THUBreast: an open-source breast phantom generation software for x-ray imaging and dosimetry

Area of Science:

• Medical Imaging Physics
• Computational Anatomy
• Radiological Dosimetry

Background:

• Realistic breast phantoms are crucial for virtual clinical trials in breast x-ray imaging.
• Existing phantoms often lack accurate glandular distribution, impacting imaging and dosimetry realism.
• This limits the reliability of virtual trials for evaluating imaging techniques and radiation risk.

Purpose of the Study:

• To develop novel breast phantoms that incorporate realistic glandular distribution.
• To create an open-source software tool (THUBreast) for generating these advanced phantoms.
• To validate the performance of THUBreast phantoms in terms of imaging texture and radiation dosimetry.

Main Methods:

• Developed THUBreast, an open-source software for generating breast phantoms with realistic structural texture and glandular distribution.
• Created three phantom groups: THUBreast, patient-based, and homogeneous.
• Simulated irradiation and calculated power-law exponents (β) for texture realism and mean glandular dose (Dg) for radiation risk using MC-GPU.

Main Results:

• THUBreast phantoms showed mean glandular dose (Dg) values with a 4% average deviation compared to patient-based phantoms.
• Homogeneous phantoms overestimated average Dg by 30% compared to patient-based phantoms.
• THUBreast phantoms yielded a mean power-law exponent (β) of 2.92 ± 0.08, consistent with prior studies.

Conclusions:

• THUBreast phantoms demonstrate good performance in both imaging texture realism and radiation dosimetry.
• These phantoms offer a more accurate representation of breast anatomy than homogeneous models.
• THUBreast provides a valuable tool for enhancing the evaluation of image quality and radiation risk in virtual breast x-ray imaging trials.