Automated detection of air embolism in OCT contrast imaging: Anisotropic diffusion and active contour based approach

Embolism can be a cause of life threatening situation for which early detection and diagnosis is of major importance. This work describes an automatic approach for air embolism detection and measurement of different morphological features of the embolus using OCT contrast imaging technique. Firstly, the channel has been segmented through morphological processing. Manually selecting the initial contour for active contour (AC) technique is time consuming. To overcome this, anisotropic diffusion (AD) is implemented to automatically select the initial contour prior to AC. A snake based AC is executed to segment out the embolus. The proposed emboli segmentation mechanism has been compared with other segmentation techniques and it has been observed that it can efficiently extract the embolus with high segmentation accuracy (92%-94%) and reduced computational time. Different morphological descriptors showing the shape properties of the embolus have been computed to perform the shape analysis of it and measuring the criticality of the blockage area. It has been experimented that this method can also track multiple emboli flowing through the microchannel, thereby facilitating the study of contrast imaging in air embolism detection.