Model-based deep learning framework for accelerated optical projection tomography

OBANDO, MARCOS; BASSI, ANDREA; DUCROS, NICOLAS; MATO, GERMÁN; CORREIA, TERESA M.

Scientific Reports

Springer

Año: 2023 vol. 13

In this work, we propose a model-based deep learning reconstruction algorithm for optical projectiontomography (ToMoDL), to greatly reduce acquisition and reconstruction times. The proposed methoditerates over a data consistency step and an image domain artefact removal step achieved by aconvolutional neural network. A preprocessing stage is also included to avoid potential misalignmentsbetween the sample center of rotation and the detector. The algorithm is trained using a databaseof wild-type zebrafish (Danio rerio) at different stages of development to minimise the mean squareerror for a fixed number of iterations. Using a cross-validation scheme, we compare the results toother reconstruction methods, such as filtered backprojection, compressed sensing and a direct deeplearning method where the pseudo-inverse solution is corrected by a U-Net. The proposed methodperforms equally well or better than the alternatives. For a highly reduced number of projections, onlythe U-Net method provides images comparable to those obtained with ToMoDL. However, ToMoDLhas a much better performance if the amount of data available for training is limited, given that thenumber of network trainable parameters is smaller.