Mitigating Under-Sampling Artifacts in 3D Photoacoustic Imaging Using Res-UNet Based on Digital Breast Phantom

Abstract

In recent years, photoacoustic (PA) imaging has rapidly grown as a non-invasive screening technique for breast cancer detection using three-dimensional (3D) hemispherical arrays due to their large field of view. However, the development of breast imaging systems is hindered by a lack of patients and ground truth samples, as well as under-sampling problems caused by high costs. Most research related to solving these problems in the PA field were based on 2D transducer arrays or simple regular shape phantoms for 3D transducer arrays or images from other modalities. Therefore, we demonstrate an effective method for removing under-sampling artifacts based on deep neural network (DNN) to reconstruct high-quality PA images using numerical digital breast simulations. We constructed 3D digital breast phantoms based on human anatomical structures and physical properties, which were then subjected to 3D Monte-Carlo and K-wave acoustic simulations to mimic acoustic propagation for hemispherical transducer arrays. Finally, we applied a 3D delay-and-sum reconstruction algorithm and a Res-UNet network to achieve higher resolution on sparsely-sampled data. Our results indicate that when using a 757 nm laser with uniform intensity distribution illuminated on a numerical digital breast, the imaging depth can reach 3 cm with 0.25 mm spatial resolution. In addition, the proposed DNN can significantly enhance image quality by up to 78.4%, as measured by MS-SSIM, and reduce background artifacts by up to 19.0%, as measured by PSNR, even at an under-sampling ratio of 10%. The post-processing time for these improvements is only 0.6 s. This paper suggests a new 3D real time DNN method addressing the sparse sampling problem based on numerical digital breast simulations, this approach can also be applied to clinical data and accelerate the development of 3D photoacoustic hemispherical transducer arrays for early breast cancer diagnosis.

Keywords: breast imaging; deep learning; image reconstruction; photoacoustic imaging; under-sampling.

Figure 1

(a) Pipeline utilizing Res-UNet model to solve under-sampling problem based on PA numerical breast phantom; (b) (i) Overview of numerical breast phantom; (ii) Extracted blood vessel structure; (iii) Numerical breast phantom components in sagittal plane.

Figure 2

(a) The diagram of photoacoustic imaging simulation using a hemispherical transducer array; (b) (i) Cross-sectional view of the light intensity distribution, denoted by ϕ₁; (ii) Cross-sectional view of light absorption distribution, denoted by ϕ₂; (iii) Average energy deposition at different depths, denoted by ϕ₃.

Figure 3

Visual representation of customized 3D Res-UNet architecture.

Figure 4

Performance of 3D deep learning network. First column: numerical breast phantom (initial pressure distributions); second column: dense reconstruction results; third column: sparse reconstruction results (under-sampling ratio 10%); fourth column: DNN results (same under-sampling ratio as the third column). (a) MAPs in the coronal plane. (b) MAPs in the sagittal plane. (c) Close-up views of the regions outlined by the white bounding boxes in (a). (d) Close-up views of the regions outlined by the yellow bounding boxes in (b).

Figure 5

Performance of 3D deep learning network for data with different degrees of sparsity. (a) MAPs of ROI in coronal plane and (b) MAPs of ROI in sagittal plane on sparse reconstruction results and corresponding DNN outputs for 256, 410 and 512 transducer arrays. First row: sparse reconstruction results. Second row: corresponding DNN outputs. (c) 3D MS-SSIM and (d) 3D PSNR evaluation metrics for transducer array numbers of 256, 310, 360, 410, 460, and 512.