Direct reconstruction of anatomical change in low-dose lung nodule surveillance

Area of Science:

• Medical Imaging
• Radiology
• Image Reconstruction

Background:

• Sequential imaging studies contain valuable prior information for enhancing current image reconstruction.
• Prior-image-based reconstruction (PIBR) methods, like reconstruction of difference (RoD), improve image quality, especially at low radiation doses.
• Challenges in PIBR include accurate registration between scans and selecting appropriate regularization parameters.

Purpose of the Study:

• To address the challenges of registration ambiguity and regularization parameter selection in PIBR.
• To develop a modified RoD framework for improved anatomical change detection in sequential CT scans.
• To apply the modified framework to the clinical problem of lung nodule surveillance.

Main Methods:

• Developed a modified deformable registration approach enforcing locally smooth/rigid registration around change regions.
• Extended analytic expressions to relate reconstructed contrast to regularization parameters and system dependencies.
• Validated the approach using digital phantoms and clinical projection data.

Main Results:

• The modified framework successfully separated patient motion from anatomical changes.
• Performance was evaluated based on the size of the registration region of interest and X-ray exposure.
• Demonstrated the efficacy of the enhanced PIBR for highlighting anatomical changes.

Conclusions:

• The modified PIBR framework effectively distinguishes patient motion from anatomical changes.
• This approach directly highlights anatomical changes crucial for lung nodule surveillance.
• The method shows promise for improving diagnostic accuracy in follow-up imaging studies.