Hao Zhang

Ph.D.

Johns Hopkins University

Postdoctoral Fellow

NLST (Learn more about this study)

NLST-329

Aug 3, 2017

Ultra-Low-Dose Lung Nodule CT Surveillance Using Prior-Image-Based Reconstruction

Prior-image-based reconstruction (PIBR) methods, which incorporate a high-quality patient-specific prior image into the reconstruction of subsequent low-dose CT acquisitions, have demonstrated great potential to dramatically reduce data fidelity requirements while maintaining or improving image quality. However, one challenge with the PIBR methods is in the selection of the prior image regularization parameter which controls the balance between information from current measurements and in-formation from the prior image. Too little prior information yields few improvements for PIBR, and too much prior information can lead to PIBR results too similar to the prior image obscuring or misrepresenting features in the reconstruction. While exhaustive parameter searches can be used to establish prior image regularization strength, this process can be time consuming (involving a series of iterative reconstructions) and particular settings may not generalize for different acquisition protocols, anatomical sites, patient sizes, etc.

We have proposed a novel approach for prospective analysis of PIBR. The methodology can be used to determine prior image regularization strength to admit specific anatomical changes without the need to perform iterative reconstructions in advance. The experimental results using a digital phantom show that the proposed analytical approach has high accuracy in predicting the admission of specific anatomical features, allowing for prospective determination of the prior image regularization parameter. We are aiming to validate its feasibility in clinical lung data.