Stereo laser speckle dissimilarity analysis using self-organizing maps

Abstract

Imaging methods based on the physical phenomenon of speckle have been gaining relevance (mostly due to technological improvements in sensors and cameras) in several areas of science, particularly in the field of Medicine. Specifically, the use of speckle laser technology has proved to be useful in the characterization of post-hypoxia perfusion states, either in the cortex of animal models or in skin perfusion experiments. This dynamic information, if added to the morphology, results in an imagery modality that allows the full (static and dynamic) characterization of the scattered surface. In this work, a stereo vision system was adopted during a laser speckle acquisition of skin surface, in order to perform a dissimilarity (i.e. left vs. right) analysis based on activity descriptors in the context of a Post Occlusive Reactive Hyperemia (PORH) test. Additionally, to perform inter-view segmentation, a registration procedure is also proposed based on high entropy regions. The results reveal that among five commonly used activity descriptors, the Shannon Wavelet Entropy (SWE) is the most consistent in characterizing the transition from an occlusive state to a hyperemic state, thus indicating its use as a dissimilarity descriptor in stereo acquisitions to identify physiological states related with reduced perfusion and its recovery.