In mammography, an automatic exposure control (AEC) is utilized with the objective of reducing radiation doses applied to the patient [1,2]. The AEC automatically determines the required tube current (mAs) for the desired image in advance, thereby preventing overexposure to X-ray [2]. Considering that the performance of the adjustment factors for the AEC varies depending on the manufacturer, it is crucial to implement optimal conditions tailored to each system. Nevertheless, in digital imaging, due to the wide latitude of the images, if the AEC adjustment factor setting is not correct, the image may be overexposed without any change in image quality [3]. The adjustment factors for AEC include the position and number of ion chambers, sensitivity, and the density and location match between the field configuration in the detector and dominant zone. These adjustment factors directly influence image quality and exposure dose [4]. Moreover, depending on the thickness, density, and atomic number, which define the subject’s physical properties, the ion chambers embedded within the AEC are directly affected by attenuated X-rays and its scatter rays in terms of electrical signal conversion. Thereby, the exposure tube current is automatically adjusted.

To optimize the AEC adjustment factors, the role of the objects is crucial. X-rays were generated in a tube and attenuated as they passed through objects until their intensity reached the set concentration threshold, as determined by the detector’s field configuration. The International Electrotechnical Commission recommended displaying the exposure index (EI) as a relative value indicating the radiation dose received by the image receptor for digital radiography devices [5-7]. Because the EI values could vary among manufacturers, setting a target exposure index (EI_T) that represents the optimal diagnostic value is important. When the EI matches the EI_T, the deviation index (DI) becomes zero, indicating that the radiation dose has been optimized. The EI_T provides information on the minimum dose that could achieve optimal image quality. The EI_T value will typically be established by the user and/or the digital radiography (DR) system manufacturer and stored as a table within the DR system [8]. The optimum EI_T value is derived from the image detected when the X-ray exposure is properly calibrated for the human body.

Nonetheless, this could cause an ethical problem, i.e., repeatedly exposing the human body to X-rays poses unnecessary radiation exposure to the subject. To avoid this problem, a phantom that closely resembles the human body is required [9]. In particular, for mammography, it is essential that the phantom resembles fine breast tissues. The existing American College of Radiology (ACR) phantoms, such as those shown in Fig. 1, have limitations as they typically indicate fibrous tissues through simple line forms and depict calcifications and masses through basic dot and circle shapes. These limitations in imaging differ from the histograms of actual human digital images and have posed challenges in optimizing the EI_T value. Especially concerning the AEC systems used for reducing radiation doses applied to the patient, it is crucial to optimize the adjustment factors simultaneously maintaining the quality of the diagnostic value in mammography images. Therefore, the breast anthropomorphic phantom (BAP) used in the AEC of a mammography system should be composed of tissue-equivalent materials that realistically and accurately represent the anatomy, tissue, and characteristics of organs [10]. Although several researchers have been developing BAPs, there have been challenges such as difficulty in distinguishing between real breast tissues and adipose tissues and limitations where internal structures are indicated merely by simple line and circle forms [11-13]. The 3D printing technology of the 4th industrial revolution is a technology that could convert actual 3D objects obtained using a scanner into sculptures that closely resemble the real object [14,15]. Therefore, we attempted to overcome the abovementioned limitations by converting mammographic images into three-dimensional images using the digital lithophane technology [16] and printing them using 3D printing technology to produce a precise human breast phantom.

Figure 1. Phantoms of the American College of Radiology.

The purpose of this study was to develop a BAP that could be utilized for dose optimization in AEC in a digital mammography system.

1. Mammography images and anthropomorphic phantom 3D printing

The 3D-printed high-density and low-density BAPs depicted in Fig. 2 were generated using the digitalized lithophane technology based on selected clinical images.

Figure 2. 3D-printed breast anthropomorphic phantom developed using the digitized lithophane technology.

2. Optimization of tissue-equivalent PMMA support thickness

The phantom was positioned at the end of the image detector in the mammography system to optimize the support structure thickness, as depicted in Fig. 3. As shown in Table 2, the auto-exposure values for kVp, mAs, EI, and DI were obtained using the AEC, based on the ACR phantom and varying the PMMA thickness of the BAP for two breast types. The support thickness of the BAP in high-density breasts approximated to the thickness of the ACR phantom was PMMA thickness of 3 cm, with an EI value of was 28,370 and a DI value of −0.096, at 28 kVp and 69.1 mAs. In low-density breasts, the support thickness was PMMA thickness of 4 cm, with an EI value of 28,565 and a DI value of −0.006, at 30 kVp and 71.4 mAs.

Table 2 . Automatic exposure control exposure parameters according to the ACR phantom and changing PMMA thickness of 3D phantom in breast types.

Breast type	Phantom	kVp	mAs	EI	DI
High-density	ACR phantom	28	55.5	29,295	0.043
	Only BAP	24	19.9	30,630	0.237
	BAP+PMMA 1 cm	25	40.0	34,040	0.696
	BAP+PMMA 2 cm	26	63.0	32,180	0.452
	BAP+PMMA 3 cm	28	69.1	28,370	−0.096
	BAP+PMMA 4 cm	30	79.9	26,920	−0.324
Low-density	ACR phantom	28	55.5	29,295	0.043
	Only BAP	24	16.1	50,115	2.376
	BAP+PMMA 1 cm	25	34.0	37,050	1.106
	BAP+PMMA 2 cm	26	54.4	35,540	0.884
	BAP+PMMA 3 cm	28	61.8	30,870	0.271
	BAP+PMMA 4 cm	30	71.4	28,565	−0.066

Figure 3. Selected clinical mammography and 3D printing products using lithophane technique, (a) high-density and (b) low- density breasts.

3. Optimization of the AEC system

1) Optimization of image quality

Table 3 shows the results of the evaluation by the two radiologists using the 7-point Likert scale for image quality. The highest total score of 5.25 points was consistently observed at density levels of −6 and −3 for the high-density BAP and at −6, −3, and 0 for the low-density BAP. As depicted in Fig. 4, mammographic images at all density levels were obtained for both high-density and low-density BAPs. These results revealed that, visually, there were no differences, and the structures of the breast were well observed without remarkable variations.

Table 3 . Qualitative evaluation results of radiologists.

Phantom	Density	Contrast	Sharpness	Total score
High-density	−6	5.5±0.5	5.0±1.0	5.25±0.8
	−3	5.5±0.5	5.0±1.0	5.25±0.8
	0	5.0±0.0	5.0±1.0	5.00±0.7
	3	4.5±0.5	4.5±0.5	4.50±0.5
	6	4.5±0.5	4.5±0.5	4.50±0.5
Low-density	−6	5.5±0.5	5.0±0.0	5.25±0.4
	−3	5.0±1.0	5.5±0.5	5.25±0.8
	0	5.0±1.0	5.5±0.5	5.25±0.8
	3	4.5±0.5	5.0±0.0	4.75±0.4
	6	4.5±0.5	4.5±0.5	4.50±0.5

Figure 4. Experiment of X-ray attenuation by changing polymenthylmethacrylate thickness, (a) 1 cm, (b) 2 cm, (c) 3 cm, and (d) 4 cm, of 3D phantom in the automatic exposure control of a digital mammography system.

2) Automatic exposure conditions according to density level

Table 4 shows the tube current values based on density level using the developed BAP. In the high-density BAP, with the tube voltage fixed at 28 kVp, the system automatically minimized the exposure to 39.1 mAs, achieving an EI value of 16,165 and a DI value of −2.542 at the −6 density level. At the −6 density level, the system automatically minimized the exposure to 49.2 mAs, achieving an EI value of 16,440 and a DI value of −2.466. At the 6 density level, the system exposed an automatic maximum of 136.4 mAs, with an EI value of 44,680 and a DI value of 1.877. With an increase in the density level, both low- and high-density phantoms showed an increase in EI and DI values.

Table 4 . Autoexposure values based on density level.

Phantom	Density	kVp	mAs	EI	DI
High-density	−6	28	39.1	16,165	−2.542
	−3		54.6	23,480	−0.917
	0		69.1	28,370	−0.096
	3		85.5	34,755	0.785
	6		101.0	41,570	1.564
Low-density	−6	29	49.2	16,440	−2.466
	−3		70.4	23,400	−0.932
	0		81.2	28,565	−0.066
	3		114.4	38,045	1.179
	6		136.4	44,680	1.877

3) Optimization of mAs

According to the evaluations of all images by the radiologists, image quality received higher ratings at −6 and −3 density levels in the AEC system. As shown in Fig. 5, the automatic exposure (mAs) differed according to changing density levels. In the high-density BAP, the current in the automatic exposure tube at −6 density level reduced by 39.4% compared with that at “0” density level, and in the low-density BAP, the current in the automatic exposure tube at −6 density level reduced by 43.4% compared with that at “0” density level.

Figure 5. Mammography on the anthropomorphic breast phantom by changing the density of automatic exposure control, (a) Density −6, (b) Density −3, (c) Density 0, (d) Density 3, and (e) Density 6, for low- and high-density phantoms.

This study aimed to optimize AEC used for dose reduction in mammography systems using BAP. It was essential to configure the AEC settings to achieve an optimal dose exposure tailored to the characteristics of the examinee in order to obtain a diagnostic-quality mammographic image. However, AEC misadjusts the dose automatically based on the density of set factors if foreign materials such as metallic objects are placed on the field configuration in the detector, thereby probably resulting in overexposure in the X-ray to reach the set control values [18,19]. To optimize these factors of AEC, the role of an object is important. The best object is a human being; however, repeated radiographic imaging of the human breast tissue to optimize AEC factors violates the as low as reasonably achievable principle, for which a physical phantom has been used. The currently used ACR physical phantoms are suitable only for quality evaluations of simple images in mammography systems and are unsuitable for implementing optimal dose of AEC factors.

Previous studies, e.g., the study by Kiarashi et al. [13], had limitations such as the inability to accurately indicate the parenchymal and adipose regions of the breast and the lengthy fabrication period of the phantom. To address these limitations, in the present study, the digitalized lithophane technology that can convert existing images into three-dimensional representations was used. Lithophane, a popular art form in the 19th-century Europe, utilized the variation in the amount of transmitted light based on the thickness of the plate. To represent images in detail, the internal space is completely filled to regulate the amount of projected visible light by adjusting the thickness of the plate [20]. In the present study, we developed an anthropomorphic phantom by allowing X-rays to be projected instead of visible light. Consequently, we overcame the limitations of previous studies by distinguishing between the breast parenchymal and adipose regions, undergoing a modeling process to create high-density and low-density BAPs that closely resemble real breast images. Furthermore, the BAP developed in this study exhibited structures resembling internal breast tissues, whereas the internal structures in the existing ACR phantom appear as simple lines, dots, or circular shapes.

In a previous study by Varallo et al. [11], the use of fused deposition modeling technology in 3D printing of structures resulted in the formation of tiny air layers inside the printed material, causing a degradation of image quality. However, in this study, we used the DLP technology, which uses a liquid resin material cured by ultraviolet light, to eliminate microgaps and minimize noise in radiographic images. This approach allowed us to achieve excellent image quality. The brightness of mammographic images depends on X-ray attenuation that is influenced by the thickness of the object. Therefore, in this study, we optimized the support thickness using PMMA, a human-equivalent material, placed underneath the BAP to derive equivalent thickness compared with that of the existing ACR phantom.

Therefore, the PMMA support thickness was optimized as 3 cm for the high-density BAP and 4 cm for the low-density BAP, resulting in equivalent thickness compared with that of the ACR phantom. It is believed that the breast parenchyma shows a greater distribution in the high-density BAP than in the low-density BAP, resulting in greater X-ray attenuation and an equivalent thickness with a relatively thin PMMA support. This improvement reflects the clinical characteristics and overcomes the limitations of existing phantoms that fail to capture the nature of the breast tissue.

The density level in the AEC system was adjusted in five steps for both high-density and low-density BAPs. Radiologists evaluated the density level with the lowest automatically exposed tube current as the optimal level, which did not affect the diagnostic quality of the image. Consequently, compared with the middle-level density, the automatically exposed tube current decreased by 39.4% for high-density breasts and by 43.4% for low-density breasts in Fig. 6. This dose reduction is attributed to the higher sensitivity to radiation at lower density level settings, allowing the formation of image contrast with minimal radiation exposure. This study was conducted through repetitive experiments using anthropomorphic phantoms instead of human subjects, thus ensuring the maintenance of diagnostic image quality.

Figure 6. Comparison of autoexposure mAs by changing the density on anthropomorphic high-density and low-density phantoms.

Nonetheless, this study may have reliability errors due to the subjective qualitative evaluation by radiologists. Another limitation is that it focused on a single manufacturer. According to previous research, variations in radiation dose can occur based on the performance of the X-ray generator and detector. Therefore, future studies must consider multiple manufacturers for a comprehensive investigation [21,22]. Furthermore, the breast size of the anthropomorphic phantom was based on the average breast size of Koreans (A cup), which is a limitation. There are also limitations, i.e., the kVp value was not constant because the compression intensity could not be consistently maintained during the experiment, and a quantitative evaluation could not be performed in the final image quality evaluation stage. However, further research could overcome these limitations, and the development of BAPs holds significant potential for advanced applications.

This study developed a BAP using digitalized lithophane and 3D printing technologies for optimizing the dose modulation factors of AEC in mammography. First, high-density and low-density BAPs were developed, which can provide images similar to actual human breast radiographic images. Second, the automatic exposure tube current reduced with optimization of the density level of AEC using the BAP. We anticipate that the 3D-printed BAP developed in this study could contribute to dose reduction by optimizing AEC simultaneously maintaining high-quality radiographic images in mammography.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-forprofit sectors.

The authors have nothing to disclose.

All relevant data are within the paper and its Supporting Information files.

Conceptualization: Youl-Hun Seoung. Data curation: Hye-Jin Kim. Formal analysis: Hye-Jin Kim. Investigation: Hye-Jin Kim. Methodology: Youl-Hun Seoung. Project administration: Youl-Hun Seoung. Resources: Hye-Jin Kim. Software: Hye-Jin Kim. Supervision: Youl-Hun Seoung. Validation: Youl-Hun Seoung, Hye-Jin Kim. Visualization: Youl-Hun Seoung, Hye-Jin Kim. Writing – original draft: Hye-Jin Kim. Writing – review & editing: Youl-Hun Seoung.