Development of hybrid 3D printing approach for fabrication of high-strength hydroxyapatite bioscaffold using FDM and DLP techniques

This study develops a hybrid 3D printing approach that combines fused deposition modeling (FDM) and digital light processing (DLP) techniques for fabricating bioscaffolds, enabling rapid mass production. The FDM technique fabricates outer molds, while DLP prints struts for creating penetrating channels. By combining these components, hydroxyapatite (HA) bioscaffolds with different channel sizes (600, 800, and 1000 μm) and designed porosities (10%, 12.5%, and 15%) are fabricated using the slurry casting method with centrifugal vacuum defoaming for significant densification. This innovative method produces high-strength bioscaffolds with an overall porosity of 32%–37%, featuring tightly bound HA grains and a layered surface structure, resulting in remarkable cell viability and adhesion, along with minimal degradation rates and superior calcium phosphate deposition. The HA scaffolds show hardness ranging from 1.43 to 1.87 GPa, with increasing compressive strength as the designed porosity and channel size decrease. Compared to human cancellous bone at a similar porosity range of 30%–40%, exhibiting compressive strengths of 13–70 MPa and moduli of 0.8–8 GPa, the HA scaffolds demonstrate robust strengths ranging from 40 to 73 MPa, paired with lower moduli of 0.7–1.23 GPa. These attributes make them well-suited for cancellous bone repair, effectively mitigating issues like stress shielding and bone atrophy.


Introduction
Millions of patients worldwide suffer bone injuries each year due to trauma, infection and tumors, resulting in the formation of bone defects that require filling with repair materials [1].Therefore, the repair of injured or defective bones is a crucial concern for orthopedic surgeons.Currently, the commonly used methods for bone defect repair are autografts and allografts [2][3][4].Autografts involve using the patient's own bone tissue, which offers several advantages such as compatibility, reduced risk of disease transmission, and absence of transplant rejection.However, autografts have limitations including limited availability, additional surgical site morbidity, and potential complications associated with the harvesting procedure [5,6].On the other hand, allografts, which involve using bone tissue from another individual, offer the advantage of readily available grafts.However, allografts also have several drawbacks, including the risk of secondary surgery, potential disease transmission, and the possibility of transplant rejection [7,8].Considering these factors, bone tissue engineering has emerged as an attractive research field to develop alternative solutions for bone defect repair [9][10][11].The objective of bone tissue engineering is to aid in tissue repair rather than replacement, particularly for large bone defects that cannot heal naturally.This has sparked considerable interest in porous scaffolds fabricated from biomaterials that are biocompatible, bioresorbable, biodegradable, and non-toxic [12][13][14].As a result, synthetic bone scaffolds have become potential alternatives to autografts and allografts.These scaffolds not only provide excellent osteoconductivity and osteoinduction, but also mitigate immune reactions and eliminate the risk of acquiring infectious diseases associated with transplanted tissues [15].
In recent years, additive manufacturing (AM) technology, also known as 3D printing, has emerged as a game-changer in rapid production [16].Fused deposition modeling (FDM) is the most widely used 3D printing technology, which involves stacking layers of designed models by heating and dissolving thermoplastic materials.In biomedical applications, FDM has also been utilized for bioprinting various musculoskeletal tissues, including bone [17][18][19][20][21][22][23][24][25], cartilage [26][27][28][29], ligament [30,31], tendon [32,33], muscle [34][35][36], and osteochondral unit [37,38].Inkjet printing employs a similar patterning technique as FDM, with the difference that the ejected material consists of distinct droplets instead of continuous fibers [39][40][41][42].However, the mechanical properties of 3D-printed tissues may be compromised by inadequate material bonding, which depends on the choice of biomaterials and stacking methods.Currently, bioprinted bone constructs in these studies do not meet native bone's mechanical characteristics [40].In addition, stereo lithography appearance and digital light processing (DLP) predominantly use photosensitive materials, which can be printed with good precision and improved mechanical properties [43].Synthetic DLP 3D-printing materials such as PLLA/BCP [44], TiO 2 [45], ZrO 2 [46], PU/ hydroxyapatite (HA) [47], and β-TCP [48] composites necessitate the addition of photosensitive resins and photoinitiators to form a paste that can be cured under ultraviolet (UV) irradiation.However, most of these materials are toxic and unsuitable for human implantation [49,50], and further investigation is needed regarding the mechanical properties of the sintered products.Selective laser melting (SLM) and selective laser sintering (SLS) enable the fabrication of products from metal or ceramic powders [51][52][53].However, SLM/SLS is not commonly used in the manufacturing of ceramics because the temperature difference during the laser sintering or melting of ceramic powders easily generates stress in the ceramic products, leading to the formation of numerous cracks within the ceramic products [54,55].
In addition to direct 3D-printing, the use of molds for the preparation of bioceramic scaffolds has gained popularity as the 3D printing method allows for the production of high-precision molds based on designs.Some studies have suggested that methods like direct coagulation casting and gelcasting can produce structurally complex and highstrength ceramics [56,57].Currently, to facilitate mold removal, degradable materials are commonly used for 3D-printed molds, such as water-soluble polyvinyl alcohol (PVA) [58] and acetone-soluble acrylonitrile butadiene styrene (ABS) [59].After the infusion of a slurry (ceramic material + binder) and subsequent processing, the final products with complex structures are obtained.The slurry can be dried and sintered at high temperatures, simultaneously burning off the binder to obtain the ceramic scaffold.Filippov et al [60] proposed a low-pressure injection molding method, where ceramic slurry was cast into FDM-printed ABS molds under low pressure, and the ABS mold was burned off through hightemperature sintering, leaving the ceramic scaffolds in place.Furthermore, the team utilized DLP to print a Gyroid model, resulting in samples that were sintered to obtain a bioceramic scaffold with macropores [61].Table S1 provides an overview of different 3D printing methods used in the literature for the preparation of ceramic materials.
A wide range of biocompatible materials suitable for 3D printing have been developed [62,63].Polylactic acid (PLA) is an affordable, non-toxic and biocompatible material, but it generates acidic byproducts during degradation, which may cause tissue inflammation and cell death [64][65][66].Poly (D, L-lactide) exhibits good biocompatibility and high mechanical strength, making it suitable for biomedical applications [67,68], while poly (ε-caprolactone) (PCL) is a low-cost and biodegradable polyester [69].Hydrogels offer tunable mechanics, degradation, and functionalization capabilities, making them valuable in various tissue engineering applications [70][71][72].Pluronic F-127 (PF-127), a hydrogel composed of poly (ethylene oxide) and poly (propylene oxide), is non-toxic, biocompatible, and biodegradable [73].It can be softened at low temperatures and cured at high temperatures, making it suitable for biomaterial studies and as a blend for bioprinting [74].Among the bioceramic materials explored for bone repair, HA has received significant attention due to its similarity to bone composition, excellent biocompatibility, and non-toxic properties [75][76][77][78].
Although there have been numerous studies on 3D printed bioceramic scaffolds, there are still areas that require improvement.For instance, while FDM offers easy and fast printing, the samples suffer from poor precision and mechanical strength, and the production time for a single model is longer.DLPprinted scaffolds address accuracy and process time issues, but the material preparation involving photosensitizing agents may lead to toxin residues on the sample.In our study, we have innovatively combined the advantages of FDM and DLP methods to rapidly fabricate non-toxic bioceramic scaffolds with exceptional mechanical properties, which is a novel approach not previously documented in the literature.In this process, HA served as the bioceramic material, and PF-127 was selected as the binder.A homogenized bioceramic slurry, consisting of HA and PF-127, was filled into FDM-printed outer molds.Compared to traditional subtractive manufacturing, the FDM method allows for rapid adjustments in the size and style of the outer mold, accommodating various machine specifications for vacuum centrifugal defoaming in subsequent steps to increase the bioscaffold density.Additionally, the outer molds, 3D printed with resin, can withstand the rigors of vacuum centrifugal defoaming.This eliminates the need for manufacturing molds with metal materials, hence significantly reducing costs.Simultaneously, a high-precision DLP printer was employed to 3D print precise struts, which were then inserted into the slurry to create the designed channels.The advantage of using DLP method is its ability to facilitate cost-effective mass production, with the designed channel sizes and porosity being adjustable by modifying the design parameters of the printed struts, including diameter and spacing.The resin used in DLP printing is not part of the bioscaffold's composition.It only comes into contact with the sample surface, and any residual traces can be easily removed through sintering, rendering the scaffold entirely non-toxic.Following strut insertion, vacuum centrifugal defoaming was applied to remove bubbles, even on a nanoscale, significantly increasing the density of the HA slurry.This step is critical in achieving the remarkable compressive strength displayed by the bioscaffold in this study.After drying and high-temperature sintering, crystallized HA scaffolds were obtained.Extensive investigations were conducted to analyze the mineralogical, morphological, physical, mechanical and biological properties, aiming to comprehend the factors and mechanisms influencing the scaffold properties.By combining these three key technologies (FDM, DLP, vacuum centrifugal defoaming), this innovative approach holds great potential for the mass production of robust and nontoxic bioscaffolds for bone tissue engineering.

Materials
In this study, HA (Alfa Aesar, USA) was used as the raw material, and PF-127 (Sigma-Aldrich, USA) was used as the binder for bioscaffold fabrication.PF-127, in contrast to high-polymer materials like PLA and PCL, exhibits reverse thermoresponsive properties that allow it to remain in a liquid state at room temperature, facilitating easy mixing with bioceramic powder.In addition, the water solubility of PF-127 allows for the avoidance of using organic solvents, such as ethanol and acetone, thereby preventing the dissolution of 3D-printed molds.PLA resin (MIN-YAU, Taiwan) was utilized to prepare the plastic outer mold.Three different DLP photosensitive resins were employed to print the struts, which create penetrating channels in the bioscaffolds.These photosensitive resins possess the following properties after the curing treatment: • FW900 (FEASUN, Taiwan): Soft and pliable • FT30B (FEASUN, Taiwan): Hard and stiff • Insta Resin Grey (Capybara Robot, Taiwan):

Tough and resilient
Based on the results, the most suitable resin material would be selected for strut printing and subsequent demolding.In addition, petroleum jelly (Chan Guare Industry, Taiwan) and lubricant (7717-GW, Taiwan Taisheng, Taiwan) were applied to the surface of outer molds and struts as release agents to facilitate the demolding of the bioscaffold.

Preparation of FDM-printed plastic outer mold and DLP-printed strut
The plastic outer mold was designed using 3D modeling software (123D Design, Autodesk, USA).The mold consisted of four individual molds, denoted as A, B, C, and D, as shown in figure 1(A).mold A resembled a hollow barrel with an outer diameter of 20 mm, an inner diameter of 14 mm, and a bottom that spread outward by 2.5 mm.mold B served as a base to be placed at the bottom of mold A, allowing for direct removal of the bioceramic samples after the demolding process.mold C was a component used to contain the assembly of mold A and mold B, which could be fixed by rotating it 45 degrees clockwise.mold D served as a safety device to secure mold A and mold B in place.The 3D models of these molds were sliced and converted to G-code using CURA software (Ultimaker, Netherlands).The sliced files were then imported into an FDM-type 3D printer (PING EDU, Linkin Factory, Taiwan) and printed using PLA resin with a nozzle size of 0.4 mm, a layer height of 0.2 mm, a print temperature of 195 • C, a print speed of 30 mm s −1 and a fill density of 30%.
The struts were designed to create penetrating channels in the bioscaffolds starting from the top of the HA slurry.3D modeling software (123D design, Autodesk, USA) was utilized to design different strut models with varying strut diameters (600, 800, and 1000 µm) and strut quantities, corresponding to different designed porosities of the bioscaffold (10%, 12.5%, and 15%).The samples were named in the format of X Y (X representing the strut diameter and Y representing the designed porosity) as follows: 600 10 , 600 12.5 , 600 15 , 800 10 , 800 12.5 , 800 15 , 1000 10 , 1000 12.5 , and 1000 15 .The design of different strut diameters and porosities is illustrated in figure 1(B).These models were imported into FlashDLPrint software (FlashForge, China) for slicing and file conversion.The converted files were printed using a DLP-type 3D printer (Hunter, FlashForge, China) and photosensitive resins.The struts were formed by the irradiation of UV source (λ = 405 nm) in the DLP printer with a layer height of 0.05 mm, light intensity of 40%, and exposure time of 26 s for the first layer and 6 s for the subsequent layers.Excess resin on the resulting struts was removed with 95% alcohol.The struts were then dried in an oven at 55 • C for 2 h.Finally, the struts were placed in a UV curing box (WH0001, Wanhao, China) and exposed to UV (λ = 405 nm) for 400 s to achieve complete curing.In addition to the struts designed to be inserted from the top of the HA slurry, another type of strut, which creates penetrating channels from the bottom of the slurry, was designed to investigate the effect of the direction of strut insertion on bioscaffold formation.This strut was designed to replace mold B (figure 1A(b)), and was a DLP-printed product with the identical base as mold B but featuring a strut with a diameter and quantity of 800 10 (figure 1B(b-1)) on the base.

Fabrication of bioscaffolds
The inner wall of the plastic outer mold was coated with a layer of lubricant to prevent the bioceramic samples from sticking to the mold during drying and demolding.The struts were dip coated with a layer of petroleum jelly at 60 • C, followed by the application of the lubricant on the surface of the petroleum jelly to create a double-layer coating of release agent.Prior to assembling the entire mold, a plastic film was placed on mold B (figure 1A(b)) to isolate the HA slurry from the base plate.
The flow chart of the bioscaffold preparation is shown in figure 2. Firstly, 1.2 g of PF-127 powder was added to 4.8 ml of deionized (DI) water.The mixture was sealed and kept at 5 • C until PF-127 fully dissolved in water, forming a liquid binder.Then, 6 g of PF-127 binder was mixed with 5 g of HA ceramic powder and kept at 5 • C for 15 min to obtain the HA slurry.A planetary centrifugal mixer (MV-300S, CGT Technology, Taiwan) was used to homogenize and defoam the slurry at 1500 rpm of revolution and 1200 rpm of rotation for 120 s.After homogenization, the HA slurry was loaded into a 10 ml syringe, and 1 g of the HA slurry was injected into each assembled plastic outer mold (figure 1(A)).The struts were then inserted into the HA slurry contained in the plastic outer mold to create penetrating channels through the bioceramics, followed by cooling storage at 5 • C for 30 min (Type A).Subsequently, a centrifugal defoaming treatment was applied at 1500 rpm of revolution and 1200 rpm of rotation for 120 s under vacuum conditions.The entire package (mold + strut + HA) was sealed with plastic wrap and kept at 5 • C for 1 h, followed by drying at zero humidity in an electronic moistureproof box under room temperature for 48 h.Another type of plastic mold assembly was used as a comparison to test the different insertion direction of the strut.In this assembly, mold B was replaced by a strut having the identical base as mold B with the strut parameters of 800 10 , and the HA slurry was directly injected into the assembly (Type B).The centrifugal defoaming was carried out under vacuum at the same rotation speeds to remove air bubbles.It is important to note that the removal of air bubbles and cooling at 5 • C significantly increased the density of the samples.After drying, the base of the DLP strut was removed using a laser cutter (Beamo, FLUX Inc, Taiwan), and the struts were completely removed after baking the sample at 90 • C for 60 min.The HA scaffold was obtained and polished using sandpaper, followed by a two-step sintering process.Firstly, the sample was calcined at 450 • C for 2 h with a heating rate of 5 • C min −1 to remove the polymers, moisture, and impurities.Then, the temperature was increased to 1200 • C (heating rate: 10 • C min −1 ) and maintained for 2 h to form a crystallized HA scaffold.A control group of HA ceramic samples without penetrating channels was also prepared and denoted as 0 0 .

Characterization
Three replicates for each sample were prepared for dimensional characterization.The channel size and spacing on the samples before and after sintering were observed by an optical microscope (IX73P1F, Olympus, Japan).The shrinkage of channels and designed porosity was analyzed using ImageJ software (National Institutes of Health, USA).The sample weight was measured by an analytical balance (AS 220.R2 PLUS, Radwag, Poland).The dimensions of the 3D printed FDM molds, DLP struts and HA scaffold were measured using a Vernier caliper.
The mineralogy of the raw HA powder, sintered HA powder and sintered HA scaffold, which was ground into powder form, was analyzed using an xray diffractometer (XRD, D2 Phaser, Bruker, USA) with CuKα radiation (λ = 1.541 84 Å) at 30 kV and 10 mA, with diffraction angles (2θ) ranging from 20 • to 70 • .The functional groups were determined by Fourier transform infrared spectroscopy (FTIR, VERTEX 70 v, Bruker, USA).The microstructure and the morphology were investigated by environmental scanning electron microscopy (ESEM, Quanta 200, FEI, USA) at an accelerating voltage of 10.0 kV.

Density analysis
The density of sintered scaffolds was measured by Archimedes' method in DI water.The weight of the sintered sample (m 1 ) was recorded after complete drying.The sample was then soaked in DI water and evacuated under vacuum conditions until there were no air bubbles in the surrounding water.The watersoaked sample was suspended in DI water using a combination of an Archimedes' principle kit and an analytical balance to obtain the immersed mass (m 2 ).After that, the sample was removed from the DI water, and the saturated weight in air (m 3 ) was measured.Bulk density (D b ) represents the dry weight of the sample divided by the volume of solid material, open pores and closed pores.The equation for bulk density is as follows: where ρ 0 is the density of DI water at 25 • C. Apparent density (D a ) is defined as the dry weight of the sample divided by the volume of solid material and closed pores.The equation for apparent density is expressed as follows: Apparent density The material volume (of solid HA can be obtained by dividing the dry weight of the sample (m 1 ) by the theoretical density of HA (D m = 3.156 g cm −3 ).The bulk volume (V b ) consists of the solid HA, open pores and closed pores, while the apparent volume (V a ) contains the solid HA and closed pores.The V m , V b , and V a can be derived from the following equations, respectively: The total volume of the sample (V t ) was obtained by summing the bulk volume (V b ) and the volume of the designed penetrating channels created by DLPprinted struts, as follows: where V d represents the volume of the penetrating channels, which was calculated by multiplying the measured channel aperture area by the sample height.The proportions of open pores, closed pores, designed penetrating channels, and total pores were presented by the open porosity (P o ), closed porosity (P c ), designed porosity (P d ) and total porosity (P t ), which can be derived from the following equations, respectively:

Mechanical strength analysis
The hardness of the sintered scaffolds was measured using a Vickers hardness tester (FM-810, Future-Tech, Japan).The hardness tester has a diamond indenter in the shape of a square-based pyramid with an apex angle of 136 • .The surface of the sintered samples was indented for 10 s with a loading force of 0.5 kg.The indentation diagonals of the resulting indent were measured to calculate the hardness.Sintered bioceramic scaffolds (Ø12 × H3 mm) were prepared for the characterization of compressive strength.The compressive strength and modulus of HA scaffolds were determined using a precision universal tester (AGS-50KNXD, Shimadzu, Japan) with a 50 kN load cell at a constant crosshead speed of 0.5 mm min −1 .Each sample group was subjected to three replicates, and the standard deviation (SD) was computed.
The MTT assay was carried out by replacing the initial EMEM in the MG-63 cell culture with 100 µl of the filtered extract.The MG-63 cells were incubated in the extracts for an additional 24 h.Following incubation, the extracts were removed, and 10 µl of a 5 mg ml −1 MTT solution in phosphate-buffered saline (PBS) was added to each well.The plates were then incubated in a 5% CO 2 atmosphere at 37 • C for 2 h.Afterward, the MTT solution was aspirated, and 100 µl of dimethyl sulfoxide (DMSO, Sigma-Aldrich, USA, 99.5%) was added to dissolve the purple formazan crystals.The optical density (OD) of the samples was measured at a wavelength of 570 nm using an ELISA microplate reader (Eon, BioTek, USA) to determine cell viability.The quantitative data were presented as the mean value along with the corresponding SD, with five replicates per experiment.As a control group, a positive control was created by adding sodium dodecyl sulfate (SDS, J.T Baker, USA, 99%), while negative control groups consisted of cells only and cells with the addition of alumina (Al 2 O 3 , Sigma-Aldrich, USA, 99.5%).The cell viability of the samples was determined by comparing their OD to that of the negative control, which consisted only of cells and was set as 100%.The OD of each sample was divided by the OD of the negative control to obtain the relative cell viability.
Cell adhesion was assessed using ESEM (Quanta 200, FEI, USA) and a live/dead dual-staining kit (Dr View, IMT Formosa New Materials, Taiwan).Before cell culture, the bioscaffold was vacuum-soaked in PS for 12 h.It was then vacuum-soaked in 95% ethanol for another 12 h, followed by autoclave sterilization at 121 • C for 20 min.For ESEM imaging, MG-63 cells were seeded on the 800 10 bioscaffold at a density of 5 × 10 3 cells well −1 in a 24-well plate and incubated for 7 d, with a change of 1000 µl EMEM every two days.Following the culture period, the sample was rinsed twice with 500 µl of PBS in a 48-well plate.Subsequently, the sample underwent a dehydration process using a series of ethanol solutions with increasing concentrations (30%, 50%, 60%, 75%, 90%, 90%, 95%), with each step lasting for 15 min at 37 • C.After the sample was thoroughly dried in a humidity-controlled chamber at zero humidity for 24 h, it was sputter-coated with gold before examination by ESEM.To perform live/dead imaging, MG-63 cells were initially seeded onto the 800 10 bioscaffolds at a density of 5 × 10 3 cells well −1 in a 24-well plate.This culture was maintained for 14 d, with a 1000 µl change of EMEM every two days.After the incubation, the samples were rinsed twice in PBS and then immersed in 1 ml of fresh PBS.Following this, 1 ml of a live/dead dye solution, consisting of a live cell dye (2 µg ml −1 calcein-AM, green) and a dead cell dye (4 µg ml −1 propidium iodide, red), was introduced into each well.The bioscaffolds were co-cultured with the dye solution for 30 min in darkness.Subsequently, the live/dead dye solution was aspirated, and the samples underwent PBS washing.Observation was performed using an optical microscope (IX73P1F, Olympus, Japan) equipped with fluorescence mirror units featuring excitation wavelengths of 488 nm (green for calcein-AM) and 531 nm (red for propidium iodide).
In degradation testing, the formula developed by Kokubo and Takadama [79] was used to prepare a simulated body fluid (SBF) at 36.5 ± 0.5 • C. To achieve a pH value of 7.4 for the SBF solution, 1 M HCl was added as necessary.Sintered bioceramic scaffolds 600 10 , 800 10 , and 1000 10 were immersed in 45 ml of SBF solution at 37 • C for 7, 14, 21, and 28 d, respectively.The samples were completely dried at 60 • C for 24 h before weighing.The weight loss of the samples was calculated as a percentage using the following formula: where W L represents the weight loss of the samples after immersion in SBF, W F denotes the final weight of the samples after SBF immersion, and W 0 represents the initial weight of the samples.Three replicates were performed for each sample group, and the SD was calculated.ESEM (Quanta 200, FEI, USA) equipped with energy-dispersive x-ray spectroscopy (EDS, Genesis XM 4i, EDAX, USA) was used to observe the morphology and newly formed HA after a 28 day immersion.

The selection of DLP photosensitive resin
In order to select the most suitable material for strut fabrication, three types of photosensitive resins (FW900, FT30B, and Insta Resin Grey) were used to print struts with different diameters by DLP method.The 3D printing and bioceramic casting results of the different DLP resins are shown in table S2.It can be observed that the struts printed with FW900 started to deform at 800 µm and failed to print below 700 µm in diameter.FW900 is a wax-like photosensitive resin commonly used in lost-wax casting, and its soft material nature was unable to support the struts during the 3D printing process, leading to difficulties in releasing the bioceramic sample from the struts.On the other hand, FT30B exhibited high toughness and precise forming during the printing process, but the bioceramic sample was vulnerable during the subsequent demolding treatment.In comparison, the Insta Resin Grey was more flexible and allowed for effective strut removal after 15 min of baking at 100 • C. Consequently, Insta Resin Grey was chosen as the ideal DLP material for the subsequent experiments.

Error evaluation for FDM molds and DLP struts
To assess the accuracy of the 3D printing techniques, measurements were taken on the FDMprinted plastic outer molds and DLP-printed struts using a vernier caliper.The purpose was to determine the deviations between the printed objects and the original design, as illustrated in figure S1 and summarized in table S3.For each set of model parameters, 6 FDM molds and DLP struts were measured.The results revealed that the actual inner and outer diameters, which measured 13.81 mm and 19.93 mm, respectively, exhibited a shrinkage of 1.4% and 0.4% compared to the designed values of 14 mm and 20 mm.These slight discrepancies in the printed products could be attributed to factors such as the temperature of the print head and the printing chamber environment when PLA resin was utilized.
Regarding the DLP-printed struts, those corresponding to a porosity of 10% demonstrated better accuracy than those with porosities of 12.5% and 15%.
Additionally, the strut models with larger diameters exhibited greater precision compared to smaller diameters.This can be attributed to the dense distribution of struts, which may result in higher shielding of UV irradiation, subsequently leading to a deterioration in strut formation.Furthermore, the pixel size of the light projection, which measured 0.0625 mm (62.5 µm), introduced a higher error for small prints.Therefore, strut models with denser strut distributions and smaller strut diameters, such as 600 12.5 and 600 15 , exhibited higher average errors (⩾10%).The SD for each model mostly fell within the range of 0.01-0.02mm, indicating the high stability of the 3D printing techniques.

The direction of strut insertion
Figure S2 shows two distinct types of plastic mold assemblies employed for efficient strut removal and the production of complete bioceramic scaffolds.The first assembly involved replacing the flat base (mold B) with a new base that matched the strut parameters of 800 10 before casting the HA slurry.The second assembly entailed inserting the struts into the already-cast HA slurry within the plastic mold.To ensure the struts could be released from the bioceramic sample, both types of DLP struts were precoated with a double layer of release agent consisting of petroleum jelly and lubricant.This coating prevented the HA bioceramics from adhering to the struts.Figures S2(a)-(c) depicts the outcomes of the first assembly.In this configuration, the HA slurry dried rapidly, and the fixed struts prevented the bioceramic sample from shrinking, leading to cracks throughout the sample.On the other hand, figures S2(d)-(f) illustrates the second assembly.Here, a plastic film was placed on the base plate to prevent the bioceramic sample from adhering to the plate, and the struts were inserted into the HA slurry.Due to the strut base reducing the drying rate of the HA slurry and the flexibility of the upper ends of the struts, the bioceramic sample could dry and shrink without sustaining damage.Thus, the second assembly type was deemed the appropriate molding technique for this study.

Shrinkage evaluation of dried and sintered HA scaffolds
The HA scaffolds were dried in an electronic moisture-proof box, as elucidated in section 2.3.It is noteworthy that when the samples were placed openly in the electronic moisture-proof box, the HA slurry deposited underneath did not dry efficiently due to faster moisture evaporation from the top compared to the bottom.To ensure a consistent drying rate for the entire sample and prevent sample breakage, the entire package (mold + strut + HA) was wrapped with plastic wrap to slow down the overall drying rate.Most of the water within the HA slurry was dehydrated after 48 h, during which PF-127 effectively acted as a binder through water evaporation, resulting in close bonding between the HA particles.Figure 3 displays the appearance of the HA scaffolds after drying and sintering.It is worth noting that the bioceramics experienced volume shrinkage during the drying stage, which could cause the 3D-printed struts to bend.The stress generated by this bending can lead to defects in the samples, and cracks may form during sample drying or demolding, ultimately affecting the attainment of a complete shape after subsequent sintering.It can be observed that in the case of 600 15 , excessive compaction occurred between the struts, resulting in the formation of cracks during the drying process.Consequently, it was not possible to achieve a fully-formed scaffold.To prevent failure after sintering, it is suggested to avoid a spacing between the struts of less than 0.8 mm in this fabrication method or carefully consider the arrangement of the struts.The weight and diameter (Ø) of the HA scaffolds after drying and sintering were measured using an analytical balance and vernier caliper, respectively.The channel size and designed porosity were evaluated by optical microscope and ImageJ software, as summarized in table S4.In addition, the shrinkage percentages of the samples after drying and sintering, compared to the weight of the injected HA slurry and the diameters of the FDM mold and DLP strut (table S3), are presented in table S5.Negative values in table S5 represent a decrease in magnitude.The original casted samples were expected to have a weight of 1 g and a diameter matching that of the FDM mold, which measured 13.81 mm.After drying, the average weight and diameter of the samples were 0.531 g and 12.5 mm, respectively, indicating a reduction of 47% in weight and 9.5% in diameter.The weight loss during drying was primarily attributed to water evaporation, theoretically amounting to 43.6%.Regarding the channel size, the penetrating channels expanded after drying, with the HA scaffolds having a designed porosity of 10% exhibiting the lowest expansion rate.This expansion could be due to the double-layer coating of release agent on the surface of the struts, leading to an increase in diameter.Additionally, a denser distribution of struts may contribute to an uneven application of the release agent.The difference in designed porosity before and after drying was not significant, ranging approximately from −1.2% to 0.4%.
After sintering, the average weight and diameter of the bioceramic samples decreased to 0.387 g and 9.31 mm, respectively, representing a reduction of 61.3% and 32.6% compared to the original casted sample.The weight loss included the removal of water and PF-127, which theoretically accounted for 54.5%.The additional weight loss of around 6.8% can be attributed to residues on the mold or struts.This explains why the 0 0 sample (without designed channels) exhibited less weight loss, as it did not have residues on the struts.When comparing samples with and without channels, the rates of diameter shrinkage were approximately the same.In terms of channel size, the three initially designed channel sizes exhibited an average inward shrinkage to 482 ± 7 µm, 640 ± 5 µm, and 811 ± 15 µm, respectively, after the sintering process.Figures 3(d)-(f) show SEM images of the channels on the sintered HA scaffolds, revealing well-defined circular shapes.These channel sizes ranged from 400 to 800 µm, which have been demonstrated to be suitable for promoting bone growth [80].The designed porosities after sintering showed no significant deviation (−0.9%-1.3%)from the initially designed values.The results indicate a stable shrinkage tendency of the samples after sintering.This should be considered in future production of various bioscaffolds to ensure they closely match the expected dimensions.

Material analysis, cell viability and cell adhesion
Phase identification was carried out by XRD to determine the mineralogy of the raw HA powder, sintered HA powder and sintered HA scaffold (ground into powder), as depicted in figure 4(A).The XRD patterns of all the samples were referenced to the inorganic crystal structure database (ICSD) standard for HA (collection code: 56306 [81]).The peak intensity of the sintered HA samples was significantly higher than that of the raw HA powder, indicating that the crystallization occurred and hence the crystallinity was increased during the sintering process [82].The crystallite sizes were calculated to be 22 nm for the raw HA powder and approximately 100 nm for both the sintered HA powder and the sintered HA scaffold based on the Scherrer equation.Moreover, the XRD pattern of the sintered HA scaffold matched the standard HA spectrum without any additional peaks, confirming the complete removal of PF-127 after sintering.
The functional groups of raw HA powder, sintered HA powder, and sintered HA scaffold (ground into powder) were characterized using FTIR, as shown in figure 4(B).All the samples had the characteristic bands of typical HA at 474-632, 961-1088, 2007-2153 and 3571 cm −1 , which are in good agreement with reported literature [83][84][85][86][87][88][89], as explained in the followings.The peak at 474 cm −1 represents the doubly degenerate ν 2 O-P-O bending mode, while the bands at 571 and 601 cm −1 correspond to the bending mode of triply degenerate ν 4 O-P-O bonds.The bands at 961 and 1043-1088 cm −1 are assigned to the nondegenerate symmetric ν 1 and triply degenerate asymmetric ν 3 P-O stretching mode, respectively.A group of bands with low intensity at 2007-2153 cm −1 is attributed to the combination and the overtone of ν 1 and ν 3 phosphate modes.The vibrational and stretching mode of hydroxyl groups are confirmed by the bands at 632 and 3571 cm −1 , respectively.These characteristic peaks became more pronounced after the sintering treatment, indicating the successful crystallization of HA.The broadband at 3300-3500 cm −1 implies the presence of water molecules absorbed by the HA lattice, and the peak at 1643 cm −1 also suggests the ν 2 bending mode of absorbed water [84,86,89].The bands at 875, 1421 and 1472 are emblematic of carbonated apatite, which can be ascribed to the substitution of CO 3 2− ions for PO 4 3− ions [83,84,89,90].Compared to the sintered HA powder and scaffold, the raw HA powder contained a large amount of water molecules and CO 3 2− ions, which may be due to its smaller crystal size being more susceptible to impurities.It has been reported that small crystal size is more prone to incorporating CO 3 2− and H 2 O into PO 4 3− sites [91,92].Thus, the bands representing water molecule and carbonated apatite were significantly reduced after the sintering treatment, supporting the removal of absorbed water and CO 3 2− ions.Furthermore, the HA scaffold did not exhibit any additional bands compared with the sintered HA powder, confirming the complete removal of PF-127 by sintering.However, a peak at 1384 cm −1 was observed in all the samples, which could be attributed to an unknown contamination.
The SEM images of different sections of the sintered 600 10 HA scaffold are presented in figure 5. From the top surface image (figure 5(a1)), it can be observed that the circular-shaped penetrating channels remained intact, while minor defects may have occurred during the removal of the struts.Scratches were found on the top surface due to sanding with sandpaper prior to sintering.Notably, distinct stripes were observed on the lateral surface of the bioscaffold (figure 5(b1)).This can be attributed to the layered structure of the plastic outer mold prepared by FDM, leading to a layered appearance on the lateral surface of the bioscaffold after molding.The advantageous effects of these layers for cell adhesion have been highlighted by Han et al [93] that untreated 3Dprinted Polyether-ether-ketone (PEEK) and carbon fiber reinforced PEEK (CFR-PEEK) samples exhibited significantly higher cell density and greater cell coverage on their surfaces compared to sandblasted and polished samples.Furthermore, cells attached in lines within the valleys of the FDM-manufactured surfaces.In figure 5(c1), the sample was dissected in half to examine the inner surface of the penetrating channels, revealing a clear laminar pattern.This pattern is a result of the stacking structure of the DLP-made struts, where the layer spacing on the channel surface is smaller than that on the lateral surface due to the different layer thickness of DLP and FDM (0.05 mm and 0.2 mm, respectively).The cross-sectional images of the bioscaffold (figure 5  To further examine cell viability, adhesion, and distribution of MG-63 on the lateral surface and within the penetrating channels after a 14 day incubation period, a live/dead staining assay was conducted, as presented in figures 5(g) and (h).As observed in figure 5(g), numerous MG-63 cells adhered to the lateral surface of the 800 10 sample, particularly within the valleys formed by the FDM-printed outer mold.In comparison to the conventional FDM-printed PEEK composite reported by Han et al [93], the valleys formed on the HA scaffold's surface during the casting process in this study were notably wider due to the FDMprinted outer mold.Consequently, they can accommodate a greater number of bone cells.In figure 5(h), MG-63 cells also adhered to the inner walls of the penetrating channels.This adhesion was facilitated by the layered structure formed on the inner walls due to the DLP-printed struts.Cells adhered abundantly within these valleys as well.The in vitro testing results confirmed the non-toxicity of the HA scaffold in this study and highlighted the enhanced adhesion of a larger population of bone cells facilitated by the surface layers.

Physical and mechanical assessment
The density, porosity, hardness and compressive strength of the sintered HA scaffolds are presented in table 1.The 600 15 sample was excluded from this study due to the close proximity of the struts during the fabrication process, leading to severe sample fracture, as shown in figure 3(a3).The bulk density (D b ) after sintering was generally in the range of 2.36-2.46g cm −3 , while the apparent density (D a ) ranged from 2.99 to 3.07 g cm −3 .The lower values of both D b and D a compared to the theoretical density of HA (∼3.16 g cm −3 ) [81] are attributed to the presence of pores, which can be explained as follows.Although PF-127 was used as a binder, ensuring strong adhesion of HA, there were no additional methods employed to compress the scaffold during preparation, except for centrifugal defoaming under vacuum.As a result, pores were formed during the sintering process when the binder burned out.
According to the causes and states of pore formation, the porosity in HA scaffolds can be classified as open porosity (P o ), close porosity (P c ) and designed porosity (P d ).P o and P c are present within the bulk of the HA scaffold, where open pores are connected to the external environment, while closed pores are completely sealed within the material.The presence of these two types of pores is attributed to the manufacturing process, where the sample was vacuum defoamed without pressurization, resulting in additional open and closed pores, as shown in figures 5(c1) and (d1).The values of P o and P c can be determined using the formulas provided in section 2.5, based on the dried sample weight (m 1 ), apparent density (D a ), bulk density (D b ), and theoretical density of HA (D m ).In this study, the range of P o for the HA samples was found to be 16.35%-20.38%,while P c ranged from 1.77% to 3.68%.This indicates that the majority of pores within the samples were open pores.It is worth noting that the average value of the sum of P o and P c , approximately 21.04 ± 1.03%, is very close to the theoretical ratio of PF-127 in the dried HA samples, which is 19.35%.This implies that there were not excessive additional pores formed and suggests that the centrifugal defoaming technique can significantly increase the density of the samples.Designed porosity (P d ) refers to the penetrating channels created by the DLP-printed struts.In this study, P d of the HA scaffolds was slightly larger than the original design, which may be attributed to the doublelayer coating of release agent applied to the strut surface, as elucidated in section 3.4.The results show that samples with larger P d tended to have smaller values of P o + P c , approaching the theoretical porosity of 19.35%, regardless of pore size.This may be due to the larger P d , indicating a higher amount of struts, making it easier for gases within the material to escape through the gaps between the struts and slurry during vacuum defoaming.In addition, the total porosity (P t ) can be obtained by summing the P o , P c , and P d , ranging from 21.98% for sample 0 0 -32.76%-36.71%for the HA scaffolds.Apart from the ability to adjust the design parameters of the pores according to specific requirements, this manufacturing method resulted in the creation of numerous small pores within the material.These small pores have the potential to enhance nutrient absorption and increase the sample's degradability.Therefore, this bioscaffold could be considered as a drug delivery medium since the presence of these pores can contribute to faster initial drug release [95].While obtaining a porous structure, the centrifugal vacuum defoaming process, employed in this study, harnessed centrifugal forces to compress the bioceramic samples while simultaneously removing nanoscale bubbles from the slurry.This procedure maximized the sample's density, significantly enhancing its strength.Remarkably, this was achieved without the need for physical compression molds, as required by uniaxial pressing.Consequently, this method successfully met objectives for rapid manufacturing and cost-efficiency.
The hardness of the samples was evaluated using a force of 0.5 kg, with three measurements taken at different positions on each sample and the average hardness values calculated.The hardness results for the different HA scaffolds are presented in table 1 and figure 6(a).It can be seen that the scaffold with a channel size of 1000 µm had higher hardness, while scaffolds with channel sizes of 600 µm and 800 µm showed lower hardness.This could be ascribed to the fact that larger channel sizes, under a fixed designed porosity, resulted in a lower number of channels, thereby reducing the likelihood of material failure during hardness measurements.However, in this study, no direct correlation was found between hardness and the values of P o , P c , and P d .This might be due to the fact that surface hardness properties did not necessarily reflect the overall structural characteristics of the material.
The compressive strength of the HA scaffolds is shown in table 1 and figure 6(b).As can be seen that despite the relatively small differences in total porosity (P t ) among the samples, their compressive strength varied significantly due to the differences in channel size and designed porosity.With an increase in channel size, the compressive strength decreased.This can be attributed to the fact that, at the same porosity, smaller and more dispersed channel sizes allowed for a more uniform distribution of stress within the material, reducing stress concentration issues.Additionally, smaller channel sizes may facilitate the rapid filling of powder particles into cracks upon fracture, thereby enhancing the compressive resistance of the material.However, under the same channel size, a higher designed porosity led to lower compressive strength, indicating that the designed structures of penetrating channel had a significant impact on the mechanical properties of the HA scaffold.Among all the samples, the 600 10 scaffold displayed the highest compressive strength, reaching 238.9 MPa, followed by the 800 10 scaffold with a strength of 201.0 MPa.To further investigate the influence of bioscaffold height on mechanical properties, the compressive strength of three samples, namely 600 10 , 800 10 , and 1000 10 , was measured after increasing their heights from approximately 2.8 mm to around 4.2 mm, as shown in figures 6(c) and (d).
The maximum stress values for these three samples were 40 MPa for 600 10 , 73 MPa for 800 10 , and 61 MPa for 1000 10 .It can be observed that the compressive strength decreased as the sample height increased, which is possibly due to the presence of additional defects.In the case of the 600 10 sample, the distribution of DLP-printed struts was denser, making it more prone to defects when the sample height was increased, resulting in a greater reduction in compressive strength compared to the 800 10 and 1000 10 samples.
Table 2 compares the compressive strength and compressive modulus of the 4.2 mm high HA scaffolds (600 10 , 800 10 , and 1000 10 ) from this study with those of human cortical bone, cancellous bone, and bioceramic scaffolds fabricated using 3D printing methods as reported in the literature.It can be observed that the channel sizes of the three HA scaffolds in this study were reduced to an average of 481 µm, 640 µm, and 811 µm, respectively, which are all larger than those of cancellous bone.In addition, the porosity of these scaffolds fell within the range of cancellous bone (30%-95%).According to Gibson's report [96], the microstructure of cancellous bone can be classified into asymmetric and columnar structures based on the direction of force loading.If the stress distribution within cancellous bone is intricate, it results in a correspondingly intricate and markedly asymmetrical trabecular network.However, in bones subjected to primarily uniaxial loading, such as vertebrae, trabeculae normally adopt a columnar structure characterized by cylindrical symmetry.These bone columns are aligned vertically, thereby imparting relatively high strength and stiffness in the loadbearing direction, while exhibiting lower strength and stiffness in the perpendicular directions.Since porosity levels significantly influence mechanical properties of bioscaffolds, the compressive strength and modulus of human cancellous bone at 30%-70% porosity were calculated based on data from the literature to compare the mechanical properties of bioscaffolds in this study with natural bone at the same porosity level, as shown in table 2. In terms of compressive properties, the strength of these HA scaffolds is significantly higher than that of the columnar structure of cancellous bone but falls within the range of the asymmetric structure.The compressive modulus of the bioscaffolds is comparable to that of the columnar structure of cancellous bone but lower than that of the asymmetric structure.These results indicate that the bioscaffolds in this study can withstand the same maximum stress as the asymmetric structure at an equivalent porosity level.However, the lower compressive modulus exhibited by these bioscaffolds implies greater deformation compared to the asymmetric structure.This also suggests that the bioscaffolds in this work have a greater capacity to absorb energy per unit volume, as indicated by the area under the stress-strain curve, highlighting their superior toughness compared to cancellous bone.The differences of compressive modulus between the HA samples can be attributed to the variation in channel size, amount and distribution.
Table 2 also compares bioprinted musculoskeletal tissues from the literature, including bone, cartilage, tendon, muscle and osteochondral units, fabricated by various 3D printing techniques, such as FDM, inkjet, DLP and casting.At present, all 3Dprinted bioscaffolds are unable to meet the compressive strength and modulus required by cortical bone.However, a few bioscaffolds manage to achieve the mechanical properties needed for cancellous bone.For instance, Zhang et al [20] and Filippov et al [61] fabricated bioscaffolds with porosities of approximately 60%-65%, composed of HA and calcium pyrophosphate using FDM and DLP technologies, respectively.Both of these exhibited a compressive strength of 5 MPa, which fell below that of natural bone at corresponding porosities.In an attempt to enhance the compressive strength, Zhang et al [20] incorporated 50% and 70% PLLA into the HA, resulting in an increase in compressive strength to 14.2 and 29.7 MPa, respectively.However, the compressive modulus decreased from 0.79 GPa to 0.04 GPa, ultimately leading to overall mechanical properties lower than those presented in this study.Furthermore, some high-strength bioscaffolds fabricated through FDM [97] and DLP [48] methods were reported to have compressive strengths ranging from 16 to 45 MPa, meeting or even exceeding the strength requirements of cancellous bone at the same 50% porosity level.However, their compressive modulus still fell short compared to cancellous bone, indicating room for improvement in this aspect.Lin et al [98] utilized FDM in combination with negative thermoresponsive hydrogels and silicone oil to enhance the density and compressive strength of the HA/β-TCP bioscaffold.Compared to this literature, the 800 10 and 1000 10 samples in our work, even at a higher porosity level, exhibited superior compressive strengths to the samples enhanced by silicone oil.Liang et al [99] utilized DLP method to fabricate HA scaffolds with a compressive strength of 22.5 MPa and a modulus of 0.4 GPa at a porosity of 70%, closely resembling the mechanical properties of cancellous bone.However, their cell viability decreased to 70% compared to the control group after 7 d of cell culture, possibly due to residual toxicity from the DLP process.In addition, this method involved an extended debinding and sintering process lasting up to 52 h, which may result in higher production costs.Currently, the compressive strength and modulus of bioscaffolds produced using the casting method tended to be relatively weak.In summary, the application of the centrifugal vacuum defoaming method in the present study has proven to be highly effective in substantially enhancing the density and compressive strength of the bioscaffold through an efficient sintering procedure, a technique not employed in prior literature.Among the HA samples, 800 10 stands out as exceptionally well-suited for prospective biomedical applications due to its superior compressive strength  and a modulus that closely aligns with cancellous bone, which helps prevent stress shielding and bone atrophy resulting from modulus mismatch with native bone [100].

In vitro degradation evaluation
Figure 7 presents the results of an in vitro test conducted by immersing the HA scaffolds in SBF.The samples with a designed porosity of 10% were selected for the in vitro test due to their higher mechanical properties (figure 6).show the magnified SEM images of these three areas at 20 000x magnification.It can be observed that both the 800 10 and 1000 10 samples exhibit abundant fuzzy deposits on the inner walls of the channels, while the 600 10 sample retains a smooth surface with HA grains.In figure 7(g), the EDS analysis of the deposits on the inner walls of the 1000 10 sample (indicated by a yellow arrow) reveals a significant presence of calcium (Ca), phosphorus (P), and oxygen (O) elements, confirming the composition as calcium phosphate (apatite, denoted as CaP).The formation of CaP deposits induced by SBF immersion has been validated in other literature [103].Zadpoor [104] compiled 33 experimental results and concluded that the ability of in vitro apatite formation has proven to be effective in predicting the in vivo performance of biomaterials.From figures 7(a)-(f), it can be observed that under the same designed porosity, HA scaffolds with larger channel diameters exhibit thicker and larger deposits of CaP.This is likely due to the larger channel diameter allowing easier access of SBF, promoting faster and more complete ion exchange reactions.Additionally, the rough surface texture formed by the 3D-printed layers in the channels may facilitate the deposition of CaP and offer the potential for improved attachment of bone cells [105][106][107][108]. Figure 7(h) presents the weight loss data of the 600 10 , 800 10 , and 1000 10 HA scaffolds after immersion in SBF for 7, 14, 21, and 28 d.The weight loss fluctuates within the range of ±1%, which is attributed to the degradation of the samples and the deposition of CaP due to the reaction with SBF.The low degradation rates of the three samples can be ascribed to the strong adhesion between HA grains.The 600 10 sample exhibits a weight loss of approximately 0.5% within 28 d, primarily owing to the degradation of HA rather than CaP deposition.The 1000 10 sample shows a weight increase of 0.5% at the 14th day, but a weight decrease of 0.5% compared to the original weight after 28 d.This suggests that although initial CaP deposition was higher due to the larger channel diameter, it was also more susceptible to degradation, leading to weight loss in the later stages.The weight of the 800 10 sample stabilizes at a steady increase of 0.5% after the 14th day, indicating good CaP deposition and low degradation rate in this sample with a specific channel diameter.

Conclusions
This study developed a hybrid 3D printing methodology that combines FDM and DLP techniques for the bioscaffold fabrication, enabling efficient mass production.The FDM method was employed to fabricate the plastic outer molds, while the DLP technique was used to print struts required for creating penetrating channels.By integrating these components, high-strength bioscaffolds with different channel sizes and porosities, based on HA, were realized using the slurry casting method.Further densification was achieved through the application of centrifugal vacuum defoaming.This work conducted extensive investigations into photosensitive resin selection, mold and strut design, and 3D-printing parameters to optimize the bioscaffold performance.XRD and FTIR analyses confirmed the successful crystallization of HA during sintering.SEM images unveiled a microstructure comprising tightly bound HA grains and a layered structure evident on both the lateral scaffold surface and the channel inner wall.Physical and mechanical evaluations revealed that HA scaffolds in this study had overall porosities in the range of 32.76%-36.71%and exhibited increasing compressive strength with the decrease in designed porosity and channel size.Compared to human cancellous bone with compressive strengths ranging from 13 to 70 MPa and moduli from 0.8 to 8 GPa at a similar porosity level of 30%-40%, the HA scaffolds exhibited robust strengths ranging from 40 to 73 MPa, while maintaining lower moduli of 0.7-1.23 GPa.These characteristics rendered them ideal for cancellous bone repair by effectively addressing concerns such as stress shielding and bone atrophy.In vitro assessments, including biodegradation and cytocompatibility testing, demonstrated favorable CaP deposition and low degradation rates due to the strong adhesion between HA grains, along with impressive bone cell viability and attachment caused by the layered surface structure.These findings suggest that this innovative approach for bioscaffold fabrication holds promise in future biomedical applications.
(d1) and (d2)) show tightly bound HA grains with an approximate grain size of 1 µm after sintering.Clear micro-pores are visible within the interior of the channels and on the fracture surfaces, as depicted in figures 5(c1) and (d1), resulting from the removal of PF-127.Both the lateral (figure 5(b2)) and channel surface (figure 5(c2)) exhibit compact grains, indicating substantial densification of the HA bioceramics achieved through vacuum centrifugal defoaming.In contrast, the top surface appears relatively loose (figure 5(a2)), which can be attributed to the sanding process with sandpaper.

Figure 5 (
e) illustrates the cell viability of the 800 10 HA scaffold, which was determined by MTT assay.The OD was compared against the positive control SDS and negative controls (MG-63 cells and Al 2 O 3 ), indicating the non-cytotoxic nature of the HA scaffold.The negative control comprising only MG-63 cells was considered as 100%.In addition, SEM imaging was performed to investigate MG-63 cell adhesion and morphology after 7 d of incubation, as depicted in figure5(f).The SEM micrograph shows that the cells adhered effectively, exhibiting a flattened morphology and demonstrating strong adhesion to the scaffold's surface.The attached cells displayed the characteristic irregular morphology of MG-63[94].

Figure 6 .
Figure 6.(a) Hardness and (b) compressive strength of sintered HA scaffolds; (c) correlation between sample height and compressive strength; (d) compressive stress at sample height of around 4.2 mm.

Table 2 .
Comparison with literature on human bone and other bioprinted scaffolds.
a To compare with the findings of this study, the compressive strength and modulus for porosities of 30%-70% were calculated based on data from the literature.b Tensile testing was conducted in this work.