Ossama Mahmoud, GH Janssen and Mahmoud R. El-Sakka, "Machine-Learning-Based Functional Microcirculation Analysis", Annual Conference on Innovative Applications of Artificial Intelligence, IAAI'2020, pp. I3326-I3331, February 2020, New York, New York, USA.
Abstract
Analysis of microcirculation is an important clinical and re-search task. Functional analysis of the microcirculation allows researchers to understand how blood flowing in a tissues’ smallest vessels affects disease progression, organ function, and overall health. Current methods of manual analysis of mi-crocirculation are tedious and time-consuming, limiting the quick turnover of results. There has been limited research on automating functional analysis of microcirculation. As such, in this paper, we propose a two-step machine-learning-based algorithm to functionally assess microcirculation videos. The first step uses a modified vessel segmentation algorithm to extract the location of vessel-like structures. While the second step uses a 3D-CNN to assess whether the vessel-like struc-tures contained flowing blood. To our knowledge, this is the first application of machine learning for functional analysis of microcirculation. We use real-world labelled microcirculation videos to train and test our algorithm and assess its perfor-mance. More precisely, we demonstrate that our two-step al-gorithm can efficiently analyze real data with high accuracy (90%).