AAPM DL-Sparse-View CT Challenge Submission Report: Designing an Iterative Network for Fanbeam-CT with Unknown Geometry

TL;DR

This paper presents a data-driven iterative neural network approach for limited-view fanbeam CT reconstruction, including a geometric calibration step, to recover breast phantom images without explicit knowledge of the forward model.

Contribution

The authors introduce a novel two-step, end-to-end neural network method that estimates unknown fanbeam geometry and reconstructs images from limited-view data.

Findings

Achieved near-exact image reconstruction from limited-view measurements.

Developed a geometric calibration method for unknown fanbeam geometry.

Demonstrated robustness in limited-view CT scenarios.

Abstract

This report is dedicated to a short motivation and description of our contribution to the AAPM DL-Sparse-View CT Challenge (team name: "robust-and-stable"). The task is to recover breast model phantom images from limited view fanbeam measurements using data-driven reconstruction techniques. The challenge is distinctive in the sense that participants are provided with a collection of ground truth images and their noiseless, subsampled sinograms (as well as the associated limited view filtered backprojection images), but not with the actual forward model. Therefore, our approach first estimates the fanbeam geometry in a data-driven geometric calibration step. In a subsequent two-step procedure, we design an iterative end-to-end network that enables the computation of near-exact solutions.