Towards Improved Control of Breast Density in Simulated Mammograms via Perlin Noise Parameterization
Tomic, Hanna LU ; Dustler, Magnus LU
and Bakic, Predrag R.
LU
(2026)
Medical Imaging 2026: Physics of Medical Imaging
In Progress in Biomedical Optics and Imaging - Proceedings of SPIE
13924.
- Abstract
Virtual Imaging Trials (VITs) provide a reproducible, cost-effective, and ethical approach for testing medical imaging technologies. In breast imaging, realistic phantoms across density levels are crucial, as density affects cancer risk assessment and imaging performance. We have previously demonstrated the use of Perlin noise to simulate breast parenchyma. This study establishes a systematic mapping between Perlin noise parameters and breast density, enabling automated phantom generation with user-specified density targets. Density control was achieved by linking Perlin noise threshold values (TVs) to the resulting volumetric breast density through a set of algorithms. Using the STELLA-R simulation pipeline, multiple Perlin noise... (More)
Virtual Imaging Trials (VITs) provide a reproducible, cost-effective, and ethical approach for testing medical imaging technologies. In breast imaging, realistic phantoms across density levels are crucial, as density affects cancer risk assessment and imaging performance. We have previously demonstrated the use of Perlin noise to simulate breast parenchyma. This study establishes a systematic mapping between Perlin noise parameters and breast density, enabling automated phantom generation with user-specified density targets. Density control was achieved by linking Perlin noise threshold values (TVs) to the resulting volumetric breast density through a set of algorithms. Using the STELLA-R simulation pipeline, multiple Perlin noise volumes were generated, superimposed, and thresholded. Because the TV influences the size and spatial distribution of dense structures, we established a quantitative mapping between TV and breast density. This mapping was then used to build a prediction model that estimates the TV required to achieve a specified density. By minimizing the difference between predicted and target densities, the algorithm determines TVs that reproduce desired breast densities in synthetic phantoms. The mapping revealed a monotonic relationship between TV and volumetric density, with variability introduced by the stochastic nature of Perlin noise. The mean density error between target and simulated output was 35%. Although this level of error indicates the need for further refinement, the proposed framework provides a reproducible and scalable foundation for density-controlled phantom customization in future virtual imaging studies. This approach supports imaging system evaluation and AI training across clinically relevant density ranges.
(Less)
- author
- Tomic, Hanna
LU
; Dustler, Magnus
LU
and Bakic, Predrag R.
LU
- organization
- publishing date
- 2026-04-02
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- keywords
- breast density, Breast phantoms, breast tumors, computer simulation, temporal changes in anatomy, virtual clinical trials
- host publication
- Medical Imaging 2026 : Physics of Medical Imaging - Physics of Medical Imaging
- series title
- Progress in Biomedical Optics and Imaging - Proceedings of SPIE
- editor
- Ganguly, Arundhuti ; Li, Ke and Abbaszadeh, Shiva
- volume
- 13924
- article number
- 139243T
- publisher
- SPIE
- conference name
- Medical Imaging 2026: Physics of Medical Imaging
- conference location
- Vancouver, Canada
- conference dates
- 2025-02-15 - 2025-02-19
- external identifiers
-
- scopus:105039292108
- ISSN
- 1605-7422
- 2410-9045
- ISBN
- 9781510697850
- DOI
- 10.1117/12.3088024
- language
- English
- LU publication?
- yes
- id
- 429234e5-a126-41e2-85bf-f3d239b7f4be
- date added to LUP
- 2026-07-07 15:44:24
- date last changed
- 2026-09-15 21:27:25
@inproceedings{429234e5-a126-41e2-85bf-f3d239b7f4be,
abstract = {{<p>Virtual Imaging Trials (VITs) provide a reproducible, cost-effective, and ethical approach for testing medical imaging technologies. In breast imaging, realistic phantoms across density levels are crucial, as density affects cancer risk assessment and imaging performance. We have previously demonstrated the use of Perlin noise to simulate breast parenchyma. This study establishes a systematic mapping between Perlin noise parameters and breast density, enabling automated phantom generation with user-specified density targets. Density control was achieved by linking Perlin noise threshold values (TVs) to the resulting volumetric breast density through a set of algorithms. Using the STELLA-R simulation pipeline, multiple Perlin noise volumes were generated, superimposed, and thresholded. Because the TV influences the size and spatial distribution of dense structures, we established a quantitative mapping between TV and breast density. This mapping was then used to build a prediction model that estimates the TV required to achieve a specified density. By minimizing the difference between predicted and target densities, the algorithm determines TVs that reproduce desired breast densities in synthetic phantoms. The mapping revealed a monotonic relationship between TV and volumetric density, with variability introduced by the stochastic nature of Perlin noise. The mean density error between target and simulated output was 35%. Although this level of error indicates the need for further refinement, the proposed framework provides a reproducible and scalable foundation for density-controlled phantom customization in future virtual imaging studies. This approach supports imaging system evaluation and AI training across clinically relevant density ranges.</p>}},
author = {{Tomic, Hanna and Dustler, Magnus and Bakic, Predrag R.}},
booktitle = {{Medical Imaging 2026 : Physics of Medical Imaging}},
editor = {{Ganguly, Arundhuti and Li, Ke and Abbaszadeh, Shiva}},
isbn = {{9781510697850}},
issn = {{1605-7422}},
keywords = {{breast density; Breast phantoms; breast tumors; computer simulation; temporal changes in anatomy; virtual clinical trials}},
language = {{eng}},
month = {{04}},
publisher = {{SPIE}},
series = {{Progress in Biomedical Optics and Imaging - Proceedings of SPIE}},
title = {{Towards Improved Control of Breast Density in Simulated Mammograms via Perlin Noise Parameterization}},
url = {{http://dx.doi.org/10.1117/12.3088024}},
doi = {{10.1117/12.3088024}},
volume = {{13924}},
year = {{2026}},
}