Related Experiment Video
Updated: Jun 11, 2025
10:22
Fabrication and Characterization of Optical Tissue Phantoms Containing Macrostructure
Published on: February 12, 2018
10.6K
Novel Perlin-based Phantoms Using 3D Models of Compressed Breast Shape and Fractal Noise
João P V Teixeira1, Telmo M Silva Filho1, Thaís G do Rêgo1
1Department of Computer Science, Federal University of Paraiba, João Pessoa, Brazil.
Summary
Virtual clinical trials using Perlin phantoms optimize digital breast tomosynthesis (DBT) systems. Custom acquisition geometries improve DBT accuracy by reducing breast volume overestimation and segmentation errors.
Area of Science:
- Medical Imaging
- Computational Phantoms
- Breast Imaging Technology
Background:
- Virtual clinical trials (VCTs) are crucial for evaluating digital breast tomosynthesis (DBT) systems.
- Realistic breast phantoms are essential for characterizing lesions and assessing cancer masking risks in VCTs.
Purpose of the Study:
- Introduce Perlin-based phantoms for optimizing the acquisition geometry of a novel DBT prototype.
- Evaluate the impact of different acquisition geometries on DBT performance using simulated breast phantoms.
Main Methods:
- Developed Perlin-based phantoms using a GPU implementation of the Perlin-CuPy library.
- Simulated 240 3D breast models with varying thickness and dimensions under mammographic compression.
- Generated DBT projections and reconstructions for two acquisition geometries to compare performance.
Main Results:
- Breast volume estimates in DBT are improved with posterior-anterior x-ray source motion.
- DBT tends to overestimate breast volume, with significant variations based on acquisition geometry.
- Segmentation errors are more pronounced in thicker and larger simulated breasts.
Conclusions:
- Custom acquisition geometries can enhance the performance and accuracy of DBT systems.
- Perlin phantoms effectively identify limitations in acquisition geometries and aid in DBT prototype optimization.
