Tomographic Foundation Model -- FORCE: Flow-Oriented Reconstruction Conditioning Engine
This work addresses the need for improved CT reconstruction techniques in medical imaging, particularly for low-dose screening and sparse-view scanning, but it appears to be an incremental advancement over existing unsupervised methods.
The paper tackles the problem of severe noise and artifacts in CT image reconstruction under challenging clinical scenarios by integrating data fidelity with a generative AI model, resulting in a novel framework called FORCE that shows superior performance in various CT imaging tasks.
Computed tomography (CT) is a major medical imaging modality. Clinical CT scenarios, such as low-dose screening, sparse-view scanning, and metal implants, often lead to severe noise and artifacts in reconstructed images, requiring improved reconstruction techniques. The introduction of deep learning has significantly advanced CT image reconstruction. However, obtaining paired training data remains rather challenging due to patient motion and other constraints. Although deep learning methods can still perform well with approximately paired data, they inherently carry the risk of hallucination due to data inconsistencies and model instability. In this paper, we integrate the data fidelity with the state-of-the-art generative AI model, referred to as the Poisson flow generative model (PFGM) with a generalized version PFGM++, and propose a novel CT framework: Flow-Oriented Reconstruction Conditioning Engine (FORCE). In our experiments, the proposed method shows superior performance in various CT imaging tasks, outperforming existing unsupervised reconstruction approaches.