Table 1. Overview of PA digital breast phantoms.

Study year	Breast modeling technique	Tissue type	Lesion model	Advantages and limitations
2014 [26]	Combination of a few simple objects	Skin and fat	None	A model easy to accomplish but oversimplified
2017 [21]	Threshold segmentation from contrast-enhanced MRI data	Skin, vessel, fat and fibroglandular tissue	None	Accurate representation of the breast anatomy but suffers from large computational burden and huge time consumption, only 50 patients’ data were acquired
2017 [27]	Combination of a few simple objects	Background soft tissue	A circular target that had absorption properties of hemoglobin	A 2D model easy to accomplish but oversimplified, the optical parameters are inaccurate
2018 [22]	Load the initial PA pressure from X-ray breast image after processing	Background soft tissue visualized by X-ray images	Benign and malignant tumors segmented from X-ray images	Large database available, easy to accomplish, while the real PA images are significantly different from the simulated ones, the X-ray images can’t give enough contrast between malignant and benign tumors as well as between vessels and other background tissues
2018 [28]	Combination of a few simple objects	Glandular tissue and vessels	None	A 2D model easy to accomplish but oversimplified
2019 [23]	Application of the second phantom	Skin, vessel, fat and fibroglandular tissue	Tumors acquired from mice studies (imaged with fluorescent imaging system)	Tumor models derived from mice studies; no distinction between tumor types
2020 [24]	Segmentation of the digital mammography (DM) dataset based on deep learning	Skin, fat and fibroglandular tissue	Segmented from DM dataset	Large database available, high efficiency but only in 2D, does not include vessel structures which are crucial for PA imaging
2021 [25]	Reconstruct the 3D breast model through a series of 2D US B-scan slices, then perform segmentation	Skin, fat and fibroglandular tissue	Segmented manually from US images	Accurate representation of the breast anatomy but does not include vessel structures, suffers from large time consumption, no distinction between tumor types