Segmentation of tomographic data without image reconstruction

Area of Science:

• Medical imaging
• Computational geometry
• Image processing

Background:

• Traditional computed tomography (CT) requires full image reconstruction, which is computationally intensive.
• Reconstructing internal object details is often unnecessary when only boundary information is needed.

Purpose of the Study:

• To introduce Geometric Tomography (GT) as a method for reconstructing object boundaries directly from tomographic projection data.
• To demonstrate that GT reduces computational load compared to conventional CT by avoiding image reconstruction.

Main Methods:

• Segmentation of object boundaries directly from raw scanner data (sinogram).
• Geometric transformation of segmented shapes into reconstructed shapes in standard space.
• Application to both synthesized and real-world tomographic datasets.

Main Results:

• Successful reconstruction of external and internal object boundaries without full image reconstruction.
• Achieved sub-pixel positioning accuracy for reconstructed boundaries.
• Demonstrated efficient 'on-the-fly' processing and real-time boundary tracking.

Conclusions:

• Geometric Tomography (GT) provides a computationally efficient alternative to traditional CT for boundary reconstruction.
• GT is particularly effective for sparse, high-contrast objects like bones and blood vessels.
• The method enables real-time analysis and tracking of object boundaries in medical imaging applications.