Given the intrinsic variability of the microcirculation 35, several sites of the organ of interest should be averaged. Ideally five sites should be examined, but at image analysis the quality of some images may be less than initially estimated and these should be discarded. Accordingly, we concluded that at least three sites that can be reliably evaluated per patient, and if possible five sites, should be obtained at each evaluation.

One may wish to increase optic magnification in order to enhance visualization of some structures (white blood cells). On the other hand, the increased microscopic precision limits the field of interest to a narrower window, which may be problematic, considering the heterogeneity of the microcirculation. In addition, movement artifacts will be magnified. Accordingly, we recommend use of 5× objectives for human sublingual microcirculation with OPS and SDF devices. In small animals, 10× objectives should be used.

Capillaries and venules are collapsible; accordingly, these vessels may be very sensitive to pressure applied to the organ. Because the microcirculation is just below the microscope, excess pressure applied to the area may collapse the microcirculation, and the investigation of the microcirculation can become unreliable in these conditions. This can result in decreased flow in large venules (venules >30 μm), which may become sluggish, absent, or alternate, or there may even be backflow. Importantly, pressure may be only focal, when the pressure is not applied globally to the preparation but only to one side. Of note, pressure artifacts can also be observed during compression of the tongue (for instance, by the investigators' finger, in an attempt to stabilize the tongue) or during contraction of tongue muscles. Interestingly, all authors have reported that venular perfusion is always preserved, whatever the severity of alteration in smaller vessels 38. Observation of an altered large venular blood flow is thus suggestive of a pressure artifact. To prevent applying pressure to the area, it is recommended that the microscope be pulled back gently until contact is lost and then to advance the probe again slowly to the point at which contact is regained. These aspects are summarized in the operational procedure proposed by Trzeciak and coworkers 5.

Several technical issues should be addressed to ensure adequate image acquisition and further analysis. Video images are usually immediately captured on a computer using a dedicated videocard, and the images should be stored at full size as DV-AVI files to allow computerized frame-by-frame image analysis and use for educational purposes. We recommend limiting recording time to 20 s because it may be difficult to maintain a clear and steady image for a longer period. In addition, longer clips should be divided for further analysis, especially if analysis is performed using software. Clips of 20 s duration are already very large (50 to 100 MB), and the need for adequate storage should be anticipated. To enhance image focusing, large external monitors should be used instead of the LCD screen of the computer. Videotaping the image (and later digitalization of the images) can also be performed if needed, but high-quality digital videotape recording and appropriate labelling of the video strips are necessary. VHS video recording or DVD recording where MPEG compression is used should be avoided because these result in loss of resolution.

It is difficult to separate venules from capillaries. Usually, these vessels are delineated according to their diameter and a cut-off value of 20 μm is used to differentiate capillaries from venules. However, the size of capillaries and venules can be affected by various factors, so this limit can fluctuate. Analyses of larger vessels are of limited interest except as a quality control measure to ensure that no excessive pressure is applied to the tissue. In larger venules, rolling and adherent leucocytes can be observed, but this requires higher magnification and different analytical methods.

Separation of the screen into quadrants (or using equidistant lines) is mandatory when analysis is done by eye. Indeed, it is very difficult to count vessels over the entire screen because the eye may be attracted by specific regions of interest. However, the altered microcirculation is usually heterogeneous, and it is thus important to have a full overview of the image. To obtain a comprehensive measure of the perfusion characteristics, it is advisable to measure both the MFI, and the PVD and PPV. The image is divided into four quadrants and flow is assessed in each quadrant to measure the MFI 15. Three horizontal and three vertical lines are drawn on the screen, and perfusion of each vessel at an intersection with lines drawn on the screen is determined to measure the PVD and PPV 3 of the image. This type of comprehensive analysis (for example, MFI, PPV and PVD) helps to generate a picture of perfusion and perfusion heterogeneity in representative types of vessels, avoiding oversimplification. (Additional files 1 to 4). Drawing quadrants or lines may be obsolete if perfusion of all vessels can be detected by software analyses.

FCD, estimated as PVD, can be calculated either as the number of perfused vessels that cross three horizontal and three vertical lines, divided by total length of lines (as in the report by De Backer and coworkers 3) or as total length of perfused vessels divided by total surface of area (with appropriate software) 20.

Perfused vessels are defined as total number of vessels - (no flow + intermittent vessels). These may be calculated for each type of vessel. Problems may be encountered when the vessel diameter is incorrectly identified and with looping vessels that may be counted twice. Calculation should be made only in images that have not been manipulated. Software can be helpful in stabilizing the image, but this procedure implies some size reduction (Figure 3). The total length of the lines will be affected by this procedure.

To calculate the total length of the lines, we must know exactly the size of the image projected on the screen. The US National Television Systems Committee (NTSC) standard and the phase alternating line (PAL) and sequential colour with memory (SECAM) standards use different displays that may affect the presentation on the screen (720 × 576 pixels for DV-PAL and 720 × 480 pixels for DV-NTSC). The resulting differences in area should be taken into account when using a scoring method worldwide. The optical field of view of SDF imaging with 5× objectives is approximately 0.94 mm × 0.75 mm. A slight difference between the magnifications between OPS and SDF explain small differences in image size. In PAL/SECAM standard the OPS system gives an image size of 1.54 mm × 1.15 mm (1.54 mm × 0.96 mm in NTSC), and the SDF gives an image size of 0.98 × 0.73 mm (0.98 mm × 0.60 mm in NTSC). Of note, the length and width of both systems can slightly vary during focusing because both OPS and SDF devices focus by moving the camera closer or further away from the tissue, altering the magnification. Usually, this effect is quite limited but it can be as large as 10% when the full range of focus (0 to 1 mm depth 1) is explored.

The ideal software, we propose, should automatically recognize all blood vessels and measure their diameters and blood flow in each individual vessel of the investigated field. This is not currently available. Semi-quantitative analysis should therefore be used; such analysis has been proven to be able to distinguish between health and disease 358.

Our consensus is that all three indices discussed above (PVD, PPV and MFI) should be measured to describe comprehensively the functional perfusion of the microcirculation. Looking at PPV allows no distinction to be made between normal, sluggish and hyperdynamic flows, but it provides information on flow heterogeneity within the image. PVD provides an accurate estimate of FCD.

In addition to making the distinction between perfusion and nonperfusion, the MFI score differentiates between the different types of continuous flows (sluggish, normal and high flow). In conditions where flow is homogeneous, the MFI score can thus provide additional information. However, with this method the capillary density, and thus FCD, is not estimated. Hence, the proportion of perfusion should be used in heterogeneous situations, whereas MFI should be preferred in more homogenous conditions because it takes into account the difference between sluggish and continuous flows.

The microcirculation is heterogeneous in many disease states, and for this reason it has been proposed that several areas be averaged. The heterogeneity can be quantified. Initially, the coefficient of variability was used 3. More recently, Trzeciak and coworkers 5 proposed another heterogeneity index, which involves evaluating three to five sites and measuring the MFI in the quadrants, taking the difference between highest MFI minus the lowest site MFI divided by the mean flow velocity of all sublingual sites at a single time point. This heterogeneity index has the advantage of taking into account extreme deviations, whereas the coefficient of variation evaluates all deviations from the mean. From a pathophysiological point of view, the heterogeneity is a key determinant of the shunted fraction, often seen in distributive shock. For this reason, taking into account the extreme deviations is more representative.

An analysis of the microcirculation (Table 3) should be reported for both total and small (<20 μm) vessels. The consensus is to report PVD, PPV and MFI to describe the functional perfusion of the microcirculatory image. The heterogeneity index (calculated as the difference between extreme values of either MFI or PPV between the three to five recordings of the organ divided by its mean value) is needed to describe the heterogeneity of perfusion in the microcirculatory area under observation.

The interpretation of these variables may sometimes be difficult, especially when discordant changes between the different indices occur during interventions.

Tissue perfusion is dependent on FCD (reflected by PVD) and blood velocity (reflected by MFI). As discussed above, vascular density is probably more important than blood velocity in determining tissue perfusion, because oxygen extraction can compensate for a decreased flow. Shunt fraction, a key determinant of oxygen extraction capabilities 162122, is reflected by blood flow heterogeneity in the investigated area by PPV and between the different areas of the investigated organ by the heterogeneity index.

Several software packages have been developed, allowing FCD calculation or reliable blood flow measurements in individual vessels. The CapImage software (Dr Zeintl software Engineering, Heidelberg, Germany) has been developed for intravital microscopy 23 but can also be used for the analysis of OPS and SDF images 2425. This software is validated for blood flow measurements in straight vessels segments only. The CapiScope software (KK Technology; Honiton, UK) has been developed for analysis of OPS images. It measures FCD, and vessel diameter and velocity. It reliably measures blood flow in individual vessels. Very stable images, without any movement artifact, should be used with these two software packages because they do not provide image stabilization. Recently, the MAS analysis system (MicroVision Medical, Amsterdam, The Netherlands) was developed. It includes a stabilization image processing, a calculation of FCD and measurements of blood flow in individual vessels. Unfortunately, these packages still require much user intervention to identify the vessels of interest. In addition, flow cannot be calculated automatically and simultaneously in multiple vessels, so that blood flow distribution histograms can not readily be obtained. In addition, it is particularly difficult to measure blood flow in capillaries, which constitute the main area of interest. Blood flow measurement is calculated as cross-sectional area (based on measurement of vessel diameter) times blood velocity. The error in determining the flow is especially large with errors in measurement of vessel diameter, since it is the square of the diameter that is used in cross-sectional area (πD²/4) calculation. Determination of vessel diameter is difficult in small capillaries, and consequently the relative error in measurements may be greatest in small vessels. Vessels are visualized because they contain red blood cells but the vessel wall is not visualized. In most vessels, multiple red blood cells flow side by side, allowing easy identification of vessel diameter. This is more complicated in small capillaries, especially when red blood cells are separated by plasma gaps. Software with time averaging of sequential frames and better imaging modalities may improve the accuracy of these measurements.

One may anticipate that in the future FCD measurement will be mechanized. Although FCD may be obtained automatically, this process is likely to require some human validation, ideally by clicking away vessels that do not appear to be perfused. The human eye can easily draw a vessel when red blood cells are separated by large plasma gap, whereas this is will probably continue to be a limitation of software in the short term.

All variables should be separated according to vessel size using a cut-off of 20 μm. Histogram of vessel diameter and vessel flow would provide not only mean values but also identify variability in the measurements.

Vessel flow measurements require a moving feature (isolated red blood cell or white blood cell) to be visible in at least three consecutive movie frames. The highest computer-aided measurable velocity is physically restricted by the video frame rate (30 frames/s for NTSC and 25 frames/s for PAL and SECAM) and by the length of the vessel part where the flow is assessed. Faster cameras with higher frame rates could overcome this physical limitation. This is important because with current conventional cameras the flow in fastest flowing vessels can not be calculated 26.

Stabilization processes incorporated in software packages are very helpful in improving image readability and computerized analysis. However, problems of isotropy (see above) are encountered when FCD is determined semi-quantitatively using the six lines methods. Independently of the analytical method used, some information will be lost. Indeed, peripheral parts of the images, not seen on successive images when movements occur, will be lost during the stabilization process (Figure 3). As a result, the final image is smaller than the original one, but the software displays this transformed image at the same size as the original image, altering its magnification. Ideally, the percentage of reduction from the original size should be provided by the stabilization software, but this information is not currently provided.