Field-of-View Extension in Dental Cone-Beam CT via Implicit Neural Representations and Diffusion Model-Based Refinement

Problem

The paper addresses the issue of truncated field of view (FOV) in dental cone-beam computed tomography (CBCT) systems, which can lead to incomplete imaging of anatomical structures. This problem is particularly critical in dental applications where accurate visualization is essential for diagnosis and treatment planning. The work is presented as a preprint and has not undergone peer review.

Method

The authors propose a method that leverages implicit neural representations (INR) to estimate the missing parts of truncated projection data. The approach consists of two main components:

• Implicit Neural Representation (INR): This component is responsible for inferring the missing data from the truncated projections, effectively filling in gaps in the imaging data.
• Iterative Reconstruction: Following the INR, an iterative reconstruction process generates a secondary volumetric image that enhances anatomical consistency, ensuring that the reconstructed image aligns more closely with expected anatomical structures.
• Diffusion Model: To further improve image quality, a diffusion model is employed, which refines the output of the iterative reconstruction, enhancing the overall visual fidelity of the images produced.

Results

The proposed method demonstrates effectiveness in several key areas:

• Reduction of Truncation Artifacts: The method effectively reduces artifacts associated with truncation, although no specific quantitative results are reported.
• Improvement in Reconstruction of Structures Beyond Original FOV: The approach shows effectiveness in reconstructing anatomical structures that lie beyond the original FOV, again without specific quantitative metrics provided.
• Image Quality Enhancement: The enhancement of image quality is noted as effective, but no numerical comparisons to baseline methods are reported.

Limitations

The authors do not report any limitations in their study. However, the lack of quantitative results and comparisons to established baselines may limit the ability to fully assess the method's performance relative to existing techniques.

Why it matters

This work has significant implications for the field of dental imaging, particularly in improving the diagnostic capabilities of CBCT systems. By addressing the limitations of truncated FOV, the proposed method could lead to better patient outcomes through more accurate imaging. Furthermore, the integration of implicit neural representations and diffusion models may inspire future research in other imaging modalities, potentially enhancing the quality and reliability of medical imaging technologies.