TY - JOUR
T1 - Quantification of differences between nailfold capillaroscopy images with a scleroderma pattern and normal pattern using measures of geometric and algorithmic complexity
AU - Urwin, S.G.
AU - Griffiths, B.
AU - Allen, J.
PY - 2016/12/23
Y1 - 2016/12/23
N2 - This study aimed to quantify and investigate differences in the geometric and algorithmic complexity of the microvasculature in nailfold capillaroscopy (NFC) images displaying a scleroderma pattern and those displaying a 'normal' pattern. 11 NFC images were qualitatively classified by a capillary specialist as indicative of 'clear microangiopathy' (CM), i.e. a scleroderma pattern, and 11 as 'not clear microangiopathy' (NCM), i.e. a 'normal' pattern. Pre-processing was performed, and fractal dimension (FD) and Kolmogorov complexity (KC) were calculated following image binarisation. FD and KC were compared between groups, and a k-means cluster analysis (n = 2) on all images was performed, without prior knowledge of the group assigned to them (i.e. CM or NCM), using FD and KC as inputs. CM images had significantly reduced FD and KC compared to NCM images, and the cluster analysis displayed promising results that the quantitative classification of images into CM and NCM groups is possible using the mathematical measures of FD and KC. The analysis techniques used show promise for quantitative microvascular investigation in patients with systemic sclerosis.
AB - This study aimed to quantify and investigate differences in the geometric and algorithmic complexity of the microvasculature in nailfold capillaroscopy (NFC) images displaying a scleroderma pattern and those displaying a 'normal' pattern. 11 NFC images were qualitatively classified by a capillary specialist as indicative of 'clear microangiopathy' (CM), i.e. a scleroderma pattern, and 11 as 'not clear microangiopathy' (NCM), i.e. a 'normal' pattern. Pre-processing was performed, and fractal dimension (FD) and Kolmogorov complexity (KC) were calculated following image binarisation. FD and KC were compared between groups, and a k-means cluster analysis (n = 2) on all images was performed, without prior knowledge of the group assigned to them (i.e. CM or NCM), using FD and KC as inputs. CM images had significantly reduced FD and KC compared to NCM images, and the cluster analysis displayed promising results that the quantitative classification of images into CM and NCM groups is possible using the mathematical measures of FD and KC. The analysis techniques used show promise for quantitative microvascular investigation in patients with systemic sclerosis.
UR - http://www.scopus.com/inward/record.url?eid=2-s2.0-85012925101&partnerID=MN8TOARS
U2 - 10.1088/1361-6579/38/2/N32
DO - 10.1088/1361-6579/38/2/N32
M3 - Article
VL - 38
JO - Physiological Measurement
JF - Physiological Measurement
SN - 0967-3334
IS - 2
ER -