 44 244 16
3 0 11 36 70

NG, Not gradable; 1, normokinesia; 2, hypokinesia; 3, akinesia.


LV measurements on HIGH and HAND image data correlated very well ( r = 0.99, P < .01). Bland-Altman analysis showed no relevant bias or scatter (interventricular septum: bias = 0.91 mm, 1.96 × SD = 2.1 mm; LV end-diastolic diameter: bias = 0.5 mm, 1.96 × SD = 4.1 mm, LV posterior wall: bias = 0.61 mm, 1.96 × SD = 2.4 mm; Figure 4 ).

Figure 4

Measurements of LV dimensions. (Left) Measurements from both types of devices correlated very well. (Right) Likewise, the Bland-Altman plot reveals no relevant bias and low variation between measurements.

Detection of Pericardial Effusion

Six study patients (1.7%) had pericardial effusions, which were detected equally with both devices ( Figure 2 D).

Regurgitant Lesions

The device’s color Doppler mode is an advantage over earlier pocket echocardiographs, which did not provide this feature ( Figure 2 E). The relatively low frame rate is compensated for by high sensitivity, so that the color Doppler performs well when screening valve lesions in clinical practice.

Using HIGH as the gold standard, a total of 330 patients with mitral regurgitation were detected, and nine mild mitral regurgitations were missed on HAND readings (2.6%). One mild mitral regurgitation was detected with HAND only (0.3%). The overall concordance for the detection of mitral regurgitations was substantial (κ = 0.80, P < .01).

Among 128 patients with aortic regurgitation on HIGH readings, eight (6.3%) were missed on HAND readings. Two regurgitations were detected only by HAND. The concordance for aortic regurgitations was very good (κ = 0.94, P < .01).

Tricuspid regurgitation was observed in 324 patients with HIGH and in 291 patients (89.8%) with HAND, resulting in moderate concordance (κ = 0.6, P < .01; Table 2 ).

Jun 15, 2018 | Posted by in CARDIOLOGY | Comments Off on Diagnostic Accuracy of a Hand-Held Ultrasound Scanner in Routine Patients Referred for Echocardiography

Full access? Get Clinical Tree

Get Clinical Tree app for offline access