| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Department of Bioengineering, University of Pittsburgh, 749 Benedum Hall, Pittsburgh, PA 15261. From the 2004 RSNA Annual Meeting. Received September 27, 2005; revision requested November 14; revision received December 21; accepted January 24, 2006; final version accepted February 1. Supported by National Institutes of Health grants 1-R01-EB00860-1 and 1-R01-HL074285-01. Address correspondence to W.M.C. (e-mail: wmchang{at}gmail.com).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
Materials and Methods: Institutional review board approval and oral and written informed consent were obtained. The sonic flashlight replaces the standard US monitor with a real-time US image that appears to float beneath the skin and is displayed where it is scanned. In studies 1 and 2, participants performed sonic flashlight–guided needle insertion tasks in vascular phantoms. In study 1, 16 participants (nine women, seven men) with no US experience performed 60 simulated vascular access trials with sonic flashlight or conventional US guidance. With analysis of variance (ANOVA) and power-curve fitting, improvement with practice rate and mean differences between techniques and tasks were examined. In study 2, 14 female nurses (mean age, 50.1 years) proficient with conventional US performed simulated vascular access trials on three tasks with the sonic flashlight and conventional US. With random assignment, half the participants used the sonic flashlight first and half used conventional US first. Mean performance with each technique and that with each task were compared by using ANOVA. In study 3, feasibility of sonic flashlight guidance for access to internal jugular and basilic veins was demonstrated in a cadaver.
Results: For study 1, learning rates (ie, decrease in access time over trials) did not differ for vascular access with sonic flashlight and conventional US. Overall, participants achieved faster vascular access times with sonic flashlight guidance (P < .007). In study 2, participants performed procedures faster overall with the sonic flashlight (P < .02) and found the sonic flashlight easier to use. In study 3, sonic flashlight–guided vascular access was gained in the cadaver.
Conclusion: Learning and performance of vascular access were significantly faster with the sonic flashlight than with conventional US, and vascular access could be gained in a cadaver; the sonic flashlight is ready for clinical trials.
Supplemental material: radiology.rsnajnls.org/cgi/content/full/241/3/771/DC1
© RSNA, 2006
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
To address this difficulty, some researchers have explored nonconventional methods for viewing the US image, patient, instrument, and operator's hands in one environment. Head-mounted display systems have been developed to display a US image as if within the patient (1–4). Despite their promise, head-mounted display systems have yet to overcome substantial obstacles, including lag time, low resolution, limited field of view, weight, and expense. Furthermore, if multiple observers are cooperating in a procedure or are involved in training, each observer requires a separate head-mounted display to observe the same in situ US image.
The sonic flashlight, a device in development at our institution, displays real-time US images inside the patient without the use of positional tracking or a head-mounted display system (5,6). The sonic flashlight fixes the relative geometry of the transducer, display, and a half-silvered mirror, which the operator looks through, to produce a virtual image of the US data inside the patient (Figs 1, 2). The US image appears to float beneath the surface of the skin. It is a virtual image in the exact optics sense of the word. For all intents and purposes, each pixel of the US image emanates from its correct anatomic location within the patient, as if being illuminated directly by the sonic flashlight (Fig 2).
|
|
|
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
Sonic Flashlight Prototype
The sonic flashlight prototype is built around a Food and Drug Administration–approved commercially available 10-MHz US probe (Terason 2000; Teratech, Burlington, Mass), a small 44 x 33-mm flat-panel organic light-emitting display (AM550L; Kodak, Rochester, NY), and a 25 x 50 x 2-mm half-silvered mirror with 30% reflectance (Edmund Optics, Barrington, NJ) (Fig 1). Since the sonic flashlight can display US images up to only the size of the flat-panel display, the current prototype is limited to a region 44 mm deep and 33 mm wide. The US probe and the flat-panel display are fixed at 80° on opposite sides of the mirror by a rigid mount.
The US probe on which this version of the sonic flashlight is built is approximately 16 x 54 x 92 mm with a 14 x 54-mm scanning footprint. The sonic flashlight retains the same scanning footprint as that of the probe of 14 x 54 mm, and the size of the entire sonic flashlight is approximately 44 x 57 x 133 mm. The US data from the transducer are transmitted to a laptop computer (Latitude C840; Dell, Round Rock, Tex), which performs the rotation, scale, and translation necessary to display the US image at its correct size and position on the flat-panel display. The US system refresh rate is 22 frames per second, and the components of the sonic flashlight add no appreciable latency (<11 msec as measured with software). The digital US data contain 512 x 128 pixels, which are displayed on the flat-panel display with 521 x 218-pixel resolution, and no loss of display resolution occurs.
Vascular Phantom
We used a custom vascular phantom (Blue Phantom; Blue Phantom Division, Advanced Medical Technologies, Kirkland, Wash) that contained three vessels, which were labeled vessel 1, vessel 2, and vessel 3 (Fig 4). Vessel 1 is a bifurcating vessel that is 5 mm in diameter and is located 9 mm from the surface of the phantom. Directly beneath vessel 1 lies vessel 2, which is a 3-mm-diameter vessel that is 20 mm from the surface of the phantom. Vessel 2 also bifurcates but in the opposite direction from vessel 1. When viewed from above, the two vessels appear similar to a letter Y stacked on top of an upside down Y. In a separate region from vessels 1 and 2 lies vessel 3, which is a straight vessel that is 4 mm in diameter and is located 15 mm from the surface of the phantom.
|
Study 1: Learning and Statistical Analysis
Sixteen medical students with no US experience were randomized to either the sonic flashlight group or the conventional US group, with eight participants per group. The sonic flashlight group included three men and five women (mean age, 23.1 years; range, 21–25 years). The conventional US group included four men and four women (mean age, 24.3 years; range, 22–28 years). Participants attended standardized tutorials about how to use their respective device and how to perform the procedures before starting. The tutorials consisted of a description about how to use their respective device; what the US image represented; how to locate, aim, and guide a needle in a cross-sectional US scan of a vessel; and, last, a demonstration. Both tutorials included instruction about how to gauge depth and size of the target.
The tutorials for the two techniques differed only in where to look at the US image and how to aim at a target within the image. For both techniques, participants were instructed to orient the probe at a 90° angle to the surface of the phantom. The participants in the conventional US group were instructed to center the target horizontally within the scanning plane and to insert the needle out of plane at an approximately 45° angle relative to the scanning plane. They were shown that the needle entry point would thus be the same distance away from the scanning plane as the target depth. The participants in the sonic flashlight group were instructed to insert the needle out of plane at any entry angle. Both tutorials lasted approximately 5 minutes. All tutorials were conducted by one individual (W.M.C., with 2 years of experience with the sonic flashlight and conventional US).
The conventional US machine used in this study was an unmodified unit with a 10-MHz probe, a probe that was identical to that used in the construction of the sonic flashlight. Participants were asked to perform three tasks: tasks A, B, and C (Fig 4). In task A, participants were asked to guide a needle between the bifurcation of vessel 1, without hitting vessel 1, and to access vessel 2. The operator was constrained to a region of the phantom with a gap of less than 23 mm between the bifurcations in vessel 1. In task B, the participants were asked to guide a needle into the right bifurcation branch of vessel 2. In task C, the participants were asked to guide a needle into vessel 3. Participants used a 21-gauge 7-cm-long needle (Boston Scientific, Natick, Mass) for all tasks. Task B was designed to be the most difficult, and task C was designed to be the easiest.
Total time from that when the probe first touched the phantom to a needle flash (ie, the correct colored fluid filled the needle hub) was recorded and included multiple attempts. If at any time the needle entered an incorrect vessel, the participant was asked to remove the needle completely from the phantom and reattempt guidance into the correct vessel. A trial was defined as the completion of one task, and time for completion was recorded from the time the probe touched the phantom until the needle flash occurred. In the course of two sessions, which were approximately 1 week (7–9 days) apart, participants were asked to perform 30 trials per session, for a total of 60 trials. Each session included a series of three-trial blocks, and each block contained one trial each of tasks A, B, and C performed in randomized order (eg, tasks B, A, and C; tasks C, B, and A; tasks A, C, and B, etc). Upon completion of the study, each participant had completed 20 successive blocks or 20 trials at each task.
The data analysis focused on three issues: First, is there a difference in the time to perform vascular access with the sonic flashlight compared with that with conventional US? Second, do users of the techniques become more proficient (ie, demonstrate learning) with time, and if so, is the rate of learning different for the two techniques? Third, are the first two questions moderated by the difficulty of the task being performed? For example, do the techniques differ most when the task is most difficult? These issues were addressed with analysis of variance (ANOVA), a method that allows the variability in a set of observations to be attributed to the manipulated variables in an experiment or, alternatively, to be attributed to noise. The test for a significant difference was applied to each variable and to interactions among variables.
We assessed effects of technique and learning by using a mixed ANOVA, which included factors of technique (guidance with the sonic flashlight vs guidance with conventional US, between-participant factor), trial (n = 20, within-participant factor), and task (n = 3, within-participant factor). In this analysis, the main effect of technique was used to test for mean differences in access time between the use of the sonic flashlight and that of conventional US. Learning, or the decrease in access time with practice, was indicated by an effect of trial, and variations in the rate of learning across techniques were indicated by the interaction between trial and technique. The interaction between task and technique indicates whether the effects of technique are comparable, given tasks of different difficulty, and the interaction between task and trial similarly indicates whether the learning rate varies with the difficulty of the task. For all ANOVAs, 
was set at a significance level of .05.
To compare the techniques after performance had stabilized, we considered only the last five trials with each target. We compared guidance with the sonic flashlight and that with conventional US with a one-tailed t test, given our a priori hypothesis of an advantage for the sonic flashlight. For this comparison, was set at a significance level of .05.
In our final analysis, we fit the access times as a function of trial number with a power function to describe the rate of learning. To quantify the learning rate, we fit a power function to the mean access time according to trial number for each technique. The function takes the following form: T = a · N–b, where T is access time, a is baseline access time, N is trial number, and b is the learning rate.
If one considers the relationship between log performance time and log of trial number, the parameter a represents the baseline access time and b is a parameter that indicates the learning rate. An alternative to the power function would be an exponential function. Generally, however, human performance curves are approximated less well with an exponential function, which assumes a constant proportional decrease in access time over trials, than with a power function, which assumes that the returns from practice diminish over trials (7). Study 1 statistics were performed by using software (StatView, version 5.0.1, 1998; SAS Institute, Cary, NC).
Study 2: Proficient Conventional US Users and Statistical Analysis
In this study, we compared the use of the two techniques in a more skilled population. The study population consisted of 14 intravenous access team nurses (14 women; mean age, 50.1 years; range, 39–66 years) from our institution who were trained in US-guided peripherally inserted central catheter placement and placed these catheters in patients at the bedside on a daily basis. By using the same vascular phantoms and tasks as in study 1, participants (with exceptions noted later) performed 24 vascular access trials with the sonic flashlight and 24 with conventional US in two sets where the technique was held constant. Across participants, order of techniques was counterbalanced. Use of the second technique immediately followed use of the first technique, with no separation in time. The time of each trial was recorded. The trials included three-trial blocks, and each block consisted of the three tasks in random order in the same manner as was used in study 1.
For each participant, the first six trials (two trials with each task) were considered practice trials to familiarize the participants with the experimental procedures and equipment. Therefore, six trials per participant for each combination of task and technique were used in the data analysis. The first three participants only completed 18 trials per technique, and thus only four trials per task with each technique were available for data analysis. This was a result of a protocol change after the study began, and this change was implemented to increase the amount of data collected. The conventional US machine used in this study was an unmodified unit with a 10-MHz probe, a probe that was identical to that used in the construction of the sonic flashlight. Before use of each technique, the participants attended the same standardized tutorial about the use of the sonic flashlight and conventional US as those in study 1 attended. All tutorials were conducted by the same individual who conducted them in study 1.
ANOVA of the mean access time was performed with three factors: technique (sonic flashlight vs conventional US, within-participant factor), task (tasks A, B, and C; within-participant factor), and order of technique (sonic flashlight in first set vs conventional US in first set, between-participant factor). The value for was set at a significance level of .05. The main effects in this analysis were used to test for differences between the techniques and the tasks, and the interaction between technique and task addresses whether the difference between guidance with conventional US and that with the sonic flashlight varies with the task. The further inclusion of order of technique as a factor addresses whether the outcomes were affected by which technique the participant used first, in which case interactions between other factors and order of technique would be found. Statistical analysis for study 2 was conducted by using the same software as was used for statistical analysis in study 1 and spreadsheet software (Excel, version 11.2, 2004; Microsoft, Redmond, Wash).
After completion of all the trials with both techniques, participants were asked to complete a questionnaire containing six subjective questions that were used to compare guidance from the sonic flashlight with that from conventional US. The questions included a set of responses that could be chosen for assessment and addressed the following: (question a) ease of procedural performance (assessed with a response scale of 1–5, where 1 signified "much easier" and 5 signified "much harder"), (question b) ease of US interpretation (assessed with the same response scale as was used for question a), (question c) degree to which the mirror impeded the procedure as a result of blocking the view (assessed with a response scale of 1–5 where 1 signified "disagree strongly" and 5 signified "agree strongly"), (question d) degree to which sighting through the mirror increased the difficulty of the procedure (assessed with the same response scale as was used for question c), (question e) effect of the smaller image from the sonic flashlight on interpretation of the US image relative to the conventional US image (assessed with a response scale of 1–5 where 1 signified "much easier" and 5 signified "much harder"), and (question f) whether the image with the sonic flashlight helped or hindered aim and guidance for needle placement (assessed with a response scale of 1–3, where 1 signified "helped" and 3 signified "hindered").
Study 3: Cadaveric Vascular Access
The goal of this study was to perform vascular access in a cadaver to validate the use of the sonic flashlight in the human. In contrast to the first two studies, this study was not performed to attempt to establish statistically reliable efficacy but rather was simply undertaken to demonstrate feasibility. The cadaver was a woman of unrevealed age and cause of death who had received heparin in the course of her treatment prior to death. The neck and right upper arm were scanned by using the sonic flashlight to identify the internal structures, and a 21-gauge 7-cm-long needle was aimed and inserted into the internal jugular vein and basilic vein, sites that would normally be used for a central catheter and a peripherally inserted central catheter, respectively. At each location (neck, arm), there were three needle insertions, and no artifacts were noted from the introduction of air. Successful entry into the lumen was determined by a needle flash in the needle hub. Vascular access was obtained by a practicing interventional radiologist with more than 10 years of experience (N.B.A.).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
. There was also an effect of task (F = 39.18; df = 2, 28; P < .001); performance of task C was fastest and that of task B was slowest for both techniques. The analysis of interaction between task and technique revealed that the advantage of the sonic flashlight varied with the task (F = 6.43; df = 2, 28; P = .005). Finally, there was a main effect of trial (F = 25.54; df = 19, 266; P < .001) and an interaction between trial and task (F = 4.13; df = 38, 532; P < .001), and these results indicated that there was a decrease in performance time across trials (ie, learning) that varied with the target vessel. Note that there were no significant interactions involving trial and technique, which indicated that the rate of decrease in access time with practice did not significantly differ between the techniques: For interaction between trial and technique, the values were F = 0.93, df = 19 and 266, and P = .542. For interaction among trial, technique, and task, the values were F = 1.09, df = 38 and 532, and P = .334.
|
Accordingly, to compare the techniques after performance had stabilized, we considered only the last five trials with each target and compared the sonic flashlight and conventional US with a t test (one-tailed, given our a priori hypothesis of an advantage for the sonic flashlight). All tasks showed a significant advantage for the sonic flashlight of, on average, 1.3, 3.6, and 1.7 seconds with P = .024, .005, and .002, for tasks A, B, and C, respectively.
To quantify the learning rate, we fit a power function to the mean access time according to trial number for each technique. The functions accounted for between 80% and 93% of the variance (R2) in performance time with trials (Table). Although, for all three tasks, the learning rate parameter, parameter b, fit to the sonic flashlight data was greater than that fit to the conventional US data, these differences were not significant with t tests (two-tailed) for each task for which we used the standard error of estimation from the power regression: For tasks A, B, and C, respectively, t = 1.59, 0.42, and 1.03 and P = .112, .674, and .303, with df = 38.
|
. There was a significant effect of task, with F = 3.82, df = 2 and 24, and P < .037. The access times were greater for task B (mean, 9.1 seconds) than for tasks A and C (mean, 7.2 and 7.5 seconds, respectively). The interaction between technique and task, which would indicate that the advantage for the sonic flashlight varied with the task, was not significant. We note, however, that the interaction approached significance, with F = 2.90, df = 2 and 24, and P = .074; this finding reflects the fact that the advantage of the sonic flashlight was greatest for the slowest task, task B.
|
Study 3: Cadaveric Vascular Access
The internal anatomy was visualized in situ by using the sonic flashlight, with the carotid artery and internal jugular vein identified in the neck and the basilic vein identified in the arm. The needle was aimed and inserted into the internal jugular vein and the basilic vein at the first attempts, and the needle tip was visualized at the expected location (Figs 7, 8). When the needle entered the veins, blood freely flowed out of the needle hub (Fig 8).
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
In study 1, we compared the effects of practice and initially hypothesized that the sonic flashlight group would show faster learning and have a faster vascular access time once they achieved proficiency. Results of this study support the general conclusion that the sonic flashlight leads to overall faster performance than does conventional US. Although the learning rate was not shown to be significantly different, the advantage of the sonic flashlight is initially present and remains essentially constant with practice (P < .006). Results in study 2 showed that proficient conventional US users performed vascular access significantly faster with the sonic flashlight (P < .02), despite that they had no prior training with the device, and that participants almost unanimously judged the procedures easier to perform with the sonic flashlight. These results strongly suggest that performance with the sonic flashlight is easier and faster than with conventional US, at least in phantoms. Findings in our feasibility study in a cadaver support this suggestion.
Our choice of a through-plane insertion of the needle reflects the approach used by our particular clinical colleagues in placing peripherally inserted central catheters and other central catheters. Venous access and other US-guided procedures often are performed by constraining the needle to the plane of the scan, with or without a needle guide, and showing the complete path of the needle to avoid hitting critical structures. The sonic flashlight also can be used to perform in-plane procedures, without the constraint of a needle guide. Furthermore, for through-plane procedures, the operator can sweep the sonic flashlight through the path of the needle to avoid critical structures and provide an in situ three-dimensional sense of the anatomy.
Our positive conclusions about the promise of the sonic flashlight should be tempered by examining the limitations of this study. The measurement used in studies 1 and 2 to compare the sonic flashlight and conventional US was the time to successful vascular access, with all trials ending in successful access. A better metric in the clinical setting would be successful versus failed access, since these are the actual outcomes with clinical importance. Failure to achieve access by the nurse at the bedside usually leads to referral to the interventional radiologist. The selection of time as the measurement for comparison was necessitated by the limitations of the current vascular phantoms. No phantom currently available, to our knowledge, can simulate the difficulty of vascular access in real tissues (eg, vessels "rolling" away from the needle and heterogeneous tissue types). Therefore, it was unlikely for a participant to fail in an access attempt.
Other measurements were considered but were deemed unreasonable or impractical to measure. For example, although the number of attempts, or sticks, per successful access might seem like a logical measurement, it is difficult to clearly define one attempt: What if the needle is only partially withdrawn from the phantom and the trajectory is re-aimed, with the needle tip remaining within the phantom? Time to success was the only consistently measurable variable. We plan to compare guidance with the sonic flashlight and conventional US in the clinical setting by performing a randomized controlled trial by using guidance with these techniques in human patients and measuring actual success and failure rates in addition to mean access times. Before this large-scale study can be performed, a small-scale safety and feasibility study will first be conducted.
In summary, we demonstrated the first comparison between the sonic flashlight and conventional US. First, novices who are learning US-guided vascular access perform consistently faster with the sonic flashlight than with conventional US throughout their learning to proficiency. Second, performance in users already proficient in conventional US guidance is faster with use of the sonic flashlight, despite no prior experience with the device, and these users find the sonic flashlight subjectively easier to use. With more experience in the use of the sonic flashlight, we would expect the differences in access times and ease of use to further increase. Third, we showed that vascular access is possible in the cadaver. Therefore, we believe that the sonic flashlight is ready for initial clinical trials.
Practical application: Performance with the sonic flashlight is faster even early in the course of learning, and this finding suggests that the device may be particularly well suited for use in areas of medicine where use of US has not been widespread. These areas include emergency medicine, critical care medicine, and anesthesiology. Although this work focused on vascular access, the sonic flashlight could be used in other applications currently performed by using conventional US guidance, and these applications include biopsy and drainage catheter placement.
|
ADVANCES IN KNOWLEDGE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
|
FOOTNOTES |
|---|
Abbreviations: ANOVA = analysis of variance
Author contributions: Guarantors of integrity of entire study, W.M.C., N.B.A., G.D.S.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, W.M.C., R.L.K.; experimental studies, all authors; statistical analysis, W.M.C., R.L.K.; and manuscript editing, all authorsG.D.S. holds a patent on the sonic flashlight through the University of Pittsburgh, through which he would receive a percentage of any revenue. The patent has not been licensed at this time.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. Thomas, L. Moya, G. Avidan, K. Humphreys, K. J. Jung, M. A. Peterson, and M. Behrmann Reduction in White Matter Connectivity, Revealed by Diffusion Tensor Imaging, May Account for Age-related Changes in Face Perception. J. Cogn. Neurosci., February 1, 2008; 20(2): 268 - 284. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
is the angle between the flat-panel monitor and the half-silvered mirror, as well as the angle between the half-silvered mirror and the US section. Point P in the section and its corresponding location on the monitor are equidistant from the mirror along a line perpendicular to the mirror (distance = d). Because the angle of incidence equals the angle of reflectance (angle = 
