Digital Motion Analysis As A Tool For Analysing The Shape And Performance of The Circulatory System in Transparent Animals

The use of photographic cine film images to analyse physiological events dates back several decades, but with respect to the documentation of circulatory phenomena or cardiac activity it became an important tool in the 1960s and 1970s. One of the first applications was probably the measurement of skin perfusion in humans using a so-called videocapillary microscope with frame-to-frame analysis (Fagrell et al., 1977; Hudetz, 1997), and both this method and the laser Doppler system were very soon used in comparative animal physiology (Koyama et al., 1975; Farrell et al., 1980; Hughes et al., 1981). Within recent years, video imaging has become the main tool for the analysis of cardiac performance in transparent amphibian and fish embryos and larvae (Faber et al., 1974; Burggren and Fritsche, 1995; Colmorgen and Paul, 1995; Hou and Burggren, 1995; Orlando and Pinder, 1995; Fritsche and Burggren, 1996; Keller et al., 1996).

The experimental procedure is quite simple: one needs a dissecting microscope or a microscope with a video adapter and a video recorder. Video recordings are then made of a beating heart or of moving blood cells, and these recordings are analysed using the replay mode and slow-motion facilities. This observer-dependent and time-consuming method provides information about red cell velocity and blood flow (Pelster and Bemis, 1992) and about heart rate, stroke volume and cardiac output (see Burggren and Fritsche, 1995). A more advanced method includes the digitisation of images using a frame-grabber card and subsequent processing of the images using computer programs such as Adobe Photoshop (MacFarlane et al., 1983; Vesely et al., 1990; Sundin, 1995; Nilsson et al., 1995; Sundin and Nilsson, 1997; Hudetz, 1997). Although digitisation makes the analysis procedure easier, it still remains an observer-dependent method in which the result may vary with the illumination and contrast of the video image. The determination of vessel diameter is particularly difficult because the vessel walls often disappear in still images of transparent tissues (Hanger et al., 1995), and in fish gills, for example, the visibility of the cell membrane of pillar cells depends on the presence of moving erythrocytes (Stensløkken et al., 1999).

The fact that the visibility of blood vessel walls in most transparent tissues is dependent on the presence of moving erythrocytes led us to the idea of visualising the movement of floating erythrocytes by generating a difference image from the odd and even fields of a video frame. This manipulation permits not only an elegant and easy determination of erythrocyte velocities but, by summing several difference images, also a visualisation of the vascular system and of the distribution of erythrocytes within the vascular bed.

The experiments were performed using the almost transparent swimbladder tissue of the European eel and various larval stages of the zebrafish. European eels (Anguilla anguilla) were obtained from a local supplier and kept in a freshwater aquarium at 12 °C. Experiments with eels were performed at 20±1 °C. Zebrafish (Danio rerio) larvae were obtained from our own breeding colony. Because of their greater transparency, poorly pigmented mutants of the zebrafish (Albino, Brass, Transparent) were used. Parent animals used to start the breeding colonies were either obtained from a local supplier or generously provided by Dr Frohnhöfer from the Max-Planck Institute for Developmental Biology in Tübingen and Ms Loos from the University of Konstanz. Breeding colonies and larvae were kept in small aquaria at 28 °C. From day 6 post-fertilisation, the animals were fed with a fine nutrient powder. Experiments with zebrafish larvae were performed at 28 °C.

Anaesthetised eels were quickly immobilised by penetrating the skull with a fine needle and spinal pithing. The animals were placed into an eel-holder (Schwerte et al., 1997), and the gills were irrigated with well-aerated water (20–22 °C) at a constant flow of approximately 1.5–2 l min⁻¹. The body wall was opened ventrally. The swimbladder was carefully exposed and freed of connective tissue. Blood vessels from other tissues entering the vein leaving the retia were ligated.

To perfuse the swimbladder with blood, the swimbladder artery was occlusively catheterised and the dorsal artery was cannulated non-occlusively (PE 50). Blood was drawn from the dorsal artery and infused into the swimbladder artery at a constant flow rate. Thus, after swimbladder perfusion, the blood was returned to the central circulation and reoxygenated in the gills. Catheter tubing was heparinized with 100 i.u. ml⁻¹ saline (see Pelster, 1994). Video recordings were made using a Leica MZ8 dissection microscope with an attached black- and-white CCD camera. The tissue was illuminated from the side to enhance image contrast due to light reflection from the erythrocytes.

For the experiments, embryos and larvae of the zebrafish 2–21 days post-fertilisation were used. The animals were anaesthetised by adding 50 mg l⁻¹ phosphate-buffered tricaine (MS222; pH 7.0) to the water. After 1 min of anaesthesia, the animals were removed and carefully mounted in the experimental chamber with a thin layer (100–200 μm) of 1 % low-melting-point agarose (Sea Plaque; gelling point 26–30 °C), containing 50 mg l⁻¹ phosphate-buffered tricaine. The gel was covered with air-equilibrated water containing 50 mg l⁻¹ phosphate-buffered tricaine. This chamber was transferred into the temperature-controlled microscope desk of an inverted microscope (Zeiss Axiovert 25CF). The inverted microscope allows a plane view of the ventral or lateral side of the larvae. The left side of the animal provided the best view of the central cardiac system.

To test for the presence of α-adrenergic receptors, a micropipette was used to eject 300 μl of a 10⁻⁶ mol l⁻¹ solution of the α-adrenergic agonist phenylephrine within 10 s very close to an intersegmental vessel in the tail region of zebrafish larvae. The mucus on the skin was removed carefully using a broken micropipette 5 min before acquiring control data.

To reduce vibratory movements, the inverted microscope was placed on a solid heavy-weight steel plate. The illumination could be set to infrared light with a wavelength of 780 or 913 nm to prevent light-induced stress reactions of the animals. The microscope was equipped with a 2/3 inch monochrome machine vision CCD ^Camera (M300, JAI, Copenhagen/Denmark; without infrared cut-off filter), which in turn was connected to the luminance input (the chrominance input was left unconnected) of an SVHS video recorder (Sony S-9500). The video recorder was remotely controlled via the RS232 serial communication port. The setting of the video recorder and the recorded images were digitised using a monochrome frame-grabber card (PX-610, Imagenation, Beaverton, USA) with a personal computer (PIII 450 MHz). All measurements and image enhancement procedures were performed using custom-made Analytical Language for Images (ALI) programs running on the Optimas 6.5 programming platform (Media Cybernetics, Silver Springs, USA). The analytical procedures were optimised for automatically exporting data and enhanced images to files. Some of the data were acquired in real time, while the video recorder stored the video data for archiving purposes. This was strictly necessary for measurements in tiny vessels because the signal from the video recorder was too noisy for appropriate resolution.

A frame from an interlaced standard CCD video camera consists of two fields. The first field contains all odd lines and the second field all even lines. The acquisition of each field takes 20 ms. The pixels within each field can be separated into static pixels, located in non-moving areas of the image where no change in gray-scale occurs compared with the subsequent field, and dynamic pixels, located in areas where movement occurs and changes in gray-scale compared with the following field take place. In consecutive fields, the gray-scale value of any pixel in a static area remains constant, while the gray-scale value of a pixel in an area where movement occurs changes. Thus, subtraction of the contents of two consecutive fields (with elimination of negative values) results in a clear separation of the static pixel and half the dynamic pixels. In a static area, the subtraction of two consecutive fields results in a gray-scale value of zero (black), while any dynamic pixel will differ from black with a gray-scale value ranging from 1 (dark gray) to 255 (white).

The sum of all dynamic pixels on a horizontal line derived from the movement of a single object (a red cell, for example) is called the shifting vector, and the length of this shifting vector represents the distance travelled by the object. Fig. 1 illustrates the generation of shifting vectors for erythrocytes floating in the bloodstream. To gain most advantage from the shifting vectors produced by interlaced video images, the movement of the erythrocytes projected onto the camera chip should occur along an axis parallel to the lines of the CCD chip. Because the length of the shifting vector represents a measure of the velocity of the object, and not the gray-scale value, the contrast of images including all shifting vectors was enhanced by application of a 3×3 median filter and subsequent binarization, which turns all pixels with a gray value higher than 1 to white. The maximum detectable velocity V_max (in ms) is dependent on the visual diameter of the object projection D_object:

Because the maximum detectable velocity is greatest in a median section of a round erythrocyte (see also Fig. 1C), only the shifting vectors from the middle area of a blood cell were used for further analysis. This determination was repeated for all fields for three complete cardiac cycles to obtain a mean value for the velocity of the erythrocyte. Once the region of interest had been defined, complete data extraction from the images was performed automatically.

The shifting vectors derived from moving erythrocytes can also be used for contrast enhancement in the vascular system. By summing the visualised non-binarized shifting vectors of a sufficiently large number of frames, a complete outline of all areas in which erythrocyte movement occurred was obtained. This is essentially a cast of the vascular system, except for those tiny capillaries that do not permit the passage of erythrocytes (Fig. 2). The number of repetitions was carefully set to prevent pixel values from exceeding gray value 255 (white).

The difference in gray-scale value between the background and an erythrocyte of a zebrafish larva was between 8 and 12. Therefore, 20–30 repetitions were sufficient to obtain a good image. To improve the image further, a 3×3 median filter was applied, and gray values lower than the number of iterations were discarded to minimise noise.

The cast of the vascular bed, obtained by summing the visualised non-binarized shifting vectors of a sufficiently large number of frames, was also used to determine the diameters of blood vessels at any region of interest. In these regions of interest, the boundaries of the blood vessel were detected by a freely defined gray-scale threshold. An algorithm was developed to find a vector that described the properties of a rectangle with the same binary moments as the extracted area found by the defined threshold. The lengths of the major axis and the minor axis were extracted for direct measurement of the diameter and the length of the vessel in the region of interest.

To compensate for minor movements of the animal, a reference point was determined in the vicinity of the measurement site (see below). This reference point was automatically relocated prior to every determination of the vessel diameter. The images used for measurement of vessel diameter were recorded using a 40× phase-contrast objective.

To visualise blood distribution, the shifting vectors were binarized. The process of binarization (the threshold was set from 5 to 12) reduces the image information on the visualisation of so-called events, where an event is the movement of an erythrocyte. By summing these binarized images, the gray-scale value of any given pixel or of a defined number of pixels in the image rises from zero to 255 depending on the number of erythrocytes passing it. The resulting gray-scale value therefore codes the number of erythrocytes that have passed the region of interest. Although the depth of the gray-scale display on the screen is limited to 8 bits, the actual range for the calculations was extended to 24 bits. Thus, differences in blood distribution in whole animals could be visualised in overview images using a 4× phase-contrast objective. The procedure ran automatically, while measurements of erythrocyte distribution could be made automatically in predefined regions of interest. The dimensions of the visible fields and the size of the pixels are listed in Table 1.

In blood vessels smaller than 100 μm in diameter, the velocity of erythrocytes was determined by multiplying the number of frames recorded while an erythrocyte travelled a given distance by 40 ms, the time needed to acquire a whole frame (frame-to-frame technique). The custom-made program automatically displayed the image sequence from the video tape, while the user tracked individual erythrocytes using the mouse. The velocity was calculated and exported automatically. The imaging system was calibrated using an object micrometer. Each measurement was repeated five times, and the readings obtained were averaged to flatten the pulses of the cardiac rhythm.

The diameter of blood vessels was determined on the video image by drawing an orthogonal line between the vessel walls. The measurements were made using the Optimas program package (Media Cybernetics). To compensate for possible movements that may have shifted the site of measurement between subsequent determinations, a reference point was defined in the vicinity of the measuring site. ‘Landmarks’ such as characteristic branching points of blood vessels were used as reference points.

The diameter of blood vessels was also determined following an intra-ventricular injection of the fluorescent probe fluorescein isothiocyanate (FITC) bound to a high-molecular-mass dextran (M_r 5×10⁵) to visualise the blood plasma. The experiments were performed with a fluorescence microscope (Zeiss Axiovert 100, Vienna, Austria). The fluorescent plasma made the measurement of vessel diameter possible by detecting the vessel walls and looking for the greatest change in contrast between the plasma and the non-fluorescent background. Once the region of interest had been defined, the complete data extraction from the images ran automatically.

Both the acquired data and the extracted data were automatically exported into an ASCII file for statistical analysis. After preprocessing the data in Microsoft Excel, statistical analyses were performed using the software package Statistica.

By subtracting the odd and even fields of one video frame, shifting vectors were obtained, the length of this shifting vector for an individual object being a measure of the distance travelled by that object within the 20 ms necessary to acquire one field of a frame. As shown in Fig. 3, there was a linear correlation between the velocity of zebrafish erythrocytes determined on the basis of the length of the shifting vector and the velocity determined using the conventional frame-to-frame technique. The velocity determined by generating a shifting vector by digital motion analysis reached a maximum value at approximately 450 μm s⁻¹. The erythrocytes of the eel are slightly larger than the erythrocytes of the zebrafish, and the maximum length of the shifting vector in the eel was reached at a velocity of approximately 600 μm s⁻¹. Thus, in zebrafish larvae, erythrocyte velocities greater than 450 μm s⁻¹ could not be measured using the digital motion analysis technique. Using the frame-to-frame technique, higher velocities could be measured as long as the density of red blood cells was low enough to identify individual cells.

In early fish larvae and in transparent animals, the blood vessels are, depending on the haematocrit and blood flow, very pale and their edges are very hard to define in video images. In still images, the edges of a vessel often almost disappear (Fig. 4A). It was not surprising, therefore, that measurements of vessel diameter showed striking variations between independent observers (Fig. 5A). Even repeated diameter determinations made by the same observer gave varying results depending on the contrast, brightness and optical focus of the image. An intensity profile of an orthogonal line through a blood vessel of a typical video image revealed that, indeed, the gray-scale histogram showed no characteristic peak for object separation (Fig. 4A).

Using digital motion analysis, the vascular bed was visualised by summing images of shifting vectors from several consecutive frames. The images of the vascular bed obtained showed a clear contrast between blood vessel and background (Fig. 4B). The gray-scale histogram of such an image (Fig. 4B) clearly indicated the location of blood vessels and enabled the diameter of a vessel to be easily determined (Fig. 5B).

Blood vessel diameter in zebrafish larvae was also determined following visualisation of the vascular bed by injection of a macromolecule-bound fluorescent probe (FITC–dextran; Fig. 4C). Compared with the image obtained by digital motion analysis, the background was much less homogeneous, and the contrast between blood vessels and background was much less pronounced. Accordingly, the gray-scale histogram was blurred at the edges of the blood vessels (Fig. 4C). Furthermore, injection of FITC–dextran turned out to be quite difficult in these tiny animals, and several hours after the injection fluorescence was always noticed in tissue cells, resulting in an increasing background signal. The fluorescence signal faded within two or three days. The cardiac activity of the larvae was often significantly reduced following the injection. Therefore, no further development of this method was attempted.

By accumulating the shifting vectors derived from moving erythrocytes, an image of the routes over which erythrocytes moved was obtained. This led to the idea that the gray-scale value of a given pixel should increase in linear proportion to the number of erythrocytes passing it, as long as the velocity of the erythrocytes or the density of erythrocytes does not exceed the time resolution of the video camera. As demonstrated in Fig. 6, there was a indeed a linear correlation between the erythrocyte count achieved using the frame-to-frame technique and the erythrocyte count based on the gray-scale value obtained as a result of the digital motion analysis procedure. Thus, differences in blood distribution in whole animals could be visualised in overview images using a 4× phase-contrast objective.

Fig. 7 shows as an example the development of the vascular system in zebrafish larvae. Four days post-fertilisation, the circulatory system is operational, and erythrocytes were encountered mainly in the yolk sac, but also in the dorsal artery and vein. The head region was also well-perfused with red cells, but segmental vessels in the muscle tissue were barely detectable. At 5 days post-fertilisation, the segmental vessels were perfused with erythrocytes, and the size of the yolk sac was reduced. At 6 days post-fertilisation, the vascular system supplying the inner organs became visible, and the yolk sac was almost completely resorbed. At this stage, the erythrocytes were more evenly distributed within the vascular bed than at days 4 and 5 post-fertilisation, and this tendency became more obvious at 7 days post-fertilisation. At 7 days post-fertilisation, the swimbladder was inflated. At all stages, the ventricle could be clearly identified.

Fig. 8 shows the blood distribution in a fed and an unfed zebrafish larva at approximately 8 days post-fertilisation. In the fed larva, erythrocytes were concentrated in the gut area, indicating a good blood supply to the gut, while in the unfed animal almost no erythrocytes could be detected in the gut. In contrast, the opposite effect was observed in the intersegmental vessels. Thus, a significant redistribution of blood between the gut and the segmental muscle tissue was possible at this stage of development.

At this stage, adrenergic control of peripheral blood vessels was established in zebrafish larvae. As shown in Figs 9 and 10, application of the α-adrenergic mediator phenylephrine resulted in a significantly lower erythrocyte count in the segmental blood vessels supplying the muscle tissue. Measurement of the blood vessel revealed that the diameter of the segmental vessel in the middle region was unchanged but that, at the entrance to the segmental vessel, the diameter was significantly reduced (Fig. 10). Many erythrocytes turned into the branch but did not pass this constricted segment and returned to the dorsal artery (Fig. 9B). The first segment of the segmental artery therefore appeared to function as a ‘sphincter’.

Video-microscopic analysis of cardiovascular performance requires the immobilisation of tissues, because any tissue movement shifts the region of interest and obscures the movement of red cells. In our study, we used infrared illumination and a high-viscosity agarose medium combined with light anaesthesia to prevent the movement of zebrafish larvae. Particularly in the early larval stages, the infrared illumination (band-pass filters of 780 nm and of 913 nm) proved to be very effective in settling the larvae. Compared with illumination using the full spectrum of a tungsten lamp, the larvae were significantly less active in the observation chamber and often remained quietly on the bottom, so that recordings could easily be made using the inverted microscope. The CCD cameras, which are very sensitive to these wavelengths, provided excellent images. Using infrared illumination, recordings of the cardiovascular system in zebrafish and minnow larvae (G. Schönweger, T. Schwerte and B. Pelster, in preparation) were possible until inflation of the swimbladder without any anaesthesia or physical restriction of the animals.

To visualise the vascular bed, the shifting vectors of several frames had to be superimposed, and this required the immobilisation of animals using a high-viscosity medium. A widespread method for mounting small animals for micro-injections is incubation in low-melting-point agarose. The gelling temperature of the agarose (26–30 °C) is within the physiological range of temperature for zebrafish. High-viscosity solutions reduce water convection and therefore increase the magnitude of unstirred layers, which significantly impairs gas exchange (Pinder and Feder, 1990; Feder and Booth, 1992). Measurements of oxygen equilibration in a 150 μm layer of 1 % agarose in comparison with a layer of stirred water indeed revealed that the time necessary to equilibrate the agarose was approximately twice as long. However, the cardiac activity of zebrafish larvae was not affected in agarose. Until 15 days after fertilisation, which is approximately the time when opercular movements start (at 25 °C), heart rate, stroke volume and cardiac output were not significantly altered following agarose incubation (T. Schwerte, unpublished results). We therefore conclude that, in the early larval stages, incubation in agarose can be used to immobilise the larvae. Older zebrafish (15 days post-fertilisation and older), however, were significantly affected by incubation in agarose, heart rate decreasing within a few minutes.

For immobilisation, approximately 50 mg l⁻¹ MS222 (tricaine) was added to the water. MS222 is a known sympathetic agonist, and an overdose (600 mg l⁻¹) can be used to stop the heart (Weinstein et al., 1996). The concentration used in our study was very low, and a comparison of the cardiac activity of MS222-incubated animals and control animals analysed under infrared illumination revealed no significant differences. This is in line with the results of Mirkovic and Rombough (1998), who reported that a concentration of 20 mg l⁻¹ suppressed spontaneous body movements in larvae of rainbow trout (Oncorhynchus mykiss), but had no significant effect on heart rate and on the rate of opercular movements even when applied 30 min prior to experiments. Similar results were also obtained for larvae of the clawed frog Xenopus laevis in which, during the early larval stages, MS222 did not affect cardiac activity (R. Fritsche, personal communication).

For measurements on very tiny vessels, the noise, even from excellent SVHS video recorders, sometimes decreases the resolution. For this reason, processing the camera signal directly provided more accurate measurements. The output from the recently available DV recorders is much closer to this live camera signal, although the video compression used sometimes produces artefacts. The programs described in this study can easily be adopted to both full- and half-frame-storing recorders.

Ultrasonic Doppler flowmeters are appropriate for making flow measurements on small arteries and veins. In our experience, this method is limited to vessels larger than approximately 100 μm in diameter. Using a focused laser beam, the laser Doppler anemometer is suitable for even smaller vessels down to a diameter of approximately 10 μm (Stücker et al., 1996). Using Doppler flowmetry, the signal-to-noise ratio decreases severely at low velocities and, especially with ultrasound, low velocities are very difficult to measure. The digital motion analysis technique can easily detect very slow movements but is limited in the upper range of velocities. The shifting vector generated by subtraction of the odd and even fields of a video frame cannot be longer than the moving object; therefore, the maximum velocity that can reliably be measured is the diameter of an erythrocyte per 20 ms. For zebrafish larvae, the maximum velocity was approximately 450 μm s⁻¹, while for eel erythrocytes a velocity of 600 μm s⁻¹ was measurable. The low range of velocities is simply limited by the pixel size. Like Doppler flowmetry, digital motion analysis provides information about instantaneous velocities, but it also permits the velocity of a single erythrocyte to be determined and, by integrating the velocity signal over several erythrocytes and/or over a user-defined time window, for example, allows cardiac-cycle-dependent velocity fluctuations to be eliminated.

Thus, digital motion analysis lends itself particularly to the analysis of red cell velocity in small transparent animals. Compared with the conventional frame-to-frame technique, it also has the advantage that it is much less dependent on the illumination and contrast of the preparation. Digital motion analysis is not observer-dependent, and this significantly increases the accuracy of the measurements. Furthermore, the analysis is much easier and less time-consuming and the velocity morphometric data on the blood vessels can be acquired almost simultaneously.

Small blood vessels in fish and amphibian larvae often have a very low contrast, and the diameter estimations even of experienced observers show large variations. This may in part be due to minor misplacements of the optical focus or to changes in the contrast and brightness of the image. The proportional error increases with a decrease in vessel diameter. Visualisation of a blood vessel by superimposing shifting vectors calculated from several frames generates a very good image of the vessel, with sharp edges, which is insensitive to minor changes in the optical focus. The disadvantage of this method is that it visualises only those vessels in which erythrocytes are encountered and it may underestimate the diameter of a vessel as a result of axial red cell accumulation and plasma sleeving (Schmid-Schönbein, 1988). This effect occurs in the presence of a tiny layer of erythrocyte-free plasma at the surface of the endothelium. On the basis of model calculations about red cell movements in a capillary system, this plasma layer is estimated to have a thickness of approximately 1 μm or less (Schmid-Schönbein, 1988). To evaluate the accuracy of our method, we compared the blood vessel thickness determined from the accumulated difference images with the diameter measured from a complete video image of the same site. The results differed by approximately 2–3 %. We conclude, therefore, that using the movement of the red cells to create a picture of the vascular bed is an extremely useful method of analysing the vascular system in larval fish.

Compared with the normal video image, the use of high-molecular-mass fluorescent probes to visualise blood plasma creates better images of the vascular system, but the gray-scale histogram of such an image shows the edges of a blood vessel less clearly than the gray-scale histogram of an image generated using digital motion analysis. Compared with the use of fluorescent probes, our method opens up the possibility of visualising the development of the vascular bed and the capillarization of tissues in long-term experiments. While fluorescent probes fade after a few days and apparently can even be taken up by cells, the visualisation of blood vessels by summing shifting vectors of erythrocytes is not invasive and can be performed repeatedly in the same animal over a period of days or even weeks, depending on the how much the pigmentation of the skin increases. In contrast to fluorescence microscopy, the use of laser scanning microscopes provides a much better spatial resolution that enables a three-dimensional reconstruction of whole vessels. But, for this imaging technique, fluorescent probes have to be injected, which is very invasive (see above). Furthermore, because of the higher light intensities required, the fading effect is much higher than for wide-field fluorescence microscopy, and the use of anti-fading agents in living animals is problematic.

Especially in long-term experiments, the relocalization of the region of interest is critical. For morphometric analysis of an image, orientation of the vascular system according to landmarks (bends in a vessel or vessel bifurcations) is an accepted procedure (MacFarlane et al., 1983; Vesely et al., 1990; Sundin, 1995; Sundin and Nilsson, 1997; Hanger et al., 1995; Hudetz, 1997; Nakazawa and Kajio, 1997). Often it is not the landmark itself where the measurement takes place, and minor errors in the exact relocalization may well contribute to the observed scatter in morphometric measurements (see Fig. 5A). Using our method, this problem is easily alleviated by defining a reference point which is automatically relocalized by the computer prior to each measurement.

Measurement of erythrocyte velocity using the frame-to-frame technique is possible in smaller blood vessels in which single blood cells can be observed, and this technique may therefore also be used to count the number of red cells present in a particular region of a blood vessel. Especially at high flow rates, counting erythrocytes is very difficult, and counting them using the frame-to-frame technique is very time-consuming. Motion-based image analysis permits red blood cells to be counted in real time because the gray-scale value of any pixel in the superimposed difference image is correlated to the number of erythrocytes encountered during the measurement. Thus, the distribution of erythrocytes within a tissue or within a larva can easily be visualised. Our experiments nicely describe the development of the peripheral vascular system in zebrafish larvae. At the time of first feeding (at 7 or 8 days post-fertilisation), food intake already resulted in a clear redistribution of blood between muscle tissue and the gut. One way of achieving a redistribution of blood is the selective distribution and activation of α- and β-adrenergic receptors within the peripheral circulation. The reduction in the red cell count within segmental blood vessels following the application of phenylephrine clearly demonstrates that this mechanism is already established in zebrafish larvae at 8 days post-fertilisation. In additional experiments, we were also able to show that the peripheral vascular system of zebrafish larvae at this stage responds to nitric oxide (R. Fritsche, T. Schwerte and B. Pelster, in preparation).

In summary, the present study demonstrates that the visualisation of erythrocyte movements by subtracting the two fields of a video frame provides a very effective tool not only for measuring erythrocyte velocity and blood flow in transparent embryos and larvae of lower vertebrates but also for visualising the capillary bed and the distribution of red cells within the vascular system. As a non-invasive technique, it can also be used for long-term observations to visualise angiogenesis and vasculogenesis in growing embryos and larvae.

The study was financially supported by the Fonds zur Förderung der wissenschaftlichen Forschung (FWF, P12571-BIO).