Adaptive processing for noise attenuation in laser speckle contrast imaging

Blood vessel visualization is an essential task to treat and evaluate diseases such as port-wine stain. Laser Speckle Contrast Imaging (LSCI) have applications in the analysis of the microvasculature. However, it is often limited to superficial depths because the tissue among skin and microvasculature introduces noise in the image. To analyze microvasculature, traditional LSCI methods compute a Contrast Image (CI) by using a shifting window of fixed size and shape, which is inadequate in images with structures different types of morphologies in it, as happens in LSCI. This work aims to reduce the noise in the CIs to improve the visualization of blood vessels at high depths (> 300 μ m).

The proposed method processes the CIs with analysis windows that change their size and shape for each pixel to compute the contrast representation with pixels more representatives to the region.

We performed experiments varying the depth of the blood vessels, the number of frames required to compute the representation, and the blood flow in the blood vessel. We looked for an improvement in the Contrast to Noise Ratio (CNR) in the periphery of the blood vessels using an analysis of variance. Finding that the adaptive processing of the contrast images allows a significant noise attenuation, translated into a better visualization of blood vessels. An average CNR of 2.62 ± 1 and 5.26 ± 1.7 was reached for in-vitro and in-vivo tests respectively, which is higher in comparison with traditional LSCI approaches.

The results, backed by the measured CNR, obtained a noise reduction in the CIs, this means a better temporal and spatial resolution. The proposed awK method can obtain an image with better quality than the state-of-the-art methods using fewer frames.