Dynamic X-ray micotomography of microfibrous cellulose liquid foams using deep learning

Abstract Using X-ray microcomputed tomography (micro-CT) as a 3D microstructural analysis tool elucidates the time evolution of foam Plateau borders and nodes, providing unprecedented vision of foam dynamics. Deep learning facilitates the capability by allowing for quantifiable images to be collected at the time scale of five minutes. The stability mechanism of microfibrous cellulose is assessed demonstrating that trapping of fibers in the foam structure results in a critical concentration that marks an arrested foam. The mechanism of arrest is explained by 3D structural information extracted from micro-CT and confocal laser scanning microscopy images demonstrating the entanglement of cellulose fibers. A detailed analysis of the micro-CT data is provided to substantiate the quantitative nature of the digital images and assess advanced deep learning approaches. It is demonstrated that deep learning allows for the dynamic imaging of liquid foams whereas traditional data processing approaches cannot capture accurate geometrical information of foam structure.