A Fundamental Study on the Fabrication of Human Model Bone Phantom using an Entry-Level 3D Printer: using FDM Method for the Femur Model

보급형 3D 프린터를 이용한 인체 모형 뼈 팬텀 제작의 기초연구: Femur 대상으로 적층형 출력 방식 이용

Abstract

The purpose of this study was to create a phantom with a HU value similar to that of the human Femur using a 3D printer to replace the existing pig bone. A total of 372 people were analyzed to determine the HU value of human Femur. Using a 3D printer, a human bone model phantom was fabricated using PLA-Cu 20% and subjected to CT examination. Pig bones were 6 months old pigs, and bones 2 days after slaughter were used. As a result of the examination, the 3D printing phantom made with 80% of the internal filling showed a similar value to all data of the human body (p<0.05), and there was a difference from the pig bone (p>0.05). In addition, in the case of the HU value of Femur by age group, it was confirmed that the value of HU decreased as the age group increased (p<0.05). 3D printing and HU values confirmed a weak negative correlation with respect to the stacking height, but confirmed a strong positive correlation (R² = 0.996) with 182.13±1.290 in the inner filling (p<0.05). In conclusion, it was confirmed that the human body model phantom using 3D printing can exhibit a similar level of HU value to the human body compared to the existing pig bone phantom, and this study will provide basic data for the production of a human body model phantom using a 3D printer.

본 연구에서는 보급형 3D 프린터를 이용하여 인체 Femur와 유사한 HU 값을 가진 팬텀을 제작하여 기존 돼지 뼈를 대체할 수 있는지 분석하고자 하였다. 인체 Femur의 HU 값을 알아보기 위해 연령별 총 372명의 데이터를 분석하였다. 보급형 3D 프린터를 이용하여 PLA-Cu 20%를 이용하여 인체 뼈 모형 팬텀을 제작하여 CT 검사하였다. 돼지 뼈는 생후 6개월 된 돼지로 도축된 지 2일이 지난 뼈를 이용하였다. 검사결과 내부채움 80%로 제작한 3D 프린팅 팬텀이 인체의 모든 데이터와 유사한 값이 나타났고(p<0.05) 돼지뼈와는 차이가 있었다(p>0.05). 또한 연령대별 Femur의 HU 값의 경우 연령대가 증가할수록 HU의 값은 줄어드는 것으로 확인 되었다(p<0.05). 3D 프린팅과 HU 값은 적층 높이에 대해서는 약한 음의 상관성을 확인 하였지만 내부 채움에서는 182.13±1.290으로 강한 양의 상관성(R²=0.996)을 확인하였다(p<0.05). 결론적으로 3D 프린팅을 이용한 인체 모형 팬텀이 기존 돼지뼈 팬텀에 비해 인체와 유사한 정도의 HU 값을 나타낼 수 있음을 확인할 수 있었으며 이에 본 연구가 3D 프린터를 이용한 인체 모형 팬텀의 제작에 기초자료를 제공할 수 있을 것이라 사료된다.

Ⅰ. INTRODUCTION

Hardware and software advances have greatly improved the speed and picture quality of computed tomography (CT). Because of faster and more accurate scans using multi-detector computed tomography (MDCT) and helical scan modes especially, CT usage has been increasing annually^[1]. CT is able to provide anatomical information in cross-sectional slices and information about linear attenuation coefficients in the body^[2]. However, because CT uses x-rays, the exposure dose always needs to be considered. The issue of x-ray exposure in CT has been raised in several studies, and diagnostic medical radiology has been reported to cause the highest exposure levels^[3]. Thus, the major studies in CT research have focused on producing the highest quality image at the lowest dose. Due to the restrictions on potentially harmful research in humans, research using human models or standardized phantoms is essential. Phantom studies have no risk of radiation exposure, meaning that multiple image acquisition parameters can be adjusted arbitrarily, to quantitatively measure the effects of each parameter on the image quality and radiation dose^[4]. These studies are also very useful for comparative analysis, because they allow for repeated scanning. However, there are several factors limiting phantom studies: because it is difficult to make human phantoms domestically, these studies depend on imports; because phantoms cannot be easily moved, there are differences with actual scanning and the quality of education; and because phantoms are expensive, it is difficult for every educational institution to own phantoms for each part of the body^[5]. To avoid these difficulties, pig bone, which shows similar properties to human bone, is often used instead of a human phantom^[6,7]. However, pig bone phantoms spoil readily, are difficult to store, and may show different states of decay at the time of slaughter. There is a need for a material to replace pig bone phantoms. Recently, quaternary industrial technology has had a massive influence on the whole field of medicine (e.g., using big data to collect patient data), and 3D printing technology especially is being actively adopted in medicine, including educational models of the human body, surgical simulation, and the development of personalized medical equipment^[8]. 3D printing is a method of product manufacture, in which, using 3D graphic design techniques, a physical, tangible object is fabricated using digitalized modeling data, CT scans, or magnetic resonance imaging (MRI) scans as virtual representations of the target object. 3D printers are devices, based on simple principles, that can generate 3D output using additive manufacturing (AM) to recreate 3D digital data in a series of 2D cross-sections^[9,10]. In this study, used an entry-level 3D printer to fabricate a phantom with similar Hounsfield units (HU) to human femurs, and analyzed whether this phantom could replace conventional phantoms using pig bone.

Ⅱ. MATERIALS AND METHODS

1. Materials and equipment

In this study, we used a CT device (Somatom Definition Flash, Simense Healthcare, Germany) to obtain HU values, and used AW VolumeShare 5 (AW 4.6 version, GE Healthcare, USA) and AW Volumeshare 7 (AW 4.7 version, GE Healthcare, USA) for post-processing of the acquired HU values.

The 3D printer for phantom fabrication was an AM device with a maximum print size of 22×22×25 cm3 (pinter A8, Samdimall, Korea), and the printing material was domestically manufactured PLA-Cu 20%[Fig 1].

Fig. 1. 3D printer for producing phantom (A) and CT equipment for scanning pig Femur and phantoms (B).

For the pig bone, we used a pig femur from a 6-month-old pig that had been slaughtered 2 days earlier. SPSS (V.18.0, IBM, USA) was used for statistical analysis of the acquired HU values.

2. Acquisition of HU values in humans

Among patients who received femur CT scans between January 1st and January 31st, 2019, we retrospectively analyzed left and right femur HU from male and female patients aged 20–79 years. The patients were divided into 12 groups by age and sex: males aged 20–29 years (24.80±2.78 years), 30–39 years (34.51±2.63 years), 40–49 years (44.12±2.76 years), 50–59 years (54.77±2.87 years), 60–69 years (64.03±2.64 years), and 70–79 years (74.06±2.80 years), and females aged 20–29 years (24.64±2.49 years), 30–39 years (34.80±2.98 years), 40–49 years (45.16±2.89 years), 50–59 years (54.29±2.75 years), 60–69 years (64.00±2.82 years), and 70–79 years (74.74±2.77 years). A total of 372 patients were included in the study, with 31 patients in each group.

Using AW VolumeShare 5 (AW 4.6 version, GE Healthcare, USA), signal intensity was measured from the coronal slice in which the femur head was largest; HU was measured 5 times from each leg of each patient, to obtain a total of 3720 HU values[Fig 2].

Fig. 2. Imaging for measuring HU by setting ROI on human Femoral head right (A) and left (B).

3. 3D printing process

After modeling of right femur DICOM files using AW VolumeShare 5 (AW 4.6 version, GE Healthcare, USA) and AW Volumeshare 7 (AW 4.7 version, GE Healthcare, USA), the model were converted to STL files. The STL files were sliced using Cura (V.15.04.6, Ultimaker, Netherlands). The printing parameters were set as follows: printing nozzle size, 0.4 mm; nozzle temperature, 185–190ºC; bed temperature, 60ºC; printing speed, 50 mm/s; shell thickness, 0.8 mm. Phantom types were defined by two methods: fixing the layer height to 0.25 mm and using an infill of 100%, 90%, 80%, 70%, 60%, 50%, or 40%, or fixing the infill to 100% and using a layer height of 0.05 mm, 0.1 mm, 0.15 mm, 0.2 mm, or 0.25 mm. The designs were converted to G-code, so that they could be understood by the 3D printer. The G-code files were sent to the printer, and a total of 11 3D printed femur phantoms were fabricated[Fig 3].

Fig. 3. 3D printing process. (A) G-code file (B) FDM method 3D printing (C) Completed real size human bone phantom.

[Table 1] shows the printing times for the different combinations of infill and layer height. At a constant layer height of 0.25 mm, the printing time for the 3D printed femur phantom was 4h 7 mins for 100% infill, 4h 1 min for 90% infill, 3h 52 mins for 80% infill, 3h 42 mins for 70% infill, 3 h 33 mins for 60% infill, 3h 23 mins for 50% infill, and 3h 14 mins for 40% infill. At a constant infill of 100%, the printing time for 3D printed femur phantom was 20 h 2 min for a layer height of 0.05 mm, 10h 5 mins for a layer height of 0.1 mm, 6h 46 mins for a layer height of 0.15 mm, 5h 6 mins for a layer height of 0.2 mm, and 4h 7 mins for a layer height of 0.25 mm.

Table 1. Production time for making phantom

Since the infill 100%, layer height 0.25 mm condition appeared in both methods, this phantom was only printed once. The completed 3D printed femur phantoms were removed from the supporter attached to the base, and sanded to make a smooth surface.

4. Method of HU acquisition from the pig bone phantoms and the 3D printed femur phantoms

After performing one CT scan each of the pig bone phantom and the 3D printed femur phantoms, coronal slices were obtained. The imaging conditions were set to 120 kV, 12 Effective mAs (Eff. mAs), and 0.8 pitch, and Care Dose 4D was used. The slice thickness was 1 mm, the B30f kernel was used, the matrix size was 512×512m the SFOV was 50 cm, and the DFOV was 74 mm. The isocenter was located at the vertical and horizontal center of the phantom.

Using AW VolumeShare 5 (AW 4.6 version, GE Healthcare, USA), based on the coronal slice in which the femur head was largest, we took 30 measurements of signal intensity from each of the eleven 3D printed femur phantoms and the one pig femur, to obtain a total of 360 HU values[Fig 4].

Fig. 4. Imaging for measuring HU by setting ROI on phantom Femoral head (A) and pig bone Femoral head (B).

5. Statistical analysis

Using SPSS (V.18.0, IBM, USA), an ANOVA test and Dunnett’s post-hoc test were performed to compare mean values according to sex and age group. An ANOVA test and Scheffé’s post-hoc test were performed to compare the mean HU values between the human patients, the pig bone phantom, and the 3D printed femur phantoms. A multiple linear regression analysis was performed to analyze the correlation between age and femur head HU. Another linear regression analysis was performed to investigate how layer height and infill affected the HU values in the femur head of the 3D printed femur phantoms. For all tests, a p-value <0.05 was considered to be statistically significant.

Ⅲ. RESULT

1. HU values of the femur head in each test

As shown in Table 2, first, by age group, the HU value of the femur head in the right leg of male patients was highest in the 20–29-year-old group (504.42 ± 39.40) and lowest in the 70–79-year-old group (407.37 ± 59.49; p<0.05). In the left leg of male patients, the HU value was lowest in the 20–29-year-old group (497.42 ± 45.91; p<0.05). In the right leg of female patients, the HU value was highest in the 70–79-year-old group (494.07 ± 51.97) and lowest in the 30–39-year-old group (458.20 ± 66.00; p<0.05). In the left leg of female patients, the HU value was highest in the 70–79-year-old group (492.85± 46.97) and lowest in the 40–49-year-old group (464.44 ± 55.75; p<0.05). Among the 3D printed femur phantoms, the HU value was highest at a layer height of 0.05 mm (753.25 ± 2.68) and lowest at a layer height of 0.25 mm (747.12 ± 2.51; p<0.05). Meanwhile, the HU value was highest with an infill of 90 % (551.29 ± 4.40) and lowest with an infill of 40% (–308.12 ± 7.98; p<0.05).

When the infill was 80%, there was no significant difference in the mean HU compared to the complete human patient dataset (p>0.05) but there was a significant difference with the pig bone phantom (p<0.05)[Table 2].

Table 2. HU value of femoral head for each test

*. p1= ANOVA test for each variable, _p1*=Post hoc test by Dunnett, p2=ANOVA test for all variables, _p2*=Post hoc test by scheffe

2. Correlation analysis of femur head HU values with human variables and 3D printer parameters

Table 3 shows the results of correlation analysis of the femur head HU values by age group in human patients. HU values showed a negative correlation with age group in the right leg of male patients (–16.00 ± 0.954; R²=0.240) and in the left leg of male patients (–14.98 ± 1.012; R²=0.190; p<0.05). HU values showed a positive correlation with age group in the right leg of female patients (6.415 ± 1.117; R²=0.185) and in the left leg of female patients (4.287± 1.013; R²=0.138; p<0.05). In the 3D printed femur phantoms, the HU values in the femur head showed a negative correlation with layer height (–1.278 ± 0.216; R²=0.186) and a positive correlation with infill (182.13 ± 1.290; R²=0.996; p<0.05)[Table 3].

Table 3. A study on the relationship between human variables and 3D printer in the femoral head HU values

*. Standard Errors(SE), Liner regression analysis of each variables

Ⅳ. DISCUSSION

The use of CT has increased due to advances in CT technology, and this, in turn, has led an increase in the radiation exposure dose. CT testing causes a greater dose of radiation exposure than plain radiography^[11]. Human phantoms are being used in efforts to reduce this dose, but these present problems such as their expensive price and dependence on imports^[5]. To alleviate these problems, many researchers use animal bones that show human-like properties, but this presents its own set of problems, such as decomposition^[6,7]. In this study, we aimed to fabricate a phantom to replace pig bones using a 3D printer, and to compare the HU of the 3D printed femur phantoms with human femurs and a pig bone phantom.

For human HU values, we compared the mean values for the femur head in the each leg of patients aged 20–79 years old, differentiating between male and female patients. The male patients showed decreasing HU in the left and right femur heads with increasing age, while the female patients showed increasing HU in the left and right femur heads with increasing age. There are several factors that can explain the increase in HU with older age in female patients, and these are discussed below.

The factors involved in x-ray attenuation in CT imaging include the thickness, atomic number, and density of the target, and the photon energy of the x-ray^[12]. Of these, bone density is the most important attenuation factor for bone imaging, and this is affected by alcohol consumption, drugs, and body fat. Kim et al. analyzed the distribution of osteoporosis across all age groups in South Korea, and reported that, the prevalence of osteoporosis gradually increases starting from 50 years of age, and that this is expected to be accompanied by an increase in the rate of medication^[13]. In our study, the HU of the left and right femur head showed a negative correlation in male patients, with decreasing HU as age increased. On the other hand, the HU of the left and right femur head in female patients increased with older age. This could be inferred from other studies. First, Park et al. compared bone density in drinking and non-drinking groups, and reported higher bone density in the drinking group^[14]. Riggs BL et al. reported that several drugs for health preservation and treatment contain estrogen, which can temporarily increase HU in CR scans^[15]. Edelstein et al. analyzed the correlation between body fat and bone density, and reported that groups with high body fat showed higher bone density^[16]. Combining the results of these studies, factors affecting bone density, such as body fat, medication, and alcohol consumption, should be controlled before the study, but because we collected the human HU values retrospectively, we were unable to control these factors, which is a limitation of our study.

When manufacturing 3D printed femur phantoms, the material, layer height, infill, and printing time need to be considered. In our study, the phantom with a layer height of 0.05 mm and infill of 100% took the longest time to print, at 1202 mins, while the phantom with a layer height of 0.25 mm and infill of 40% was the quickest to print, at 194 mins. Thus, the printing time is faster with taller layer height and less infill. In addition, we analyzed the correlation of femur head HU with layer height and infill, and found that the infill was very strongly correlated with HU (R2=0.996). Specifically, as the infill increased, the HU increased. Yoon et al. experimented with using a 3D printer to make phantoms for use in medical imaging tests, and reported that phantoms similar to standard phantoms could be fabricated using 100% infill^[17]. However, in our study, phantoms with an infill of 80% showed similar HU values to human femur heads. This discrepancy between the two studies is thought to be due to differences in the printing material. Yoon et al. used PLA, whereas we used PLA-Cu 20%.

We also analyzed the HU of a pig femur head, which has conventionally been used as a phantom to replace human bone, and compared the results with scans from human patients. In our analysis, the pig bone phantom showed different mean HU compared to human data classified by sex and age group. Therefore, the pig bone phantom can be considered inappropriate as a substitute for the human body, given that it not only presents difficulties in reproducibility and storage, but also shows different HU values from actual human patients.

One limitation of this study is that we only used a single PLA-Cu 20% material, and that we could not diversify the 3D printing method. However, there is value in the demonstration that we could produce more accurate phantoms than conventional pig bone phantoms using only an entry-level 3D printer.

Ⅴ. CONCLUSION

In this study, we demonstrated the feasibility of using an entry-level 3D printer to manufacture femur phantoms that are closer to in vivo human HU values than pig bone phantoms. When manufacturing 3D printed femur phantoms, the HU was strongly correlated with infill. We investigate the HU of phantoms fabricated from PLA-Cu 20% while varying the layer height and the infill. Given that HU values differ between different organs in the body, if the HU values of diverse 3D printed materials were analyzed, it would be possible to fabricate other human-like phantoms using 3D printing. When future studies investigate the fabrication of CT phantoms using different materials and printing methods, this study will provide basic supporting data.