Application of image analysis for evaluation of red blood cell dynamics in capillaries

MICROVASCULAR

RESEARCH

Application CHRISTOPHER

d‘t,

214-225 (1992)

of Image Analysis for Evaluation of Red Blood Cell Dynamics in Capillaries G. ELLIS,* MARY L. ELLSWORTH,? ROLAND N. PITTMAN,* AND WAYNE L. BURGESS*

*Department of Medical Biophysics, University of Western Ontario, and The Microcirculation Laboratory, John P. Robarts Research Institute, London, Ontario, N6A Xl Canada; TDepartment of Pharmacological and Physiological Science, St. Louis University School of Medicine, St. Louis, Missouri, 63104; and #Department of Physiology, Medical College of Virginia, Virginia Commonwealth University, Richmond, Virginia, 23298-0551 Received March 23. 1992

We have devised a method to display and directly evaluate red blood cell (rbc) dynamics in capillaries using the same dual camera intravital video microscopy system employed to determine rbc oxygen saturation (Ellis et al., 1990). Capillary images are recorded on videotape and an interactive graphics system is used for analysis. Data are sampled once a frame for 60 set using a window (one pixel wide (0.93 pm) and 100 pixels high) positioned along the axis of a capillary. The resulting data are displayed as sequential space-time images 100 pixels high by 300 pixels wide (10 set). The space-time images thus created represent the dynamics of the rbc’s in a single comprehensive static image in which the rbc’s appear as dark, diagonal bands separated by light bands representing plasma gaps. From these images one can obtain information on velocity of individual rbc’s (pm/set), lineal density of rbc’s (rbc/mm), and rbc supply rate (rbc/sec). This information can be used to delineate the temporal and spatial heterogeneity of hemodynamics in capillary networks. These data can then be combined with coincident data on red blood cell oxygenation to provide a complete picture of oxygen transport in capillaries or it can be used atone as a tool for the evaluation of basic in vivo and in vitro rheological questions. o tsar Academic PEW, Inc.

INTRODUCTION The convective transport of oxygen through the microvasculature is determined primarily by the movement of red blood cells (rbc) and by their oxygen content. Based on theoretical considerations (Popel, 1989), one would expect that the transport of oxygen from the red blood cells to the surrounding tissue would be affected by the presence of significant heterogeneity in any or all of the primary convective transport parameters: red blood cell velocity (pm/set), lineal density (rbc/mm), and supply rate (rbc/sec). The presence of heterogeneity within an individual capillary should affect the uniformity of oxygen delivery to the surrounding tissue over time. Hemodynamic differences among capillaries should impact on the total amount of oxygen delivered even if the red blood cell dynamics within each capillary were temporally uniform. 214 Copyright Q l!W2 by Academic Press, Inc. All rights of reproduction in any form reserved. Printed in U.S.A.

IMAGE

ANALYSIS

OF RED

BLOOD

CELL

DYNAMICS

215

Despite studies reporting heterogeneity in a number of these parameters within individual capillaries and among groups of capillaries (Ellis et al., 1990; Ellsworth et al., 1988; Honig et al., 1977; Tyml et al., 1981), the impact of these factors on oxygen delivery to tissue in viva has not been fully tested, What is required for a detailed analysis is a reliable means of obtaining simultaneous data on both the dynamics and the oxygen saturation of individual red blood cells as they traverse a capillary. Several techniques currently exist for measuring these parameters individually using either manual frame-by-frame analysis of videotapes (Ellis et al., 1987; Sarelius, 1986) or automated techniques (Ellis et al., 1984, 1986; Intaglietta et al., 1989, 1990; Tyml and Sherebrin, 1980). Although frame-by-frame analysis can provide accurate, detailed data on capillary blood flow, the analysis of large amounts of such data is impractical. Furthermore, none of the automated techniques enables one to easily verify the quantitative results. We have developed a method for obtaining simultaneous data on red blood cell dynamics and oxygen saturation in a capillary using an enhanced version of the image analysis approach originally developed for the measurement of rbc oxygen saturation in capillaries in vivo (Ellis et al., 1990). This approach enables us to represent the dynamics of red blood cells in a comprehensive static image which can be readily analyzed to provide quantitative information on red blood cell velocity, lineal density, and supply rate. For example, the velocity of individual red blood cells can be followed along the length of a capillary or supply rate can be measured at a particular point in a capillary over an extended period of time. This information can then be combined with oxygen saturation measurements for each red blood cell in the same capillary to obtain a complete description of convective oxygen transport. METHODS Video images of red blood cells passing through capillaries of the resting hamster retractor muscle were recorded on videotape using the dual camera video microscopy system described previously for the measurement of oxygen saturation in red blood cells perfusing capillaries (Ellis et al., 1990). The video recordings obtained using the isosbestic wavelength (420 nm) were subsequently analyzed off-line for the corresponding hemodynamics using an image analysis system consisting of a video analyzer (Colorado Video Analyzer, Model 321), a programmable video field counter (University of Western Ontario), and a microcomputer (Compaq Deskpro 386/25) equipped with a 12-bit analog to digital convertor (A/D board, Analog Devices Model RTI-815-A), a graphics board (AT1 VGA Wonder graphics card with 512K memory, 640 x 480 fixed resolution with 256 colors), and a high-resolution computer monitor (NEC Multisync II). The “slowscan” output from the video analyzer is proportional to the light intensity along the vertical cursor line at its intersection with the horizontal television (TV) lines. The programmable video field counter extracts video synchronization information (at the beginning of each video field and horizontal TV line) from the composite video input and uses these pulses to provide timing information to the computer for digitizing the slow-scan signal (Fig. 1). Light intensity information along a region of capillary 100 TV lines long is collected once a frame (30-Hz framing rate) for 60 set providing a total of 1800

216

ELLIS ET AL.

computer

video monitor

monitor

microcamputer

vi&b caeeette

1

recorder

video anafyzer

I

1

programmable video field eoanter

FIG. 1. Schematic diagram of the equipment setup used in the analysis of the video images. (See text for details.)

light intensity va111es at each spatial ~ve~~~~ location (Fig. 2). These data obtained are then redisplayed on the Hugh-resolution computer monitor as a single graphic space-time image with sample location along the vertical axis and time along the horizontal axis using 64-grey scales to represent the light intensity levels. The dark diagonal bands on the image represent the red blood cells as they traverse the video monitor

computer

mowor

reconstructed image

z

0 time

kscondsj

Fro. 2. Schematic image of the generation of a short time segment of a space-time image. (See text for details.)

IMAGE ANALYSIS

OF RED BLOOD CELL DYNAMICS

217

93 pm Velocity (pmlsec)

Lineal

Z

Density

(rtxd mm)

Supply Raie Mdsec)

6 rbdsee 0

0.5 time

1.0

1.5

0

(sex)

FIG. 3. Schematics of the derivation of convective parameters: (top) red blood cell velocity; (middle) lineal density; and (bottom) red blood cell supply rate. The instantaneous velocity of a single red blood cell is the slope of the image of the red cell (i.e., dark band) in the space-time image. The lineal density is the number of red blood cells lying along a vertical line positioned at a specific time in the image expressed in units of the number of red cells per mm. The red blood cell supply rate is the number of red cells crossing a specific horizontal location in the image during a 1-set time interval.

sampling window. The adjacent light bands represent the plasma gaps between cells. The slope of these bands depends on the direction of flow on the video monitor (positive slope, flow up; negative slope, flow down). The space-time image can be thought of as a composite picture composed of 1800 images, each image 1 pixel wide by 100 pixels high, which displays in a single static image the dynamics of capillary red blood cell flow for a 1-min time segment. At the magnification used, one TV line represents a distance of 0.93 pm, hence the sample window corresponds to a 93-pm-long segment of the capillary. All software has been written in C (Microsoft C, Version 5.1) and Assembler (Microsoft Macro Assembler, Version 4.0). From these images all of the hemodynamic parameters can be determined by applying the basic definitions of each parameter to the space-time image as described in the legend to Fig. 3. The key aspect of the analysis of the flow of individual red blood cells through a capillary is the ability to distinguish a red blood cell from the plasma that surrounds it. Variations in the background light intensity along the length of a capillary (a significant problem in striated muscle preparations) can make it difficult to define the red blood cell-plasma interface. Our approach has been to transform light intensity data into optical density values and then apply an optical density threshold which defines the red blood cell-

218

ELLIS

ET AL.

plasma interface as described previously (Ellis et al., 1990). All optical density values below the cutoff level (plasma) are set to zero. This strict definition of the red blood cell-plasma interface overcomes the subjective decision one would have to make with light intensity images. Visual tracking of cells through the window is relatively straightforward based on the continuity of the red blood cell band from frame to frame. However, the automated tracking of cells by the computer is difficult due to the unique properties of red blood cell dynamics in capillaries, e.g., the merging of individual cells into a train of abutting cells or the reverse, the separation of red blood cell trains into single cells. Automated cell tracking requires the development of sophisticated algorithms specific for the range of red blood cell dynamics found in capillaries. We determined red blood cell velocity from measurements of the “instantaneous” slope (based on approximately three frames of data) of each dark band at 1-set intervals using standard morphometric image analysis software (Java, Version 1.31; Jandel Scientific). Red blood cell supply rate was determined by visually counting the number of red blood cells crossing the vertical midpoint of the image over each 1-set interval. An optical density profile (plot of optical density vs time) at that location was available to assist the user. Lineal density was measured once a second by automated analysis of the optical density profile (optical density vs vertical position) along the length of the capillary. A unimodal axial profile of optical density values extending for more than 5 pm was assumed to be a single red blood cell. Bimodal profiles were considered to be either two overlapping or abutting cells and multimodal profiles were treated as multiple cells. Clearly, any capillary studied must be stationary (e.g., no respiratory movement) and it must be in focus over the entire sampling region during the sampling time interval. Since the depth of focus for the objective we used (Zeiss UD40) is 1.4 pm (Slayter, 1970), capillaries which are in good focus all lie in the same plane. As was true for the analysis of oxygen saturation (Ellis et al., 1990), the image of a capillary to be analyzed must be aligned on the video monitor within 45” of vertical. This requirement is imparted by the video analyzer whose moveable vertical cursor pivots about a point at the top of the screen with a maximum 45” angular excursion on either side of vertical. APPLICATION:

ANALYSIS OF A SINGLE SPACE-TIME

IMAGE

Presented in Fig. 4 is a single representative space-time image of a capillary segment 93 pm long, observed for 60 set and displayed in three strips, each 20 set long. The utility of the image analysis approach for the evaluation of capillary hemodynamics will be demonstrated based on the analysis of this single image. Clearly, another space-time image would lead to different values for the various parameters but the analytical approach would be identical. The dark bands shown in Fig. 4 represent individual red blood cells that are moving up the monitor as indicated by their positive slopes. The horizontal light and dark bands are indicative of variations in the background light intensity. The first impression one gets from this image is that of marked variability in the horizontal distance between the dark bands and in their slopes throughout the 60-set sampling interval. One also notices regions of the image where groups of

IMAGE ANALYSIS

OF RED BLOOD CELL DYNAMICS

219

time (seconds)

FIG. 4. Space-time image from a single capillary showing the movement of red blood cells through the sample window over a 60-set time period displayed as three, 20-set segments.

cells are moving through the capillary together and other regions where the cells are travelling as discrete entities. More detailed examination of the image shows that at time zero the red blood cell velocities are low, as indicated by the shallow slopes. Velocities then increase dramatically for the next 1.5 set, followed by a significant decrease in velocity during the next 4 sec. The velocities then increase and remain fairly constant except for short periods of reduced velocity beginning at approximately 8.5, 28, 39, 55, and 58 sec. One brief period of flow reversal occurs at approximately 24.3 sec. The horizontal thickness of the dark bands is also highly variable; the bands become very thick when the cells slow down as occurs between 4 and 6 set, and very thin when the velocity increases as occurs between 6 and 7 sec. The horizontal thickness of a dark band is a measure of how long it takes a cell to pass a specific location in the capillary, i.e., the cell’s residence time at that position. Another interesting feature of this image is the merging of the two bands which originate at 3.5 and 4 sec. These two cells entered the window separated by approximately 5 pm but within 1.2 set the gap between the cells was closed. This indicates that the trailing cell was travelling slightly faster than the leading cell. The results of the quantitative analysis of Fig. 4 are shown in Fig. 5. Figure 5a shows the velocity of each red blood cell in the sample window evaluated every second for 60 sec. The most striking feature of this graph is that the velocity shows considerable variability with time, ranging from 11 to 304 pm/set (mean 95 & 45 (SD) pm/set). In addition, the velocities of individual cells in the capillary at the same instant of time are not the same. In Fig. 4 the temporal variability is readily apparent but the cell-to-cell variability is less evident. The number of

220

ELLIS ET AL.

a

“T-,

0

IO

.

.

.

.

20

30

40

50

time

.-

60

(SW)

FIG. 5. Results of the analysis of the space-time image in Fig. 4 for velocity, lineal density, and supply rate presented as time series records (a, b, and c, respectively).

data points at each time increment in Fig. 5a indicates the number of cells in the sample window at that time. These data, which represent the lineal density of rbc’s, are shown in Fig. 5b expressed as the number of rbc’s per millimeter. The lineal density was quite variable, ranging from 11 to 85 rbc/mm (mean 62 2 17 (SD) rbc/mm). The supply rate was evaluated every second as indicated in Fig. 3 by counting the number of rbc’s passing the midpoint of the window during a 1-set time interval. The red blood cell supply rate, shown in Fig. 5c, ranges from 1 to 9 rbc/sec (mean 5 k 2 (SD) rbc/sec). All three parameters show the greatest variability during the first 25 sec. DISCUSSION Figure 4 was chosen for presentation here because it is representative of the quality of image we routinely acquire with the video system and clearly demonstrates a number of the characteristics of red blood cell dynamics in capillaries. The use of a 420-nm interference filter ensures that high-contrast space-time images will be generated. In excess of 500 of these images have been analyzed for oxygen saturation and hemodynamics to date, with each image providing different values for the various parameters. The maximum velocity that can be theoretically measured by this technique is the maximum observable red blood cell displacement in one frame: sample window length (93 pm) minus twice the red blood cell length (approximately 9 pm for

IMAGE

ANALYSIS

OF RED

BLOOD

CELL

DYNAMICS

221

lime Iseconds)

FIG. 6. Space-time images obtained by automated sampling of light intensity data (top) and by manual frame-by-frame analysis {bottom) of the video recording corresponding to the first 10 set of the space-time image shown in Fig. 4. The space-time image in (a) was generated using the method described here. The space-time image in (b) is a plot of cell location (midpoint of each cell with error bar corresponding to cell length) versus time. (See text for details.)

hamster red blood cells in 4- to 5-pm-diameter capillaries) divided by the sample time interval (l/30 set). For the data presented in Fig. 6, this yields an ideal maximum velocity of 2250 pm/set [(93 Frn - 2$9 pm)) + (l/30) set]. However, this assumes that one can identify the red blood cell from one frame to the next over a very large displacement (approximately eight times the cell length) in the presence of other red blood cells. A practical limit on the maximum cell displacement for a single cell that can be easily followed in the space-time image is a displacement equal to the cell length, giving on average a maximum velocity of 270 gm/sec [9 pm/(1/30 set)]. This would include the normal velocities found in capillaries of resting striated muscle in hamster (94 + 48 (SD) pm/set; Ellsworth et al., 1988), rat (105 + 65 (SD) pm/set; Ellis et al., 1990), and frog (157 + 57 (SD) pm/set; Ellis ec al., 1990). There is no ~i~ifu~~o~ on the dete~ination of velocity imposed by zero or reverse flows. For lineal density, the upper limit is determined by the number of cells that can be detected in the sample window. The ma~mum lineal density with no overlap between cells can be calculated from the sample window length divided by the mean length of the cells in that capillary. For Fig. 4, this would be 10 red blood cells or 107 rbcfmm. If the red blood cells are overlapped to the extent that there are no minima in the optical density waveform, then the method cannot determine the true lineal density. However, one can set an upper limit on the maximum lineal density achievable in a capillary of given diameter with red blood cells of given volume and surface area. The maximum lineal density measured for the sample data set was 8 rbc/93 ym (85 rbc/mm) indicating, on average, a slight gap between the cells. The normal lineal density measured in hamster

222

ELLIS

ET AL.

retractor muscle was 63.2 ? 29.9 (SD) rbc/mm (Ellsworth et aE., 1988), in rat gracilis it was 61.4 + 22.3 (SD) rbc/mm (Ellis et al., 1990), and in frog sartorius it was 18.6 + 7.8 (SD) rbc/mm (Ellis et al., 1990). Theoretically, the maximum number of cells that can be detected passing a point in the capillary per second is one red blood cell per frame or 30 rbc/sec. However, scanning along a single horizontal line in the space-time image means moving from one frame to the next. This makes it impossible to determine whether one has detected the same cell or a new cell in the next frame without further information. By carefully examining the space-time image one can follow the progress of each cell through the capillary and thus use this information to detect supply rates approaching 30 rbc/sec. An alternate method of determining the red blood cell supply rate (rbc/sec) is to compute the product of red blood cell velocity (mm/set) and lineal density (rbc/mm). Based on our conservative estimates of the maximum measurable velocity (0.270 mm/set) and lineal density (107 rbc/mm) given above, the maximum computed supply rate would be 29 rbc/sec. Comparison of the visually measured supply rate with the computed supply rate is a simple means of detecting measurement errors in any of the three hemodynamic variables. The normal supply rate measured in hamster retractor was 6.7 2 4.6 (SD) rbc/sec (Ellsworth et al., 1988), in rat gracilis it was 6.2 + 4.2 (SD) rbc/sec (Ellis et al., 1990) and in frog sartorius it was 3.2 ? 2.1 (SD) rbc/sec (Ellis et al., 1990). Frame-by-frame analysis of high-speed tine film or video recordings is a very accurate method for measuring the dynamics of red blood cell flow through a capillary and this method generally has been considered the standard with which all other image analysis approaches should be compared. Manual analysis of the space-time image may be regarded as equivalent to manual frame-by-frame analysis if the frame-by-frame analysis is based on visual information from along the axis of a straight capillary segment. Under these conditions both methods are relying on the same light intensity information in identifying the location of each individual red blood cell. The video recording corresponding to the first 10 set of the space-time image shown in Fig. 4 was analyzed using a manual frame-by-frame approach (Ellis et al., 1987). Using an interactive computer graphics system and visual interpretation of still-frame video images, the user selected the location of both ends of each red blood cell in every frame for a total of 300 frames. These data were then used to determine red blood cell velocity and lineal density for comparison with the corresponding space-time image data. There was no statistical difference (twotailed t test with paired data; P < 0.001) between the measurements of red blood cell velocity (difference = 4.7 + 3.2 pm/set SE) or lineal density (difference = 1.0 + 2.2 rbc/mm SE) obtained with these two techniques. The data on cell location obtained from the frame-by-frame analysis were used to generate the space-time image shown in Fig. 6b using Sigmaplot (Version 4.1; Jandel Scientific). The midpoint of each cell was plotted as a function of time (horizontal axis) and position (vertical axis) with cell length being displayed by error bars. The line width for the error bars was chosen to give the appearance of a continuous band for each cell in the image. For comparison, the matching space-time image (first 10 set of Fig. 4) is shown in the top panel. As expected, the two space-time images shown in Fig. 6 show striking similarities despite the fact that they have been generated in two completely different ways. This com-

IMAGE ANALYSIS

OF RED BLOOD CELL DYNAMICS

223

parison has been made to highlight the fact that the space-time image obtained by sampling light intensity data during real time playback of the videotape contains information equivalent to that obtained with manual frame-by-frame analysis of the same video segment. There are several advantages to our approach of storing the sampled light intensity data for later analysis and display as a graphic image. First, one has immediate visual confirmation of the quality of the light intensity sample by examining the space-time image. Any problems, such as loss of focus or tissue movement during the sample interval, which may be overlooked during normal playback of the videotape, are easily recognized in the space-time image. Careful examination of the images in Figs. 4 and 6 show that the spatial variation in light intensity along the window, not associated with the red blood cells, remains fixed, i.e., the horizontal dark and light bands are straight. Respiratory motion, for example, would produce a characteristic oscillation in the background image. Second, the validity of any particular measurement can be easily verified by visually comparing the analysis results with the static image. For example, the velocity record in Fig. 5a shows that the velocity was very low, approximately 25 pm/set, during the time interval from 4 to 9 set except for a sudden increase at 8 set to nearly 150 pm/set. By direct comparison with the space-time image, one can easily verify the low velocity during this time interval and the very abrupt increase in velocity between 7.1 and 8.5 sec. Because our method is designed to analyze moving red blood cells, there are two limitations imposed on it by our use of continuous illumination with SIT cameras. The image tube integrates optical information (charge image) during the time between scans. Thus a moving object under continuous illumination will appear elongated because its image will be a history of its position during the previous l/30 set (assuming no image persistence), the amount of elongation being equal to [velocity (pm/set) x 0.033 set] pm. The use of strobed or shuttered illumination will eliminate this problem. In the typical video camera tube, the charge image is raster-scanned by an electron beam, with any one point being scanned every l/30 sec. Ideally, this electron beam should completely discharge the target and erase the charge image; however, this is not usually the case, leading to lag or image persistence. There are two types of image persistence: decay lag (bright to dark transition) and buildup lag (dark to bright transition) (Inoue, 1986). Because we observe the movement of a train of dark cells separated by bright plasma gaps, both decay and buildup lag are present. This makes it difficult to accurately define both the leading and the trailing edges of the cells. However, we have used the same optical density criteria to define the leading and trailing edges of a cell and have used these two endpoints to determine the midpoint of the cell. The error in estimating the velocity is minimized by using the slope of the dark band based on the midpoint of a red blood cell in three successive frames. We note that our velocity estimates from the space-time image are not statistically different from those of the frameby-frame approach. Charge coupled device (CCD) cameras offer the best performance in terms of virtually no image persistence and, thus, their use should eliminate the problem. A SIT camera was used in this study because our video spectrophotometric system for measuring hemoglobin oxygen saturation requires extremely low light level cameras (Ellis et al., 1990).

224

ELLIS

ET AL.

It is important to note, however, that the analysis of the space-time image as described here, is entirely independent of the means by which the data are sampled. Although image integration and persistence problems can lead to image artifacts, careful examination of Figs, 4 and 6a demonstrates that these artifacts did not significantly affect our ability to ascertain the dynamics of the red blood cells. However, the use of strobed or shuttered illumination with an Intensified CCD camera in our setup would enable one to analyze space-time images for other features of red blood cell dynamics in capillaries such as red blood cell length and cell separation without applying a correction factor for the cell velocity. It should also be noted that the analysis of the space-time image is not restricted to manual techniques but would easily lend itself to automated analysis. SUMMARY The image analysis approach to the evaluation of red blood cell dynamics described here enables one to visualize and quantify a number of important convective transport parameters simultaneously. Since the method is based on the analysis of recorded video images, simultaneous results can be obtained from other capillaries in the same field of view. From the space-time images, the temporal and spatial heterogeneity of red blood cell velocity, lineal density, and red blood cell supply rate are readily apparent. The added visual dimension that this technique provides should be useful to investigators interested in the determinants of blood flow through small diameter vesselsor glass tubes. In combination with our oxygen saturation technique (Ellis et al., 1990) this approach provides the means for evaluating oxygen supply and oxygen extraction in individual capillaries and capillary networks. ACKNOWLEDGMENTS We thank Chris Stein for providing the image used in the analysis (Fig. 4). This research was supported by an operating grant from the Heart and Stroke Foundation of Ontario to Dr. Ellis and grants from the National Institutes of Health to Dr. Ellsworth (HL-39226) and Dr. Pittman (HL18292).

REFERENCES ELLIS, C. G., ELLSWORTH, M. L., AND PITTMAN, R. N. (1990). Determination of red blood cell oxygenation in viva by dual video densitometric image analysis. Am. J. Physiol. 258, H1216-H1223. ELLIS, C. G., FRASER, S., HAMILTON, G., AND GROOM, A. C. (1984). Measurement of lineal density of red blood cells in capillaries in vivo, using a computerized frame-by-frame analysis of video images. Microvasc. Res. 27, 1-13. ELLIS, C. G., TYML, K., AND BURGESS,W. (1987). Quantification of red cell movement in microvessels: A new application of interactive computer graphics. Microvmc. Res. 33, 428-432. ELLIS, C. G., TYML, K., AND GROOM, A. C. (1986). Computer-assisted analysis of video images: A new tool for microvascular measurement. In “Microcirculatory Technology” (C. H. Baker, and W.

L. Nastuk, Eds.), pp. 229-244. Academic Press, New York. ELLIS, C. G., WRIGLEY, S. M., POTTER, R. F., AND GROOM, A. C. (1990). Temporal distributions of red cell supply rate to individual capillaries of resting skeletal muscle in frog and rat. Int. J. Microcirc. Clin. Exp. 9, 67-84. ELLSWORTH,M. L., POPEL, A. S., AND PITTMAN, R. N. (1988). Assessment and impact of heterogeneities

IMAGE

ANALYSIS

OF RED

BLOOD

CELL

DYNAMICS

225

of convective oxygen transport parameters in capillaries of striated muscle: Experimental and theoretical. Microvasc. Res. 35, 341-362. HONIG, C. R., FELDSTEIN, M. L., AND FRIERSON, J. L. (1977). Capillary lengths, anastomoses, and estimated transit times in skeletal muscle. Am. J. Physiol. 233, H122-H129. INOUE, S. (1986). “Video Microscopy.” pp. 191-228. Plenum, New York. INTAGLIETTA, M., BREIT, G. A., AND TOMPKINS,W. R. (1990). Four window differential capillary velocimetry. Microvasc. Res. 40, 46-54. INTAGLIETTA, M., MIRHASHEMI, S., ANDTOMPKINS,W. R. (1989). Capillary fluxmeter: The simultaneous measurement of hematocrit, velocity, and flux. Znt. J. Microcirc. Clin. Exp. 8, 313-320. POPEL, A. S. (1989). Theory of oxygen transport to tissue. Crit. Rev. Biomed. Eng. 17, 257-321. SARELIUS, I. H. (1986). Cell flow path influences transit time through striated muscle capillaries. Am. J. Physiol. 250, H899-H907. SLAYTER, E. M. (1970). “Optical Methods in Biology,” pp. 275-279. Wiley, New York. TYML, K., ELLIS, C. G., SAFRANYOS, R. G., FRASER, S., AND GROOM,A. C. (1981). Temporal and spatial distributions of red cell velocity in capillaries of resting skeletal muscle, including estimates of red cell transit times. Microvasc. Res. 22, 14-31. TYML, K., AND SHEREBRIN, M. H. (1980). A method for on-line measurement of red cell velocity in microvessels using computerized frame-by-frame analysis of television images. Microvasc. Res. 20, l-8.