Zero-Shot Domain Adaptation in CT Segmentation by Filtered Back Projection Augmentation

TL;DR

This paper introduces FBPAug, a simple augmentation method in sinogram space that significantly improves zero-shot domain adaptation in CT segmentation by emulating different reconstruction kernels, enhancing prediction consistency.

Contribution

The paper presents FBPAug, a novel sinogram space augmentation technique that effectively addresses domain shift caused by reconstruction kernel variability in CT images.

Findings

Consistency increased from 0.54 to 0.92 with FBPAug.

FBPAug outperforms other augmentation methods in zero-shot domain adaptation.

No specific source or target domain data preparation needed.

Abstract

Domain shift is one of the most salient challenges in medical computer vision. Due to immense variability in scanners' parameters and imaging protocols, even images obtained from the same person and the same scanner could differ significantly. We address variability in computed tomography (CT) images caused by different convolution kernels used in the reconstruction process, the critical domain shift factor in CT. The choice of a convolution kernel affects pixels' granularity, image smoothness, and noise level. We analyze a dataset of paired CT images, where smooth and sharp images were reconstructed from the same sinograms with different kernels, thus providing identical anatomy but different style. Though identical predictions are desired, we show that the consistency, measured as the average Dice between predictions on pairs, is just 0.54. We propose Filtered Back-Projection Augmentation (FBPAug), a simple and surprisingly efficient approach to augment CT images in sinogram space emulating reconstruction with different kernels. We apply the proposed method in a zero-shot domain adaptation setup and show that the consistency boosts from 0.54 to 0.92 outperforming other augmentation approaches. Neither specific preparation of source domain data nor target domain data is required, so our publicly released FBPAug can be used as a plug-and-play module for zero-shot domain adaptation in any CT-based task.