Automatic protocol for quantifying the vasoconstriction in blood vessel images.

Vascular targeted photodynamic therapy (V-PDT) has been successfully utilized for various vascular-related diseases. To optimize the PDT dose and treatment protocols for clinical treatments and to elucidate the biological mechanisms for V-PDT, blood vessels in the dorsal skin-fold window chamber (DSWC) of nude mice are often chosen to perform in vivo studies. In this study, a new automatic protocol to quantify the vasoconstriction of blood vessels in the DSWC model is proposed, which focused on tracking the pixels of blood vessels in pre- V-PDT images that disappear after V-PDT. The disappearing pixels indicate that the blood vessels were constricted, and thus, the vasoconstriction image for pixel distribution can be constructed. For this, the image of the circular region of interest was automatically extracted using the Hough transform. In addition, the U-Net model is employed to segment the image, and the Speeded-Up Robust Features algorithm to automatically register the segmented pre- and post- V-PDT images. The vasoconstriction of blood vessels in the DSWC model after V-PDT is directly quantified, which can avoid by the potential of generating new capillaries. The accuracy, sensitivity and specificity of the U-Net model for image segmentation are 90.64%, 80.12% and 92.83%, respectively. A significant difference in vasoconstriction between a control and a V-PDT group was observed. This new automatic protocol is well suitable for quantifying vasoconstriction in blood vessel image, which holds the potential application in V-PDT studies.