Machine learning assisted blood vessel segmentation in laser speckle imaging (Conference Presentation)

We are introducing an application of a machine learning approach for express analysis of Laser Speckle (LS) images. This application can be utilized for real-time visualisation of vascular beds in vivo. This research used Waikato Environment for Knowledge Analysis (Weka) integrated with Fiji/ImageJ software. A large number of acquired LS images are averaged, then used as references for training Weka classifiers. Subsequently, a bundle of these Weka classifiers are produced. We defined the minimal number of raw LS images based on a phenomenological model to minimize the time needed for LS data analysis. Finally, a new perceptually uniform color coding approach is developed for highlighting targeted blood vessels. The developed LS processing approach is especially convenient, because of its high potential for blood vessel visualisation during real-time intraoperative vascular imaging in vivo.