Reconstructing 3D Voxels from Ultra-sparse-view 2D X-ray Images Using a Dual-domain Balanced Network

Bodong Bi, Bo Yang†, Zhixin Liu, Junmin Lyu, Feng Bao, Chao Liu Measurement and Control

#Medical Imaging #Deep Learning #Inverse Problems #Reconstruction

Summary

Cone-beam CT (CBCT) reconstruction degrades sharply as the number of X-ray views drops, yet every view adds radiation dose. This work develops a U-Net framework supporting 2–4-view inputs with dual-domain optimization, reducing radiation dose by approximately 50% and achieving SSIM > 0.94.

Publication

Published in Measurement and Control. DOI: 10.1177/00202940261493222.