Fluorescence imaging in the second near-infrared window (NIR-II) offers μm resolution blood vessel information noninvasively, which is crucial for the diagnosis and surgery treatment of some blood vessel-related diseases. However, only a few blood vessel segmentation algorithms have been done for the NIR-II images so far. Here, we proposed a vessel segmentation algorithm that used multi-scale enhancement and fractional differential to enhance capillaries, and then segmented vessels based on the blood vessels' tubular characteristics. Experimental results showed that this method could effectively suppress the point and lump tissue noise influence during vascular segmentation. The accuracy of vessel identification by other algorithms dropped below 30%, while our algorithm still achieved an accuracy of around 50% in deep vessel segmentation experiments with the 6.5 mm Intralipid. So it had the advantage of accurately detecting deep and dim blood capillaries. Meanwhile, the vascular density quantization algorithm had been successfully applied to the mice's ischemic stroke evaluations for the first time. In addition, this algorithm can provide the quantified vessel features under physiological or pathological conditions, which could be used to accurately evaluate the stroke drugs' therapeutic effect in the future.
Keywords: Fractional differential; Mice's ischemic stroke evaluations; Multi-scale enhancement; Tubular characteristics; Vascular segmentation.
Copyright © 2022 Elsevier Ltd. All rights reserved.