Semi-automatic assessment of skin capillary density: Proof of principle and validation.

. Numbers and letters correspond to the affiliation list. Click to expose these in author workspaceOpens the author workspaceE.H.B.M.Gronenschild, D.M.J.Muris, M.T.Schram, Ü.Karaca, C.D.A.Stehouwer, A.J.H.M.Houben

 

Background

Skin capillary density and recruitment have been proven to be relevant measures of microvascular function. Unfortunately, the assessment of skin capillary density from movie files is very time-consuming, since this is done manually. This impedes the use of this technique in large-scale studies. We aimed to develop a (semi-) automated assessment of skin capillary density.

Methods

CapiAna (Capillary Analysis) is a newly developed semi-automatic image analysis application. The technique involves four steps: 1) movement correction, 2) selection of the frame range and positioning of the region of interest (ROI), 3) automatic detection of capillaries, and 4) manual correction of detected capillaries. To gain insight into the performance of the technique, skin capillary density was measured in twenty participants (ten women; mean age 56.2 [42–72] years). To investigate the agreement between CapiAna and the classic manual counting procedure, we used weighted Deming regression and Bland–Altman analyses. In addition, intra- and inter-observer coefficients of variation (CVs), and differences in analysis time were assessed.

Results

We found a good agreement between CapiAna and the classic manual method, with a Pearson's correlation coefficient (r) of 0.95 (P < 0.001) and a Deming regression coefficient of 1.01 (95%CI: 0.91; 1.10). In addition, we found no significant differences between the two methods, with an intercept of the Deming regression of 1.75 (− 6.04; 9.54), while the Bland–Altman analysis showed a mean difference (bias) of 2.0 (− 13.5; 18.4) capillaries/mm2. The intra- and inter-observer CVs of CapiAna were 2.5% and 5.6% respectively, while for the classic manual counting procedure these were 3.2% and 7.2%, respectively. Finally, the analysis time for CapiAna ranged between 25 and 35 min versus 80 and 95 min for the manual counting procedure.

Conclusion

We have developed a semi-automatic image analysis application (CapiAna) for the assessment of skin capillary density, which agrees well with the classic manual counting procedure, is time-saving, and has a better reproducibility as compared to the classic manual counting procedure. As a result, the use of skin capillaroscopy is feasible in large-scale studies, which importantly extends the possibilities to perform microcirculation research in humans.