VALIDATION OF AN ANALYSIS SOFTWARE FOR OPS-IMAGING USED IN HUMANS 1S.Schaudig , 1B.Dadasch , 2K.R. Kellam & 1F. Christ Clinic for 1Anaesthesiology, Ludwig-Maximilians University Munich, Germany and 2KK Technologies, UK

BACKGROUND: Orthogonal polarization spectral imaging (OPS) is a new technology for intravital microscopy, which allows imaging of the microcirculation in humans and animal models without the use of fluorescent dyes ^[1]. This is a technique, which currently lacks an adequate software for measurement of microvascular diameter and red blood cell velocity. Therefore we adapted an existing analysis software (Capiscope^(r)) for an semi-automated measurement of these parameters.

GOAL: We carried out this study with the purpose to validate the Capiscope^(R) software against the previously used analysis system Cap-Image^(R) .

METHODS: As a basis for our study we chose images from the sublingual mucosa of patients who underwent coronary artery bypass grafting (CPB). For studying microvascular changes during CPB the images, obtained with the Cytoscan^(R) model E II, had formerly been recorded on s-VHS videotapes for later off-line analysis of vessel diameter (DIA [µm]) and red blood cell velocity (VEL [µm/s]). 346 sublingual postcapillary venules of 6 patients had been analysed at 4 points in time: immediately after induction of anaesthesia, at the beginning of cardiopulmonary bypass, during the last 30 minutes of CPB and one hour after reperfusion. These time points had been selected, since especially during these phases of coronary artery bypass surgery changes in diameter and velocity were expected ^[2]. Afterwards both parameters, DIA and VEL, were measured in identical vessels using Capiscope^(R) and CapImage^(R).

DIA is measured by calculating its mean from a gray level scan performed along the length of the vessel (see figure 3) and VEL by conducting an autocorrelation routine (see figure 1). After having finished the calculation a dimension list, which is a summary of the evaluated data is displayed (see figure 2).

Statistics: The correlation was assessed using a Pearson-Product moment and a Bland-Altman-Plot ^[3].

Figure 1 shows the line scan, allowing the assessment of velocity.

Figure 2 shows the dimension list, a summary of the evaluated data, which is displayed on the screen after the calculations have been finished

Figure 3 shows a typical example of an image found with OPS during CPB when applied to the sublingual tissue, where postcapillary venules and capillaries can be seen. After the calculation has been finished the different vessel diameters (DIA) and red cell velocities (VEL) are displayed along the sides of the measured vessels (red). After having finished the semi-automated calculation we remeasured the diameters manually (green lines), in order to achieve an additional validation.

Next to the image a grey level profile is displayed, which reflects the diameter measurement

RESULTS: A good correlation between the two methods was found for diameter r²=0.95 and the Bland and Altman plot showed an acceptable deviation [3,4] (see figure 4).

Velocity however was measured lower with Capiscope^(R) compared to CapImage^(R) particular at higher values of velocity (see figure 5) and the correlation was poor (r²=0.12).

Figure 4 :Correlation of diameter measured with Capiscope^(R) and CapImage^(R) and the corresponding Bland-Altman-plot.

Figure 5 :Correlation of velocity measured with CapImage^(R) and Capiscope^(R) and the corresponding Bland-Altman-plot.

CONCLUSION: Both software packages do not represent a gold standard. Capiscope^(R) however maybe more sophisticated and accurate since it follows clearly defined statistical margins and considers the whole length of the vessel, a fact that is different to the manual analysis of CapImage^(R).

REFERENCES:

[1] Groner W et al. Nat Med 1999;5:1209-12

[2] Christ F et al. Prog Appl Microcirc 2000;24:82-93

[3] Bland JM et al. Lancet 1995; 346: 1085-7