Skip to Main Content
A methodology for segmentation and extraction morphologic feature from nailfold capillaroscopic images is presented. The main characteristic of the images studied here is the low contrast between the background and the capillaries.For this reason, three fundamental steps were applied in the preprocess: correction of the illumination, highlight and smoothing. For segmenting these images, Laplacians of the most contrasted component in each color space and the connectedness by threshold (region growth) were integrated. The extraction was carried out using image processing techniques such as principal components analysis (PCA), fractal geometry and tortuosity index (TI); their properties were proven. Tortuosity index is a clinical variable subjective to the expert, it is presented as the ratio between the area and the fractal dimension (FD) of the capillary region. Other features obtained were width and height, density of capillaries, area and perimeter, orientation and polarity. The work was carried out on 300 capillaries obtained from images of subjects that do not suffer vascular diseases of the connective tissues and 250 capillaries of patients that have Lupus erythematosus (SLE). Images were taken from the third and fourth fingers of both subjectpsilas hands. The application of the automatic segmentation allowed the classification of the capillary tortuosity and the comparison to the manual segmentation which was made on 47 capillary images by an expert in dermatology.