Evaluation of the Morphological and Biological Functions of Vascularized Microphysiological Systems with Supervised Machine Learning

AngioMT: A MATLAB based 2D image-to-physics tool to predict oxygen transport in vascularized microphysiological systems

“…affect the spatial distribution and functioning of microvascular networks [ 10 , 24 , 25 ]. Therefore, we analyzed over 500 images of vascular networks with different experimental conditions that have been previously published by our group [ 11 , 26 ] and examined if AngioMT was sensitive to the extent and morphological variability of the vascular networks ( Fig 2A and 2B ) . For example, for conditions that resulted in extremely low vascularization with disconnected networks, AngioMT predicted a poor oxygen delivery to the tissue ( Fig 2A ; first column).…”

“…Interestingly, at higher values of vessel coverage, the vessel and tissue oxygenation reached saturation, suggesting that vessel coverage values greater than 40% resulted in well-distributed networks within each image and received sufficient oxygen from the inlets. Overall, our program revealed high sensitivity of computed oxygenation to a morphological marker often used in scientific literature [ 11 , 18 ] to evaluate the performance of a vascular network ( S2 Table ).…”

“…The spatial assessment of oxygen tranposrt offered by AngioMT bolster its application as a tool for “scoring” networks. AngioMT was recently used by us in conjunction with other physiological vessel metrics, for example vessel length, area, segment diameter, number of branchpoints, to create more physiologically relevant scoring metrics [ 11 ]. We have recently also demonstrated another application where data from AngioMT served as the ground truth data to create an anrtificial-intelligence based program to evaluate vascular networks in biological systems [ 26 ].…”

“…Therefore, evaluation of the performance of the microvasculature is critical in vascular biology as impaired transport in microvessels is the primary cause of organ failures and necrosis [ 8 , 9 ]. Specifically in the context of engineered vascular systems like organ-chips and vascularized organoids, the ability to measure comparative performances of different vascular conditions is essential [ 10 , 11 ]. Measuring the transport of dissolved nutrients, growth factors, cytokines, drugs etc.…”

Machine learning chained neural network analysis of oxygen transport amplifies the physiological relevance of vascularized microphysiological systems

“…Recently, we applied several incrementally complex supervised statistical and machine learning algorithms and showed that many of these metrics are inconsistent with each other, user-dependent, and poorly correlate to the oxygen transport phenomena of vascular networks when applied independently or in combination. 9 While some groups have also directly measured perfusion capacity of networks with fluorescent beads, 10,11 this experimental strategy can be extremely difficult in practice as some capillaries might be smaller than the beads (tracer particles are typically orders of magnitude larger than oxygen molecules).…”