Towards retrieving dispersion profiles using quantum-mimic optical coherence tomography and machine learning

Abstract

Artefacts in quantum-mimic optical coherence tomography are considered detrimental because they scramble the images even for the simplest objects. They are a side effect of autocorrelation, which is used in the quantum entanglement mimicking algorithm behind this method. Interestingly, the autocorrelation imprints certain characteristics onto an artefact – it makes its shape and characteristics depend on the amount of dispersion exhibited by the layer that artefact corresponds to. In our method, a neural network learns the unique relationship between the artefacts’ shape and GVD, and consequently, it is able to provide a good qualitative representation of object’s dispersion profile for never-seen-before data: computer-generated single dispersive layers and experimental pieces of glass. We show that the autocorrelation peaks – additional peaks in the A-scan appearing due to the interference of light reflected from the object – affect the GVD profiles. Through relevant calculations, simulations and experimental testing, the mechanism leading to the observed GVD changes is identified and explained. Finally, the network performance is tested in the presence of noise in the data and with the experimental data representing single layers of quartz, sapphire and BK7.

Extracting Group Velocity Dispersion values using quantum-mimic Optical Coherence Tomography and Machine Learning

“…We cast this problem as a Machine Learning problem with an FFT stack as an input and a dispersion profile as an output. Our neural network is based on a modified VGG-16 architecture 17 which we showed is capable of interpreting layer-specific behaviour and output a depth-resolved dispersion profile of the objects 16 . In the modified architecture, batch normalization is added after each convolutional layer and residual blocks are used to sum up the outputs of the pooling layer and convolutional layers.…”

“…It was shown that Machine Learning 16 - when combined with Qm-OCT - enables to estimate qualitatively GVD value distribution, i.e. a dispersion profile, within an A-scan.…”