Augmented active surface model for the recovery of small structures in CT

Area of Science:

• Medical imaging
• Image processing
• Computational anatomy

Background:

• Conventional active contour methods struggle with small objects in low-resolution, high-noise medical images.
• Standard gradient magnitude-based energy functions perform poorly at small object scales.
• The point spread function (PSF) significantly impacts image quality and structure recovery.

Purpose of the Study:

• To develop an augmented active surface model for improved recovery of small structures in challenging imaging conditions.
• To address the limitations of traditional methods in low-resolution and high-noise computed tomography (CT) data.
• To incorporate prior knowledge, including the PSF and CT number constancy, into the active surface model.

Main Methods:

• Devised an augmented active surface model incorporating PSF and prior knowledge of CT numbers.
• Evaluated model performance on clinical computed tomography (CT) data.
• Compared results against ground truth data acquired using micro-CT.

Main Results:

• Demonstrated that conventional active contour methods yield non-optimal results for small structures.
• Showcased the inadequacy of blind gradient magnitude-based energy at small object scales.
• Validated the critical role of accounting for the point spread function (PSF) in image recovery.

Conclusions:

• The proposed augmented active surface model enhances the recovery of small structures in low-resolution, high-noise CT imaging.
• Incorporating PSF and CT number assumptions improves model robustness and accuracy.
• The findings highlight the importance of tailored regularization and prior knowledge in medical image analysis.