Artifact Reduction in Undersampled 3D Cone-Beam CTs using a Hybrid 2D-3D CNN Framework

TL;DR

This work tackles artifacts in undersampled cone-beam CT by introducing a hybrid 2D-3D CNN framework that first uses a slice-wise 2D U-Net to remove artifacts and then stacks the resulting feature maps to form a 3D representation processed by a 3D decoder. The two-stage approach combines the efficiency of 2D processing with the volumetric context of 3D modeling, improving inter-slice consistency while keeping computation reasonable. Quantitative results show substantial improvements over sparse-view reconstructions, with the 2D path achieving slightly higher PSNR/SSIM and the 3D decoder enhancing volumetric fidelity; visual results also confirm better axial-to-coronal/sagittal consistency. The method is evaluated on RSNA pulmonary embolism data, and code is available on GitHub, highlighting its practical potential for fast, high-quality 3D CT post-processing in clinical pipelines.

Abstract

Undersampled CT volumes minimize acquisition time and radiation exposure but introduce artifacts degrading image quality and diagnostic utility. Reducing these artifacts is critical for high-quality imaging. We propose a computationally efficient hybrid deep-learning framework that combines the strengths of 2D and 3D models. First, a 2D U-Net operates on individual slices of undersampled CT volumes to extract feature maps. These slice-wise feature maps are then stacked across the volume and used as input to a 3D decoder, which utilizes contextual information across slices to predict an artifact-free 3D CT volume. The proposed two-stage approach balances the computational efficiency of 2D processing with the volumetric consistency provided by 3D modeling. The results show substantial improvements in inter-slice consistency in coronal and sagittal direction with low computational overhead. This hybrid framework presents a robust and efficient solution for high-quality 3D CT image post-processing. The code of this project can be found on github: https://github.com/J-3TO/2D-3DCNN_sparseview/.

Figures (2)

• Figure 1: Overview of the 2D-3D training pipeline. A) First, a 2D U-Net is trained on 2D axial slices to remove artifacts. B) The encoder of the trained U-Net is used to extract features at different levels of an entire CT volume by looping through $N$ axial slices. Subsequently, the extracted features from a volume are stacked together, effectively transforming the 2D feature maps into 3D feature maps. C) The extracted 3D feature maps are used as input along with the sparse volume to train a 3D decoder, which returns an artifact-reduced CT volume.
• Figure 2: An axial, coronal, and sagittal slice of the full-view reconstruction from the test split (first row), 128-view reconstruction (second row) with artifact reduction by the 2D U-Net (third row) and by the Hybrid 2D-3D CNN framework (last row), respectively. Images are in the mediastinum window (Width:350 HU, Level:50 HU). The dotted lines indicate the location of the different views. All inserts are 75x75 pixels.