We describe a novel software system that utilizes automated algorithms to perform edge detection and wall tracking of high-resolution B-mode arterial ultrasound images, combined with synchronized Doppler waveform envelope analysis, to calculate conduit arterial blood flow (BF) across the cardiac cycle. Furthermore, we describe changes in brachial arterial BF to the resting forearm during incremental cycle ergometry in eight subjects. During exercise, peak BF during the cardiac cycle increased at each workload (P < 0.001), because of increased velocity in the presence of unaltered cross-sectional area. In contrast, mean BF calculated across each cardiac cycle decreased at lower workloads before increasing at 100 and 160 W (P< 0.001). Differences in the pattern of peak and mean cardiac cycle flows were due to the influence of retrograde diastolic flow, which had a larger impact on mean flows at lower workloads. In conclusion, BF can be measured with high temporal resolution across the cardiac cycle in humans. Resting brachial arterial flow, including retrograde flow, increases during lower limb exercise.
- high-resolution ultrasound
measurement of peripheral blood flow (BF) in humans has traditionally relied on fluid-filled or strain-gauge plethysmography. Given that appropriate controls are instituted (1), this method can be reliably used to measure relative changes induced by pharmacological agents that modify flow to a resting muscle bed (9). However, the technique is based on several assumptions regarding the geometric shape of the forearm or calf and the consistency of cross-sectional area (CSA) changes in proximal and distal segments and is also intolerant of motion artifact (23). Alternative methods have relied on arterial cannulation and constant infusion of an indicator, with subsequent calculation of flow from dilution kinetics in the venous effluent. Although this method has been used to successfully measure BF to human skeletal muscle at rest and during exercise, it possesses several limitations (21), is invasive, and does not provide data with good temporal resolution (17).
Recently, ultrasound-Doppler methodology has been utilized to measure BF through conduit arteries in humans (5, 6, 18-20). This method relies on BF calculation from CSA, derived from ultrasonic assessment of arterial diameter (D), and blood velocity (ν) measurement with the use of Doppler. It possesses the advantages of being noninvasive, provides absolute BF measures, and improves temporal resolution (17). However, previous studies have estimated arterial D based on relative diastolic and systolic blood pressure phases and/or derived time-averaged mean blood ν, to calculate weighted composite flow (5, 18-20). Alternatively, ν and D have been determined independently and then reassembled and time aligned post hoc to provide beat-to-beat average flow (6). Neither approach provides continuous assessment of BF changes across each cardiac cycle.
In the present study, we present a novel software analysis system that utilizes automated edge detection and wall tracking of high-resolution (12–15 MHz), B-mode arterial ultrasound images, combined with synchronized Doppler waveform envelope analysis at 20–30 Hz, to calculate conduit arterial BF across the cardiac cycle. This system was used during incremental lower limb cycle ergometer exercise to characterize mean and peak BF responses and to measure the magnitude of antegrade and retrograde brachial arterial BF to the resting forearm during each cardiac cycle.
We describe the hardware and software components of the BF analysis system, followed by details of a study undertaken in eight young (22.4 ± 4.4 yr), healthy male volunteers.
The hardware configuration consisted of a Pentium IV 1.7-GHz personal computer (PC) with 256-MB random access memory, G400 32-MB AGP Matrox Millenium graphics card, and a 76-GB hard disk running Windows 2000 SP1, used in conjunction with a standard National Instruments IMAQ-PCI-1407, single-channel, 8-bit monochrome image-acquisition board. The board was accessed through a National Instruments NI-IMAQ 2.1 application program interface, which handled all of the necessary timing, gain control, and memory transfers. The PC's internal serial port was used as the RS-232 interface for controlling the S-VHS JVC SR-3888E video recorder, when analysis was performed from video tapes. High-resolution, longitudinal B-mode ultrasound images of the brachial artery, along with Doppler ν and electrocardiogram traces, were simultaneously acquired with an Acuson Aspen ultrasound Doppler machine (Mountain View, CA) and 12- to 15-MHz linear array probe.
BF Analysis Software
Either ultrasound studies were recorded on an S-VHS tape within the ultrasound machine and then played back on a separate S-VHS video recorder for analysis, or, alternatively, the video signal was taken directly from the ultrasound machine and, with the IMAQ-PCI-1407 card, was directly encoded and stored as a digital DICOM file on the PC. Subsequent software analysis of this data, at ∼20–30 frames/s, was performed by using an icon-based graphical programming language (LabVIEW 6.02, National Instruments, Austin, TX) and toolkit (IMAQ, National Instruments), which are used to build software programs called virtual instruments (VIs).
To perform the analysis, the operator selects four regions of interest (ROIs) on the B-mode images, which are accessed from the video or stored DICOM files and displayed on the PC (Fig.1).
The “calibrate diameter” ROI allows the observer to convert the image size on the computer, measured in pixels, to the actualD of the artery in millimeters. This is performed by drawing a region on the screen between two marks on the ultrasound image that are a known distance apart (typically 1 cm).
Similarly, the “calibrate Doppler” ROI allows the observer to calibrate the Doppler ν by drawing a region that encompasses the calibration markers on the ν axis and nominating the appropriate scaling factor (typically between 1 and 3 ms).
The “diameter” ROI is used by the operator to select the most stable portion of the arterial B-mode display for automatedD analysis. A sophisticated IMAQ parallel-prong, rake VI algorithm is used to perform parallel line detection to determine theD of the artery within each 256 gray-scale frame. Each of the detected parallel lines, representing the near and far arterial walls in each frame, is interpreted with a quadratic spline technique. As a result, within a typical ROI, 200–400 parallel points are subtracted, and the median score is calculated as the representativeD for that frame. The software then plots the parallel lines from which overall internal vessel D has been calculated, to provide visual feedback to the observer, who can verify that the system is accurately tracking the arterial walls during frame processing.
Finally, a rectangular “Doppler” ROI is drawn around the Doppler waveform strip. Within this region, an IMAQ automatic thresholding VI is used to filter the gray-scale image. Each column of pixels within the ROI is then analyzed to detect the first white pixel in the vertical array, thereby detecting the waveform envelope. The detected points are then plotted to provide visual feedback regarding the precision of the envelope detection.
Data collected from all ROIs are used in subsequent acquisition of S-VHS or DICOM image files from which synchronized D and ν measures are stored for each analyzed frame, at 20–30 Hz.
Display of results.
Once the study has been acquired, a data “display” VI plots a graph of the arterial D and ν against time (Fig.2 A). In addition, these synchronized ν and D measurements are used to calculate and display BF as a continuous plot across the cardiac cycle (Fig.2 A).
Operator-controlled cursors can then be used to select and zoom in on sections of the data set that are of interest (e.g., exercise epochs), and clearly erroneous data points may be manually removed by the observer, or a smoothing algorithm applied, if required. Finally, data displayed between the cursors are analyzed and presented in a number of formats (Fig. 2). “Mean” BF [mean forearm BF (MFBF)], ν, andD are calculated as the algebraic mean of all data points between the cursors, which may be placed on either side of a discrete cardiac cycle or at the beginning and end of an array of such cycles. Note that these values incorporate those during both systole and diastole, so that MFBF is influenced by the magnitude of retrograde, diastolic flow (as described in Antegrade and Retrograde Brachial Arterial Flows During Incremental Cycle Exercise). Area under the curve data for BF, D, and ν are calculated as the time integral of each trace. A “peak” (systolic) BF [peak forearm BF (PFBF)], D, and ν detection VI is used to identify and display the maximum data point within each cardiac cycle and to subsequently calculate the average of these peaks. Finally, the area under the curve of all positive and all negative BF data points that lie between the cursors is presented to provide a global index of the volume of antegrade and retrograde BF per minute (Fig. 2).
BF Assessment in Vitro
Comparison of Doppler ultrasound flow to flow through a phantom artery.
A phantom BF system was constructed, employing a Perspex tube of known internal D (2.1 mm) suspended in a water bath. The ultrasound probe was placed in the water bath, and an image of the Perspex tube was obtained in B mode. A saline-cornstarch suspension was passed through the tube to mimic blood-derived Doppler signals, with the flow of this solution controlled via its connection to a hemodialysis pump, the rate of which was varied through 50, 60, 70, 80, and 90 cycles/s. Five measures of flow through the tube were collected over a 1-min period at each cycle rate, with the amplitude of flow on each occasion varied by altering flow through a parallel circuit. After passage through the tube, the solution volume was collected and directly measured. These volumes were compared with the net area under the curve (antegrade-retrograde area) measures of flow derived from the Doppler ultrasound system, which simultaneously recorded flow over each minute of collection.
BF Assessment in Vivo
Reproducibility of BF assessment and comparison to plethysmography.
To evaluate the day-to-day reproducibility of measurements on individuals with the Doppler ultrasound analysis system described above, we undertook repeat studies, performed 3 days apart at the same time of day, on 6 young (29 ± 2 yr), healthy adults. Each experimental session consisted of BF assessment at rest and during reactive hyperemia (RH) induced by a 5-min period of forearm ischemia. To compare techniques of BF measurement, assessments were simultaneously undertaken with the Doppler ultrasound methodology and strain-gauge plethysmography. Further methodological details of the latter technique are available elsewhere (1, 14, 15).
Treatment and analysis of data.
Between-day variability of BF measures derived with the Doppler ultrasound technique was determined by performing a Student's pairedt-test for determination of statistical difference and by calculating a coefficient of variation between the paired data sets after initial calculation of the technical error of measurement, according to the method of Kahn and Sempos (10). Comparisons between the Doppler ultrasound and plethysmography techniques were undertaken by performing Student's pairedt-tests and correlation coefficients on both absolute forearm BF (FBF) data and percent changes from resting baseline.
Effect of lower limb exercise on upper limb BF.
The further purpose of the study was to describe the changes in MFBF and PFBF to the resting forearm during lower limb exercise and the magnitude of antegrade and retrograde flow in the brachial artery across the cardiac cycle as exercise intensity increased. This study also served to develop reference data for a separate experiment regarding the role of nitric oxide in vascular function (D. J. Green, C. Cheetham, L. Mavaddat, K. Watts, R. R. Taylor, and G. O'Driscoll, unpublished observations) and for future studies.
Subjects and screening measures.
The eight subjects were healthy men with no evidence or history of vascular disease. Those enrolled had the following characteristics: age, 22.4 ± 4.4 (SE) yr; height, 178.8 ± 4.8 cm; weight, 87.8 ± 10.2 kg; resting heart rate, 71 ± 7 beats/min; systolic blood pressure, 125 ± 9 mmHg; and diastolic blood pressure, 83 ± 7 mmHg. The study procedures were approved by the Ethics Committee of Royal Perth Hospital, and all subjects gave prior written consent.
Subjects received an information sheet instructing them to abstain from food within 4 h of testing and alcohol and/or caffeine within 12 h of testing. Investigations were conducted in a quiet, temperature-controlled laboratory. A 20-gauge arterial cannula (Arrow, Reading, PA) was introduced into the brachial artery of the nondominant arm under local anesthesia with <2 ml of 1% lidocaine (Astra Pharmaceuticals) to transduce pressure, which was monitored continuously (Transpac, Abbot Laboratories) throughout the study.
After cannulation, saline was infused to maintain patency throughout a 30-min stabilization period during which subjects were seated quietly on an electronically braked bicycle ergometer (Orival 400, Lode). After this, a 2-min baseline recording of brachial arterial D and ν was undertaken with the Acuson Aspen ultrasound-Doppler machine. This was followed by a 15-min lower limb cycling exercise protocol consisting of five 3-min incremental epochs (40, 60, 80, 100, and 160 W). Brachial arterial D and ν were continuously recorded throughout the exercise protocol. Subjects had their dominant arm passively suspended, with no contact with the handle bars taking place whatsoever, while the nondominant arm was also passively supported and immobilized while Doppler ultrasound measurements were taken.
Experimental measurements: FBF assessment.
A 12- to 15-MHz multifrequency linear array probe attached to the Aspen was used to visualize the artery in the distal one-third of the upper arm. Ultrasonic parameters were set to optimize longitudinal, B-mode images of the luminal-arterial wall interface. Once set, these parameters remained constant throughout the session, and the probe was held in a constant position with a stereotactic clamp. Continuous Doppler ν assessment was also displayed by the Aspen and was collected by using the lowest possible insonation angle (always <60°), which did not vary during each study. The ν waveform was automatically corrected for the insonation angle.
BF, calculated as the product of CSA and ν, was derived from synchronized B-mode ultrasonography and Doppler ν measures, by using the suite of software packages (VIs) described in detail above inBlood Flow Analysis Software. CSA was calculated from the software-derived arterial D measures with the equation CSA = π · radius2. Continuous measures of ν, D, and BF were plotted throughout the cardiac cycle for the baseline and each of the exercise intensities (Fig. 2). PFBF and MFBF measures for the baseline and exercise workloads were calculated across the final 20 s of each period.
Treatment and analysis of data.
Results are expressed as means ± SE. Statistical changes in peak BF, mean BF, and antegrade and retrograde flows between baseline and exercise workloads were determined by using one-way ANOVA (SPSS). Post hoc t-tests were used to determine differences between each workload and baseline data. P < 0.05 was considered significant.
Typical BF, ν, and arterial D traces at baseline and during exercise stages are presented in Fig. 2. Satisfactory ultrasound and Doppler images were obtained for all subjects and at all workloads, including 160 W.
Comparison of Doppler Ultrasound Flow to Flow Through a Phantom Artery
The means (±SE) of actual fluid measurement and Doppler ultrasound flow assessment were similar in absolute terms: 25 ± 1 vs. 29 ± 1 ml (P = not significant), respectively, for 1-min collections. These data were significantly correlated (r = 0.98, P < 0.005).
Reproducibility of BF Assessment and Comparison to Plethysmography
Paired absolute blood BF data, averaged over the final 30 s (steady state) of the rest period and between 45 and 75 s postcuff deflation for the RH measure, taken on each day, did not significantly differ when compared by t-test. In addition, high correlations were evident between rest data collected on each day (r = 0.91, P < 0.01) and RH collected on each day (r = 0.94, P < 0.01). When rest and RH data were pooled, the coefficient of variation between days was 24.3%, a value that is similar to that previously reported for Doppler ultrasound reproducibility (22).
When simultaneously collected Doppler ultrasound data were compared with plethysmographic assessment, the magnitude of changes in flow from baseline measured with the use of each technique was not significantly different (P = 0.2) and was highly correlated (294 ± 34 vs. 206 ± 15%, r = 0.83, P < 0.01). Absolute resting baseline flows were also comparable: 40.3 ± 5.1 vs. 30.6 ± 2.9 ml/min for plethysmography and Doppler ultrasound, respectively. These data were not significantly different by paired t-test (P = 0.2) and were highly correlated (r= 0.97, P < 0.01). Plethysmographic and Doppler ultrasound data after RH were also highly correlated (r= 0.83, P < 0.01), although, in absolute terms, they were significantly different (99.6 ± 7.9 vs. 58.0 ± 5.4 ml/min, respectively; P < 0.05). Differences in absolute flows are perhaps not surprising, as plethysmographic measures are principally used to provide information on relative changes in flow from baseline (1).
Changes in Resting MFBF and PFBF During Incremental Cycle Exercise
ANOVA revealed a significant effect of lower limb exercise workloads on PFBF (P < 0.001, Fig.3). When individual workloads were compared with baseline data, PFBF significantly increased at 40 W (P < 0.05), 60 W (P < 0.05), 80 W (P < 0.001), 100 W (P < 0.001), and 160 W (P < 0.001). MFBF responses were also significantly affected by lower limb exercise (P < 0.001, one-way ANOVA; Fig. 4). Post hoc analysis revealed that MFBF was lower than baseline at 40 W (P < 0.01), similar at 60 W (P = 0.07) and 80 W (P = 0.8), and significantly higher than baseline at 100 W (P < 0.01) and 160 W (P < 0.001). These patterns of change in PFBF and MFBF were due primarily to changes in peak and mean ν of BF; arterialD did not significantly change relative to baseline at any workload (Fig. 5). In turn, the difference in patterns of PFBF and MFBF depended partly on changes in retrograde (diastolic) flow as described below.
Antegrade and Retrograde Brachial Artery Flows During Incremental Cycle Exercise
Figure 6 presents the area under the curve data time integral for antegrade and retrograde flows at baseline and during each exercise intensity. ANOVA revealed significant effects of exercise workload on both antegrade (P < 0.001) and retrograde (P < 0.01) flows. Relative to baseline, the magnitude of antegrade flow during each cardiac cycle increased significantly at 60 W (P< 0.05) and at all workloads between 80 and 160 W (P< 0.01). The magnitude of retrograde flow increased at each workload compared with baseline (all P < 0.01). This pattern of increase in antegrade and retrograde flows was present in all eight subjects.
The present study utilized a novel software analysis system to assess conduit arterial D and blood ν and to calculate BF across the cardiac cycle from this synchronized data. We have demonstrated for the first time in vivo, by using a combination of high-resolution B-mode ultrasonography and Doppler ν assessment, that BF can be measured noninvasively and continuously, with high temporal resolution in humans.
A major aim of the present study was to describe changes that occur in FBF during lower limb exercise. Our laboratory (14,15) and others (11-13) have recently reported improvements in upper limb vascular function during lower limb exercise training programs, which largely excluded physical conditioning of the forearm musculature. These studies suggested that exercise training may exert a generalized conditioning effect on the vasculature, and we have speculated that this may be due, in part, to the long-term impact of repetitive bouts of physical activity that, via hemodynamic modulation, influence shear stress on the vascular wall. Although previous studies have measured upper limb BF during lower limb exercise (2, 3, 8,16), these have utilized plethysmography, a technique that yields relative changes in flow averaged over several seconds, and does not provide data regarding changes in arterial ν, CSA, or BF within the cardiac cycle. Using the technology derived for this study, we were able to automatically plot changes in the BF trace, subsequently detect peak flow within each cardiac cycle, and average these, or calculate mean flow across each cycle. It is important to note that the frequency of data acquisition, at 20–30 Hz, is presently limited by the processing capacity of the computer, which can reasonably be expected to improve in the future.
We observed an increase in PFBF to the resting forearm as cycle ergometer exercise intensity increased. Somewhat to our surprise, given these PFBF findings, the pattern of change in MFBF was biphasic, demonstrating an initial decline relative to baseline, followed by a significant increase at higher workloads. The explanation for this became clear on closer examination of the BF traces (see Fig. 2). At rest, blood ν and BF in all subjects were positive during systole, falling to near zero during diastole. With increasing intensities of exercise, systolic or peak flow became progressively more positive, particularly at the higher workloads. However, in contrast to the situation at rest during diastole, negative ν, indicative of retrograde or reverse BF through the brachial artery, were observed during all intensities of exercise. Whereas the magnitude of negative diastolic flow increased with exercise intensity, its impact on MFBF was greatest at lower workloads when positive flows were modest. Hence, a biphasic MFBF response was observed.
This is the first study to quantify the magnitude of antegrade and retrograde flow through a conduit artery across the cardiac cycle in humans. Whereas a recent study observed retrograde flows during exercise, these were in the femoral artery during leg exercise and were attributed to either pulse-wave reflection or the impedance associated with rhythmic muscle contraction and consequent vessel occlusion (5). The latter cannot be the case in the present study because the forearm was at rest. This raises the possibility that, during exercise, retrograde flow in the brachial artery, and possibly other systemic vascular beds, may be due to wave reflection. An intriguing alternate hypothesis is that retrograde BF in the brachial artery may be due to a “steal” phenomenon during diastole associated with lower limb exercise in the upright posture. That is, during phasic lower limb skeletal muscle relaxation, a suction effect may occur as arterial vessels tethered to surrounding muscular connective tissue are forcibly opened. Such an effect, with the added impact of hydrostatic pressure in the arterial column, could markedly improve perfusion of the active muscle while “stealing” BF from the inactive upper limbs. These hypotheses warrant further investigation.
Regardless of the precise physiological explanation for the appearance of retrograde brachial arterial flow during lower limb exercise, this phenomenon is likely associated with greater shear stress on the vessel wall, compared with continuous laminar or pulsatile antegrade patterns of flow (7). The impact of phasic antegrade-retrograde flow on vessel wall mechanosensors, regulation of wall stress, and release of vasoactive paracrine agents such as nitric oxide, prostanoids, or endothelin, is yet to be determined.
In summary, we present a novel, noninvasive technique that, for the first time, allows assessment of BF across the cardiac cycle in humans, with high temporal resolution. We have observed differences between the patterns of change in resting upper limb peak and mean BFs across the cardiac cycle during lower limb exercise because of the impact of retrograde diastolic BF that occurs during low- and moderate-intensity cycle ergometer exercise. The systemic physiological role and significance of the observed retrograde flows await future investigation.
This study was supported by the National Heart Foundation (Australia) and Medical Research Fund of Western Australia.
Address for reprint requests and other correspondence: D. Green, The Dept. of Human Movement & Exercise Science, The Univ. of Western Australia, Parkway Entrance No. 3, 35 Stirling Highway, Crawley, WA 6009, Australia.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2002 the American Physiological Society