3D printing technology in hepatobiliary surgery education: a systematic review

Abstract

Background

Three-dimensional printing (3DP) is increasingly applied in surgical training; however, its effectiveness and reliability in hepatobiliary surgery education remain uncertain.

Objectives

To systematically review the literature on the application and effectiveness of 3DP in hepatobiliary surgical training and anatomical education.

Method

PubMed, Embase and the Cochrane Library were searched for publications dated 1 January 1995, to 19 January 2025. Two investigators independently performed the literature search and screened eligible studies that reported the use of 3DP models for surgical training or anatomical education. Study quality was assessed using the Medical Education Research Study Quality Instrument (MERSQI), and levels of evidence were graded using the Oxford Centre for Evidence-Based Medicine (OCEBM) framework. Extracted data included study design, participant numbers, type of surgical training, type of anatomical education, and reported subjective and objective outcomes.

Results

Twenty-two studies investigated the application of 3DP in clinical surgical training and 12 studies focused on anatomical education. Laparoscopic procedures were the most frequently reported training context (59%). In anatomical education, tumor localization was the most common task (50%). Ten studies reported that 3DP significantly improved operative performance, reducing operation time by 34–54.6% in laparoscopic and choledochoscopic training. In anatomical education, 3DP models improved students’ performance by 57–70% in liver segment memorization.

Conclusion

3DP models demonstrated substantial value in both surgical training and anatomical education for hepatobiliary surgery. Although still developing, 3DP has already achieved broad acceptance among surgeons and medical students as an effective adjunct for skill acquisition and anatomical understanding.

HIGHLIGHTS

1. 3DP models are applied in both surgical training and anatomical education, demonstrating broad educational value in hepatobiliary surgery.

2. 3DP models significantly shorten operation time and improve surgical skills in hepatobiliary training.

3. In anatomical education, 3DP models enhance tumor localization and outperform CT/VR models.

Introduction

Hepatobiliary surgery is one of the most technically demanding subspecialties within general surgery. Considerable anatomical variability exists in the hepatic artery, portal vein, biliary tract and hepatic veins, all of which complicate surgical approaches. These complexities, coupled with high operative risks, technical difficulty, frequent postoperative complications and prolonged training requirements, make hepatobiliary surgery particularly challenging. Traditional teaching tools – such as atlases, videos and CT imaging – remain widely used. However, due to the intricate liver anatomy and marked interindividual differences, these conventional approaches often fail to provide learners with an adequate understanding of hepatobiliary structures. Consequently, educational outcomes are suboptimal, and the acquisition of surgical skills remains difficult.

Additive manufacturing, commonly referred to as three-dimensional printing (3DP), enables rapid fabrication of high-fidelity anatomical models through specialized printers [1,2]. With ongoing advances in printer hardware, software and material diversity, 3DP has gained broad applicability across medicine. Historically, medical students’ comprehension of radiologic images has been limited to two-dimensional representations in textbooks or on computer screens [3,4]. By contrast, 3DP allows direct conversion of medical imaging data, such as CT scans, into accurate three-dimensional structures [5,6]. This capability makes 3DP a powerful adjunct or alternative to traditional teaching methods in medical education [7,8]. Improvements in resolution, processing speed and material costs have steadily enhanced its accessibility [9,10]. Modern 3DP models not only achieve higher fidelity and anatomical precision but also simulate human tissues with notable realism [11], thereby holding strong potential for surgical simulation and skills training.

Evidence from multiple randomized controlled trials [12–15] across diverse surgical disciplines demonstrates that 3DP models facilitate both early skills acquisition and advanced procedural training. Studies in neurosurgery, otolaryngology, gastrointestinal surgery and neurology have shown measurable improvements in operative efficiency when trainees practiced with 3DP-based simulations [3,16–19]. Within hepatobiliary surgery, 3DP models have been applied in laparoscopy, endoscopy, preoperative planning, anatomical teaching, tumour localization and patient education. High-quality, reproducible simulation is critical for building both decision-making capacity and technical proficiency. Yet, traditional training for surgical residents faces increasing constraints, including restricted work hours and expanding subspecialization. Integration of 3DP models into hepatobiliary training may help address these challenges by providing additional opportunities for safe, hands-on practice [20–23]. By enabling detailed three-dimensional visualization of organ anatomy, 3DP represents a potentially transformative tool in surgical education [24–27]. While surgical training has historically relied on apprenticeship models with direct observation in the operating theatre, contemporary studies highlight the value of simulation in providing safe, effective and transferable procedural training [28–30]. Nonetheless, limitations remain: high-fidelity models are costly, concerns persist regarding accuracy and standardized manufacturing protocols are lacking.

Prior systematic reviews [31,32] have primarily addressed the application of 3DP in anatomical replication and preoperative planning. Christou et al. [33] specifically explored 3DP and 3D bioprinting in hepatocellular carcinoma, while Alkhouri et al. [34] summarized its use in paediatric liver disease and transplantation. However, no systematic review has comprehensively examined the role of 3DP in hepatobiliary surgical education, particularly from the perspective of skill acquisition and anatomical knowledge enhancement aimed at improving both technical performance and cognitive understanding.

The present study systematically reviews published evidence to address two objectives: To evaluate the role of 3DP in improving surgical skills relevant to hepatobiliary surgery and to assess the contribution of 3DP to anatomical education in hepatobiliary surgery.

Materials and methods

The protocol for this review was pre-registered in the PROSPERO database and conducted in strict accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [35] guidelines and the Assessing the Methodological Quality of Systematic Reviews (AMSTAR) framework [36].

Search strategy

A systematic literature search was conducted in PubMed (all fields), Embase (all fields) and the Cochrane Library (all text) by J Lin and W Li. The search strategy employed the keywords ‘3D model’, ‘liver/hepatic’ and ‘teaching/education/sim’. The complete strategy is provided in Appendix 1 in Supplemental data. The search was restricted to publications in English from1 January 1995, to 19 January 2025, with the final search executed on 19 January 2025. Additionally, the reference lists of included articles were manually screened to identify further eligible studies.

Study selection

To comprehensively capture empirical investigations on 3DP applications in hepatobiliary surgery education and training, a predefined eligibility framework was established. Studies were included if they (1) constituted primary research with original data; (2) explicitly documented the use of 3DP in surgical education and (3) involved medically trained participants. Exclusion criteria comprised (a) non-empirical publications such as editorials, commentaries and opinion pieces, and (b) secondary literature including systematic reviews and meta-analyses, in order to preserve the integrity of primary data synthesis. This step was completed by J Lin and W Li.

Quality assessment

Two assessors (X Yu and Y Zhang) independently evaluated study quality, resolving discrepancies through consensus. Owing to heterogeneity in study design, multiple appraisal instruments were employed. Levels of evidence and recommendations were classified using a modified Oxford Centre for Evidence-Based Medicine (OCEBM) educational framework, where Grade A represents the highest and Grade D the lowest level of recommendation [37] Methodological rigor was further assessed using the Medical Education Research Study Quality Instrument (MERSQI), which evaluates the quality of quantitative medical education research [38].

Data extraction and analysis

Data extraction was performed by J Lin and W Li with a standardized template to ensure consistency across studies. Extracted parameters included: (1) bibliographic information (authors, year, country); (2) methodological characteristics (design, sample size); (3) technical specifications (materials, costs); (4) educational components (tasks performed with the models) and (5) validation outcomes, both objective (fidelity assessments, performance measures) and subjective (expert/trainee feedback). For studies with incomplete technical details on 3DP hepatobiliary models, missing information was inferred from supplementary materials or cross-referenced with methodological precedents in related surgical education literature.

Results

The initial searches yielded 673 studies and 160 articles were excluded as duplicates; a further 436 were excluded following screening of abstracts and 5 were unable to retrieved. A total of 34 studies met the inclusion criteria: 22 evaluated 3DP applications in clinical surgical training and 12 in anatomical education (Figure 1). Of these, 19 studies originated from China.

Figure 1.

PRISMA statement flow diagram of the search and study selection process. PRISMA, Preferred reporting items for systematic reviews and meta-analyses.

Quality assessment

The included studies spanned OCEBM evidence levels 1–4, with 10 studies [3,39–47] rated Level 1b, 7 studies [22,48–53] Level 2b, 1 study [54] Level 3a, 2 studies [55,56] Level 3b, 9 studies [57–65] Level 4 and 5 studies [21,23, 66–68] Level 5. As detailed in Supplementary Table 1, the mean MERSQI score was 11.6 (range: 7–13.5), with complete inter-rater agreement (100%), reflecting a consistently low risk of bias.

3DP model preparation information

Table 1 summarizes the characteristics of the included 3DP models. Most models were reconstructed from CT scans, and all liver models were segmented from CT-based structures. A few studies also reported the use of MRI [65] or MRCP [46,57,67] as imaging sources. Approximately 38% of models required manual refinement during reconstruction, and 3 studies [57,58,61] specifically described manual segmentation. In terms of fabrication, 56% of models were constructed in-house, while 26% were outsourced.

Table 1.

Preparation information for 3DP models.

Authors	Materials	Data source	Reconstruction software	Modeling approach	Self-made or outsource	3D printer
Dhir et al. [57], 2015	Polycarbonate	MRCP	CAD	Manual modeling	Self-made	Viper SI2 stereo-lithography system
Holt et al. [58], 2015	Polymer resin, Silicone	N/A	CAD	Manual modeling	Self-made	Connex 260v
Burdall et al. [59], 2016	Silicone	Digital hepatic anatomy images	CAD	N/A	Outsource	3D systems project 660pro
Li. et al. [3], 2018	N/A	CT	Hisense computer-assisted surgery system	Software modeling	Self-made	Stratasys Connex3350
Wei et al. [39], 2019	N/A	CT	N/A	N/A	Outsource	N/A
Kwon et al. [23], 2020	Silicone	CT	MeshLab MeshMixer	Software modeling	Self-made	3DM DW-06, 3DMaterials, Zeron2500
Wei et al. [66], 2020	N/A	CT	N/A	Software modeling Manual polishing	Outsource	N/A
Murillo et al. [67], 2021	Acrylonitrile butadiene styrene, P53 silicone rubber, Elmer’s slime	MRCP	CADD	Software modeling Manual polishing	Self-made	3D Zortrax
Wang et al. [60], 2022	N/A	CT	N/A	Software modeling Manual polishing	Outsource	N/A
Micallef et al. [61], 2022	Ecotough™ polylactic acid, Ecoflex™ 00–20 FAST silicone, Ease Release 200, SilcPig™ coloring	N/A	Fusion 360™	Manual modeling	Self-made	Ultimaker S5
Yu et al. [48], 2022	Silicone	CT	E3D digital medical modelling	Software modeling Manual polishing	N/A	N/A
Yang et al. [22], 2022	Silicone	CT	E3D digital medical modelling	Software modeling Manual polishing	N/A	N/A
Fechner et al. [55], 2023	Silicone-based soft resin, silicone	CT	Blender Fusion360	Software modeling Manual polishing	Self-made	Form 3B Creality CR 10S Pro
Xia et al. [49], 2023	Silicone	CT	Magic 24 Zbush	Software modeling Manual polishing	Self-made	FDM
Shen et al. [50], 2023	Silicone	CT	N/A	N/A	Outsource	N/A
Lu et al. [62], 2023	ACGE	CT	Yorktal digital medical imaging	Software modeling	Self-made	Prismlab, RP400
Shen et al. [40], 2024	Silicone	N/A	N/A	N/A	Outsource	N/A
Elisei et al. [63], 2024	FDM/PLA	CT	InVesalius Fusion 360	Software modeling Manual polishing	Self-made	Fortus 380 mc
Gu et al. [51], 2024	Silicone	CT	N/A	Software modeling Manual polishing	Outsource	Silplot-S400
Yang et al. [41], 2024	Agilus30, Vero Vivid, Vero Ultra	CT	Magics	Software modeling Manual polishing	Self-made	Stratasys J850
Aranovich et al. [64], 2024	Thermoplastic elastomer TPU-95	CT	N/A	N/A	N/A	N/A
Cao et al. [65], 2024	Silicone	MRI	E3D digital medical modelling	Software modeling Manual polishing	Self-made	SilplotS400
Kong et al. [42], 2016	ZP150	CT	N/A	N/A	N/A	N/A
Kong et al. [43], 2016	ZP150, transparent jelly wax	CT	Geomagic 12	Software modeling Manual polishing	Self-made	Spectrum ZTM 510
Streba et al. [21], 2018	nGenN/AFlex black flexible filament	CT MRI	Meshmixer	Software modeling Manual polishing	Self-made	Tiertime
Yang et al. [44], 2019	VeroClear™, RGD720	CT	N/A	Software modeling	Self-made	Stratasys J750TM
Chedid et al. [45], 2020	N/A	N/A	N/A	N/A	N/A	N/A
Bati et al. [46], 2020	N/A	MRCP	Analyze 12.0	Software modeling	Self-made	Mass Portal Pharaoh xd 20 Form Labs2
Huettl et al. [47], 2021	Polyurethane rubber, acrylonitrile butadiene styrene	CT	Synapse 3D	N/A	Self-made	Cella Medical Solutions
Lopez-Lopez et al. [52], 2021	N/A	N/A	3D-MSP	N/A	N/A	N/A
Song et al. [54], 2023	Vero Family, photocurable resin	CT	N/A	Software modeling	Outsource	Stratasys J826
Cheng et al. [68], 2023	Photosensitive resin	CT	E3D digital medical modeling	Software modeling	Self-made	SL600
Shahbaz et al. [56], 2023	RGD720	Digital human specimen	N/A	N/A	Outsource	J401Pro
Bao et al. [53], 2023	Photosensitive resin, VeroCyanV, VeroYellowV, VeroMagentaV, VeroPureWhite, VeroUltraClea	CT	N/A	Software modeling	Self-made	Stratasys J850

Note: 3D: three-dimensional; 2D: two-dimensional; 3DP: three-dimensional printing; CT: computed tomography; MRI: magnetic resonance imaging; N/A: not available.

Surgical training

A total of 22 studies evaluated the role of 3DP models in surgical training (Table 2). Training tasks most frequently included bilioenteric anastomosis (27%) and pancreaticojejunostomy (18%). Of these studies, half (n = 11) reported fidelity scores ≥4/5, and 18 (81.8%) indicated that participants rated 3DP models as useful for skill acquisition (scores ≥4/5). Quantitative outcomes showed that 3DP training reduced operation time by 34-55% and improved OSATS scores by 33–73%. Comparative studies further highlighted these benefits: Li et al. [3] demonstrated that residents trained with 3DP models achieved higher anatomical recognition scores than those using images (57 vs. 16), while Shen et al. [50] found that residents trained with 3DP models performed better than those using TLCST, with shorter operation times (20.8 vs. 25.8 min) and higher performance scores (28.8 vs. 19.7).

Table 2.

Application of 3DP models in surgical training.

Authors	Participants	Allocation	Task	Fidelity	Self-evaluation usefulness	Validity	Others
Dhir et al. [57], 2015	20 endosonographers	No	Endoscopic ultrasonography-guided biliary drainage	Overall realism: 13/15, Puncture sensation: 4/5	4/5	N/A	N/A
Holt et al. [58], 2015	16 endoscopists	No	Endoscopic ampullectomy	Overall realism: 3.2/5	N/A	Core procedural steps: 3.1/5	Self-evaluation self-confidence improvement: 2.2→2.9/5
Burdall et al. [59], 2016	20 senior pediatric surgical trainees	No	Laparoscopic choledochal surgery	N/A	7.4/10	N/A	Recommendation: 100%
Li. et al. [3], 2018	20 residents	3DP group, images group	Choledochoscopy techniques	Anatomical accuracy: 4.5/5	4.7/5	Operation time:3DP 29→12mins Anatomic recognition accuracy: 3DP > Images (57 > 16)	N/A
Wei et al. [39], 2019	3 surgeons	2 rounds 3DP training group, 1 round 3DP training group, 0 round 3DP training	Laparoscopic pancreaticojejunostomy	N/A	N/A	Operation time: 2 rounds < 1 round < 0 round(20 < 30 < 40 mins) Operation level: 2 rounds > 1 round > 0 round	N/A
Kwon et al. [23], 2020	/	No	Endoscopic retrograde bilioenteric anastomosis	N/A	N/A	N/A	N/A
Wei et al. [66], 2020	3 residents 3 fellows	No	Laparoscopic bilioenteric anastomosis	N/A	N/A	N/A	N/A
Murillo et al. [67], 2021	13 practitioners	No	Laparoscopic cholecystectomy	Overall realism:4.5/5 Texture:4.5/5	N/A	N/A	Recommendation: 90% Durability: 4.3/5
Wang et al. [60], 2022	1 junior surgeon	No	Laparoscopic pancreaticojejunostomy	N/A	N/A	OSATS score: 15→26, Operation time: 1,734→1,142 s	N/A
Micallef et al. [61], 2022	6 surgeons 2 residents 1 medical student	No	Bile duct anastomosis	Anatomical accuracy: 3.56/5	4.5/5	N/A	Feasibility:3/5
Yu et al. [48], 2022	5 attendings 5 fellows 5 residents	Attendings group, fellows group, residents group	Laparoscopic pancreaticojejunostomy	Overall realism: 3.96/5 Elasticity: 3.88/5 Tearability: 3.83/5	N/A	Operation time: residents > fellows > attendings (1254.8 > 797.8 > 569.2s) Operation scores: residents < fellows < attendings (14.4 < 17.2 < 18.8) N/ASA-TLX: residents > attendings > fellows (261.6 > 265.4 > 412.8)	Recommendation: 100%
Yang et al. [22], 2022	4 attendings 4 fellows 8 residents	Attendings group, fellows group, residents group	Laparoscopic pancreaticojejunostomy	Overall realism: 4.22/5 Tactile: 4.58/5 Breakthrough sensation: 4.5/5 Tearability: 3.75/5	N/A	Operation time: residents > fellows > attendings (106.2 > 62.5 > 37.8) Operation scores: residents < fellows < attendings (19 < 19.5 < 23)	N/A
Fechner et al. [55], 2023	20 residents	No	Percutaneous transhepatic cholangial drainage	Overall realism: 3.67/4 Anatomical accuracy: 3.67/4	3.55/4	KAP: 677→262 mGy/cm2 Puncture time: 16:15 → 7:42 min Fluoroscopy time: 175 → 52 s	Self-evaluation self-confidence improvement: 3.83/4
Xia et al. [49], 2023	5 attendings 5 fellows 5 residents	Attendings group, fellows group, residents group	Laparoscopic bilioenteric anastomosis	Overall realism: 4.17/5 Texture: 4.17/5 Tactile: 4.00/5	4.83/5	Operating time: residents > fellows > attendings (39.84 > 19.92 > 13.32 min) OSATS scores: residents < fellows < attendings (19.8 < 26.8 < 29.2)	Self-evaluation self-confidence improvement: 4.33/5 Recommendation: 100%
Shen et al. [50], 2023	5 attendings 5 fellows 5 residents	Attendings group, fellows group, residents group	Laparoscopic bilioenteric anastomosis	Overall realism: 4.70/5 Tactile: 4.9/5 Breakthrough sensation: 4.60/5 Tearabilty: 4.3/5	N/A	Operating time: residents > fellows > attendings OSATS score: residents < fellows < attendings	N/A
Lu et al. [62], 2023	A novice doctor	No	Laparoscopic liver tumor excision	N/A	N/A	N/A	N/A
Shen et al. [40], 2024	15 surgeons	3DP group, simple suture group, video group	Laparoscopic bilioenteric anastomosis	N/A	N/A	Operation time: 3DP < simple suture < video (29.4 < 42.4 < 45.4 min) Operation score: 3DP > simple suture > video (8.2 > 6.8 > 5.2) GOALS score: 3DP > simple suture > video (21.2 > 17.8 > 15.6)	N/A
Elisei et al. [63], 2024	33 residents and specialist surgeons	No	Diagnosis, biopsy, drainage	Overall realism: 31/33	> 4.5/5	N/A	Durability: >90%
Gu et al. [51], 2024	4 residents 4 senior surgeons 4 experts	Experts group, senior surgeon group, residents group	Laparoscopic and robotic bilioenteric anastomosis	Overall realism: 4.5/5	4.70/5	Operation time: residents:59.3→34.8 min, senior surgeons: 38.8→24.5 min, experts: 27.5→17.6 min OSATS score: residents: 2.1→2.8, senior surgeons:3.3→3.8, experts: 4.3→4.6 min	Operability: 4.43/5
Yang et al. [41], 2024	20 residents	3DP group, TLCST group	Laparoscopic bile duct exploration	Overall realism: 8/10 Tissue pliability: 6/10 Texture: 6/10	4.5/5	Operation time: 3DP < TLCST (20.8 < 25.8 min) Operation scores: 3DP > TLCST (28.8 > 19.7)	Teaching satisfaction: 3DP > TLCST
Aranovich et al. [64], 2024	10 residents	No	Hepatic packing	N/A	N/A	Completion rate: 90%	N/A
Cao et al. [65], 2024	2 experts	No	Complex liver cancer resection	N/A	N/A	N/A	N/A

Notes:2D: two-dimensional; 3DP: three-dimensional printing; OSATS: objective structured assessment of technical skills; NASA-TLX: NASA Task Load Index; KAP: Kerma Area Product; TLCST: traditional laparoscopic simulation training group; N/A: not available.

Anatomical education

A total of 12 studies involved the application of the 3DP model in anatomy education (Table 3). Most compared 3DP with CT/MRI/atlas-based teaching (58.3%), and 25% compared 3DP with VR models. Medical students trained with 3DP models demonstrated 57–70% improvements in liver segment memorization compared with CT/MRI/atlas. Among residents, 3DP training reduced tumour localization time by 28-67% and improved localization scores by 28–135%. Representative models for surgical training and anatomical education, as well as the preparation workflow, are shown in Figure 2.

Table 3.

Application of 3DP models in anatomy education.

Authors	Participants	Allocation	Task	Fidelity	Self-evaluation usefulness	Validity	Others
Kong et al. [42], 2016	61 medical students	3DP group, digital models group, atlas group	Liver segment memory	Overall realism: 4/5	N/A	Teaching effect: 3DP > atlas	N/A
Kong et al. [43], 2016	92 medical students	Type 1 group, Type 2 group, Type 3 group, atlas group	Liver segment memory	Anatomical accuracy: Type 3 > Type 2> Type 1	N/A	Examination sores: Type 3 > Type 1 > Type 2 > atlas	Overall satisfaction: Type 3 > Type 2 > Type 1
Streba et al. [21], 2018	12 residents 43 medical students	Residents group, medical students group	Tumor location identification	Overall realism: students: 4.13/5, residents: 4.11/5 Texture: students: 2.84/5, residents: 2.17/5	Students: 3.79/5 Residents: 4.17/5	Surgical planning improvement: students: 3.98/5, residents: 4.42/5 Knowledge improvement: students: 3.58/5, residents: 4.08/5	N/A
Yang et al. [44], 2019	3 residents	3DP group, VR group, CT group	Tumor location identification	N/A	N/A	Completion time: 3DP < VR < CT (93.42 < 223.12 < 286.10 s) Completion score: 3DP > VR > CT (80.92 > 55.25 > 34.50)	N/A
Chedid et al. [45], 2020	116 physicians	3DP group, CT group	Liver segment memory	N/A	N/A	Liver segment recognition scores: 3DP: 42→71; CT: 39→69	Preference: 3DP: 39.6%
Bati et al. [46], 2020	19 residents	No	Anatomical identification, diagnosis, and preoperative preparation	N/A	N/A	Anatomical positioning: 100% Surgical planning improvement: 100%	N/A
Huettl et al. [47], 2021	5 experts 5 fellows 10 residents 10 medical students	Experts group, fellows group, residents group, medical students group	Tumor location identification	N/A	3.67/5	Surgical planning improvement: 90% Completion time: experts < students, 3DP < VR < digital model (509 < 702 < 704 s) Accuracy rates: experts > students 3DP > VR > digital model	Preference: 3DP: experts: 60%, fellows: 60%, residents: 20%, students: 0
Lopez-Lopez et al. [52], 2021	75 medical students	3DP group, CT/MRI group, digital model group	Tumor location identification	Tumor-vascular branch distance was consistent with CT/MRI.	N/A	The average performance: 3DP > CT/MRI > digital model	N/A
Song et al. [54], 2023	3 staffs 3 residents	Staffs group, resident group	Tumor location identification	T stage was consistent with the pathology report.	Staff: 4.60/5 Resident: 4.33/5	Completion scores improvement: 3DP: staffs: 3.96→4.75, residents: 3.84→4.37	N/A
Cheng et al. [68], 2023	62 interns	3DP group, VR group, CT group	Tumor location identification	N/A	4.03/5	Completion score: 3DP > VR > CT (89.4 > 75.8 > 69.5)	Satisfaction: 86.2% Interest: 92.1%
Shahbaz et al. [56], 2023	26 surgeons	No	Anatomy of hepatic blood vessels and biliary tracts	N/A	N/A	Surgical planning improvement: 100% Training young surgeons: 84%	N/A
Bao et al. [53], 2023	8 interns 10 standardized training trainees 12 professional training trainees	Interns group, standardized training trainees group, professional training trainees group	Tumor location identification	N/A	3.65/5	Understanding key points: interns < standardized training trainees < professional training trainees (3DP: 23 < 33 < 36; 2D: 17 < 27 < 28)	N/A

Notes: 2D: two-dimensional; 3DP: three-dimensional printing; CT: computed tomography; MRI: magnetic resonance imaging; VR: visual reality; Type 1: liver segmental vascular models without parenchyma; Type 2: liver segmental vascular models with transparent parenchyma; Type 3: liver segmental vascular models with partitions; N/A: not available.

Figure 2.

Typical 3DP models and process of preparation The DICOM files obtained from CT/MRI will be converted into STL files and imported into modeling software. After modeling, the next step will be carried out according to the production plan: A. the 3D model would be sliced for 3DP and be printed; B. the 3D model would be printed to create a mold, followed by material pouring, ultimately resulting in the final model. CT: computed tomography; MRI: magnetic resonance imaging; DICOM: Digital Imaging and Communication on Medicine. a, b, c, d: Reprinted with permission under the open access [41,47, 49,51]; e, f: Reprinted with permission through Copyright Clearance Center’s RightsLink^Ⓡ service [43,54].

Discussion

This systematic review critically evaluates the role of 3DP models in hepatobiliary surgical training and anatomical education. Current evidence highlights their extensive use in endoscopic and laparoscopic skill acquisition, with broad consensus among both trainees and experts regarding their pedagogical value. In anatomical teaching, 3DP models surpass conventional atlas-based methods, particularly by enhancing spatial visualization of hepatic segmentation and tumour localization.

Our analysis of the technical workflow reveals that while CT and MRI are the predominant imaging sources, the journey from DICOM data to a functional educational model is non-trivial. Only a few [52,54] validated printed models against the source imaging data, while the majority required manual post-processing (e.g. edge smoothing) to achieve clinical or educational suitability. It highlights a significant challenge in automated segmentation for complex hepatobiliary structures. In general, training objectives dictated model requirements: anatomical precision was essential for tumour localization, whereas tactile fidelity [47] outweighed visual accuracy in basic skills such as suturing [61]. More advanced applications, such as tumour resection training, required not only anatomically precise replication but also the integration of functional features, including simulated blood circulation and fluorescent tumour labelling [65].

Current printing materials provide realistic haptic feedback for cutting and suturing, with self-healing properties enabling repeated practice. However, their mechanical and thermal responses remain inconsistent with those of living tissue, limiting their application in advanced procedures such as intraoperative bleeding control or bile leakage management. Animal models, by contrast, inherently reproduce such complications, underscoring a critical limitation in the translational capacity of current 3DP platforms.

A pivotal strength of 3DP, as evidenced in this review, is its performance relative to established teaching modalities. In anatomical education, 3DP models consistently outperformed CT, MRI, and atlas-based learning, particularly in spatial tasks like liver segment memorization and tumour localization [42,44]. 3DP offers superior retention of spatial knowledge through haptic-visual multisensory integration, enabling students and residents to observe and understand hepatobiliary anatomy from multiple perspectives, thereby deepening their understanding of anatomical structures. 3DP models allow for surgical procedure practice in simulated environments, such as laparoscopic surgery, choledochoscopy and tumour resection, helping residents become familiar with surgical procedures and techniques, thus demonstrating higher proficiency in actual surgeries. Additionally, these models provide a low-stakes platform for identifying and rectifying technical errors, effectively mitigating the risk of intraoperative mishaps and complications in clinical settings. When compared to VR, 3DP offers the distinct advantage of physicality, allowing for authentic instrument-tissue interaction which is crucial for practicing surgical manoeuvres like anastomosis. However, VR may surpass 3DP in visualizing internal structures non-destructively. This suggests that these technologies are not mutually exclusive but potentially complementary, and future research should explore hybrid simulation strategies.

For 3DP to transition from a novel tool to a cornerstone of hepatobiliary training, its integration into standardized curricula is essential. Isolated training sessions, as seen in many included studies, provide proof of concept but may not maximize long-term skill retention. The technology’s greatest potential may lie in structured, progressive training pathways – from basic anatomy recognition on low-fidelity models to advanced resection planning on high-fidelity, patient-specific simulators. A significant barrier to this integration is the current ambiguity surrounding cost-effectiveness. The frequent omission of detailed cost analyses in the literature is a major limitation. While the initial investment in printers and materials is quantifiable, a comprehensive economic evaluation must weigh these against the potential benefits: reduced operative time, decreased error rates in the operating room, diminished reliance on cadavers and animal models, and, ultimately, improved patient outcomes.

Despite encouraging results, our review identifies several translational gaps. Evidence remains constrained by small sample sizes, short training exposures and limited longitudinal data on model durability. Cost analyses are inconsistent, and the frequent omission of economic considerations prevents comprehensive evaluation. Assessments are predominantly based on subjective expert ratings, with limited use of blinded designs or objective outcome measures such as blood loss or iatrogenic injury. To address these limitations, future research must prioritize well-designed, multi-institutional randomized controlled trials (RCTs) with blinding where possible. These trials should employ standardized, objective assessment tools, including AI-driven performance analytics [69]. Beyond validation, the technological frontier is rapidly advancing. The emergence of bioprinting, multi-material composites (e.g. ACEG-x resin) and self-healing hydrogels [62] promises to better replicate parenchymal biomechanics and enhance model durability. Finally, developing standardized manufacturing and validation protocols will be crucial for ensuring consistency and reliability across different training centres.

Limitations

This review has several limitations. The included studies showed methodological heterogeneity, generally low quality, and often lacked blinding. Most trials were small and qualitative, and cost data were frequently omitted. In addition, the included studies varied in design (e.g. RCTs vs. single-group cross-sectional), which may have influenced the interpretation of results. Moreover, differences in objectives and material choices meant that 3DP applications could not be meaningfully compared across studies, further limiting synthesis. Well-designed, adequately powered RCTs with standardized outcomes are needed to establish the effectiveness of 3DP in hepatobiliary surgical education.

Conclusion

This review consolidates compelling evidence for 3DP as a transformative tool in hepatobiliary surgical education, significantly enhancing both anatomical understanding and technical skill acquisition. It provides a safe, reproducible platform for complex procedural rehearsal, bridging a critical gap in contemporary training constrained by work-hour limits and ethical concerns. Further studies should prioritize multi-centre trials with standardized assessment tools and establish high fidelity, cost-effectiveness 3DP models for advanced hepatobiliary surgical procedures curricula development.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Funding Statement

This work was supported by hospital-enterprise joint project of Zhujiang Hospital, Southern Medical University (LCYJ240304).

Disclosure statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The authors report no financial or personal conflicts of interest.

Data availability statement

The data that support the findings of this study are available from the corresponding author, Jing Tian, upon reasonable request.