Statistical iterative pattern generation in volumetric additive manufacturing based on ML-EM

Volumetric Additive Manufacturing (VAM) is a novel promising technology for photopolymerization-based 3D printing, which forms 3D objects inside photopolymer resin by rotative exposure instead of traditional layer-by-layer printing operation. Computed Axial Lithography (CAL) is a widely used modality for VAM which first generates projection pattern set from a target 3D model and then back projects the patterns through DLP device to reconstruct desired light dose distribution inside the resin. The algorithms for CAL pattern generation applied in recent developments are inspired by computed tomography. Though tomographic-based methods have performed well in recent studies, they still subject to some instinctive shortcomings such as negatives in patterns and vulnerable reconstruction to the absorption of resin. We propose a more robust approach for pattern generation using statistical iterative optimization based on maximum likelihood-expectation maximization (ML-EM). We also introduce the way to implement our pattern generation algorithm and reconstruction simulation on GPU efficiently. The reconstruction results show a high fidelity for various printing geometries and flexibility to the resin attenuation with our method in our simulation work. These features demonstrate the better performance and higher efficiency of our method and show a wide application prospect in VAM.