Breast tumor segmentation in high resolution x-ray phase contrast analyzer based computed tomography.

Area of Science:

• Medical Imaging
• Radiology
• Image Analysis

Background:

• Phase contrast computed tomography (PC-CT) offers superior breast tumor visualization compared to mammography at equivalent radiation doses.
• Current segmentation techniques do not leverage the unique properties of the phase contrast signal for breast tumor analysis.
• High-resolution PC-CT images of excised human organs present significant data challenges for segmentation.

Purpose of the Study:

• To introduce a novel mathematical framework for human-guided breast tumor segmentation tailored to phase contrast signals.
• To apply and evaluate this framework on high-resolution, multi-gigabyte PC-CT datasets of excised human breasts.

Main Methods:

• Development of a segmentation procedure utilizing the viscous watershed transform.
• Application of the viscous watershed transform to analyzer-based phase contrast images.
• Demonstration of tumor segmentation in two full human breasts as a practical application.

Main Results:

• Precise and accurate identification of tumor boundaries was achieved.
• Segmentation results were validated through independent manual contouring by four experienced radiologists.

Conclusions:

• The viscous watershed transform effectively segments tumors in high-resolution PC-CT breast images.
• Integrating this segmentation procedure with PC-CT enhances diagnostic value by combining detailed morphological information with precise boundary identification.
• This multistep approach holds significant potential for future medical diagnostic applications in breast imaging.