Dual Branch Prior-SegNet: CNN for Interventional CBCT using Planning Scan and Auxiliary Segmentation Loss

TL;DR

This paper introduces Dual Branch Prior-SegNet, a CNN that leverages planning scans and auxiliary segmentation to improve sparse view interventional CBCT reconstruction, demonstrating significant performance gains and robustness to misalignments.

Contribution

It extends the Dual Branch Prior-Net by adding segmentation guidance and robustness to misalignments, enhancing CBCT reconstruction quality in interventional imaging.

Findings

Outperforms other models by >2.8dB PSNR

Maintains robustness with up to +-5.5deg rotations

Effectively incorporates planning scans and segmentation loss

Abstract

This paper proposes an extension to the Dual Branch Prior-Net for sparse view interventional CBCT reconstruction incorporating a high quality planning scan. An additional head learns to segment interventional instruments and thus guides the reconstruction task. The prior scans are misaligned by up to +-5deg in-plane during training. Experiments show that the proposed model, Dual Branch Prior-SegNet, significantly outperforms any other evaluated model by >2.8dB PSNR. It also stays robust wrt. rotations of up to +-5.5deg.