Fast and accurate sparse-view CBCT reconstruction using meta-learned neural attenuation field and hash-encoding regularization

TL;DR

A fast and accurate sparse-view CBCT reconstruction (FACT) method to provide better reconstruction quality and faster optimization speed in the minimal number of view acquisitions and a new regularization technique is utilized to reconstruct the details of an anatomical structure.

Abstract

Cone beam computed tomography (CBCT) is an emerging medical imaging technique to visualize the internal anatomical structures of patients. During a CBCT scan, several projection images of different angles or views are collectively utilized to reconstruct a tomographic image. However, reducing the number of projections in a CBCT scan while preserving the quality of a reconstructed image is challenging due to the nature of an ill-posed inverse problem. Recently, a neural attenuation field (NAF) method was proposed by adopting a neural radiance field algorithm as a new way for CBCT reconstruction, demonstrating fast and promising results using only 50 views. However, decreasing the number of projections is still preferable to reduce potential radiation exposure, and a faster reconstruction time is required considering a typical scan time. In this work, we propose a fast and accurate sparse-view CBCT reconstruction (FACT) method to provide better reconstruction quality and faster optimization speed in the minimal number of view acquisitions ($<$ 50 views). In the FACT method, we meta-trained a neural network and a hash-encoder using a few scans (= 15), and a new regularization technique is utilized to reconstruct the details of an anatomical structure. In conclusion, we have shown that the FACT method produced better, and faster reconstruction results over the other conventional algorithms based on CBCT scans of different body parts (chest, head, and abdomen) and CT vendors (Siemens, Phillips, and GE).

Figures (10)

• Figure 1: Illustration of the FACT reconstruction with the meta-initialization and hash-encoding regularization. (a) By meta-learning the CBCT reconstruction functions (i.e., meta-initialization), the FACT reconstruction quickly optimizes each target scan to a specific CBCT image (dotted line vs. solid line). (b) In the FACT reconstruction, the hash-encoder, which includes feature vectors of multiresolution 3D grid points, generates feature embeddings (hash-encoded feature) of the query points (e.g., red point $p_2$) using tri-linear interpolation for each resolution (interpolated features). In the FACT method, a novel regularization technique (solid line and red box) for the hash-encoder is incorporated, which masks out the interpolated features depending on the optimization epochs, resulting in the masked hash-encoded feature.
• Figure 1: Example CBCT reconstruction results of 30 views for the three CT vendors (Siemens, Phillips, and GE).
• Figure 2: 3D SSIM and 3D PSNR values according to the change in the number of input views (50, 40, 30, 20, and 10 views) for the SART, ASD-POCS, NAF, and FACT methods. Regardless of the number of views, the FACT method outperformed the others regarding SSIM and PSNR. Especially in the graph of SSIM, FACT revealed more gains as the number of views decreased.
• Figure 2: CBCT optimization graphs (3D SSIM or 3D PSNR vs. optimization epochs) between the NAF and FACT methods for the head and abdomen CBCTs.
• Figure 3: Example CBCT reconstruction results between the different CBCT reconstruction methods (FDK, SART, ASD-POCS, NAF, and FACT). In the zoomed-in plots (yellow boxes), the FACT method demonstrated better image quality, successfully reconstructing anatomical structures and suppressing artifacts than the other methods (e.g., results of NAF vs. FACT in 30 views).