Deep learning reconstruction of few-view X-ray CT measurements of mono-material objects with validation in additive manufacturing

The large acquisition times needed for high-quality XCT measurements remain a stumbling block for high-throughput inspection tasks. This paper therefore presents a deep learning reconstruction algorithm to improve the quality of fast, few-view XCT measurements. The proposed method is validated on both simulated and experimental XCT measurements of additively manufactured cranio-maxillofacial implants. The validation demonstrates a drastic reduction in noise and streaking artifacts associated with few-view acquisitions. Therefore, the potential to maintain high reconstruction quality while reducing acquisition times by more than one order of magnitude is confirmed.