A Virtual Imaging Framework for Three-Dimensional Quantitative Optoacoustic Tomography Using Stochastic Numerical Breast Phantoms

TL;DR

This work addresses the need for realistic, end-to-end evaluation of 3D optoacoustic tomography (OAT) for breast imaging by extending the stochastic optoacoustic numerical breast phantom (SOA-NBP) to include skin-tone variability and benign lesions, enabling robust system design studies. It introduces a comprehensive virtual imaging framework that couples stochastic NBPs with photon transport to compute $p_0(\boldsymbol{r}, \lambda)$ via $p_0(\boldsymbol{r}, \lambda)=\Gamma(\boldsymbol{r})\,\mu_a(\boldsymbol{r}, \lambda)\,\phi(\boldsymbol{r}, \lambda)$, and simulates acoustic propagation with transducer SIR and EIR using a $k$-space method, including attenuation described by a power law. The study provides a public dataset of $1{,}020$ NBPs with multi-wavelength fluence, $p_0$, and measurements, and demonstrates the framework through a case study comparing imaging-system designs, highlighting that incorporating the transducer impulse response during reconstruction is essential for maintaining resolution with larger elements. The framework offers a versatile platform for advancing qOAT methods and system design, with broad applicability to optical/acoustic inversion tasks and potential extensions to lesion-vascular and functional analyses in breast imaging.

Abstract

Optoacoustic tomography (OAT) is a promising modality for breast cancer diagnosis because tumor angiogenesis and, potentially, hypoxia can be visualized using quantitative OAT (qOAT) techniques. Clinically meaningful inference generally requires accurate image reconstruction, which depends on measurement quality and the imager design. Virtual imaging offers a cost-effective alternative to experimental prototyping for system design evaluation and supports computational method development. This work presents a comprehensive virtual imaging framework for breast qOAT, extending a stochastic numerical breast phantom (NBP) generator by incorporating skin tone variation and both benign and malignant lesions. It enables end-to-end simulation, modeling transducer spatial and acousto-electric impulse responses. Its utility is demonstrated through a case study comparing two system designs. A total of 1,020 NBPs and associated measurement data have been made publicly available to accelerate research in optoacoustic and optical imaging. The framework provides a versatile platform for advancing computational methods and guiding system design optimization.

Figures (9)

• Figure 1: Virtual imaging framework for 3D qOAT of the breast. Anatomical, functional, optical, and acoustic NBPs are generated using the SOA-NBP tool Park2023, with extended features indicated by orange-outlined boxes. The generated optical NBPs are used to compute the optoacoustically induced initial pressure field. The resulting distribution and the acoustic NBPs are then used as inputs to simulate acoustic wave propagation and optoacoustic measurements, accounting for the impulse responses of the transducers.
• Figure 2: Anatomical numerical lesion phantoms (NLPs): (a) malignant tumor with a homogeneous composition (viable tumor cells only), (b) malignant tumor with a heterogeneous composition (viable tumor cells with a necrotic core and a peripheral angiogenesis region), (c) fibroadenoma, and (d) simple cyst. In (b), the lesion volume is clipped along a central plane to reveal the internal composition, whereas the full volumes are displayed in (a), (c), and (d). Volume rendering was performed in ParaView ParaView.
• Figure 3: Optical fluence distribution $\phi$ simulated using a newly implemented linear segment illuminator, approximated as a continuous line emitter with a conical angular distribution, shown from two different viewing angles. A 20 mm-long illuminator with a numerical aperture of $0.89$ was placed on the surface of the water tank. Volume rendering was performed in ParaView ParaView with logarithmic scaling and value-dependent opacity adjustment applied for effective visualization.
• Figure 4: Experimentally measured (black solid line) and modeled EIR (red dashed line) of a piezoelectric transducer (Imasonic, France). The time-domain response is shown at the top and the normalized frequency-domain magnitude at the bottom. The measured EIR (black solid line) was acquired by TomoWave Laboratories (Houston, TX), and the modeled EIR (red dashed line) was obtained employing the complex exponential method Tallavo2011, with the model order $\hat{N}=14$ selected based on AIC Tallavo2011 (see Appendix \ref{['APPENDIX:D']} for details).
• Figure 5: Representative distributions of functional, optical, and acoustic properties: (a) tissue labels, (b) oxygen saturation $s$, (c) blood volume fraction $f_b$, (d) melanosome volume fraction $f_m$, (e) optical absorption coefficient $\mu_a$, and (f) sound speed $c$ of type A, D, and C breasts (top to bottom). The type A breast (top) is naturally shaped, has skin phototype III ($f_m=8.575$%), and contains one malignant tumor with heterogeneous tissue composition. The type D breast (center) is naturally shaped, has skin phototype V ($f_m=15.25$%), and contains two malignant tumors with heterogeneous tissue composition. The type C breast (bottom) has a shape constrained by a hemiellipsoidal stabilizer cup and skin phototype I ($f_m=1.9$%) and includes four lesions: a malignant tumor with homogeneous tissue composition, a malignant tumor with heterogeneous tissue composition, a fibroadenoma, and a simple cyst. All lesions were positioned at $x=0$ except for the two malignant tumor in the type C breast. To visualize internal structures, only half of each breast volume is shown. The malignant tumor with heterogeneous tissue composition in the type C lies outside the displayed field of view. In panel (a), lesion tissues are illustrated with the same colors as in Fig. \ref{['FIG:2']}, while healthy tissues are shown in yellow. ParaView ParaView was used for volume rendering.